Enable k2v feature flag in CI

Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/380

.drone.yml

@@ -2,68 +2,26 @@
 kind: pipeline
 name: default
 
-workspace:
-  base: /drone/garage
-
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
-- name: nix_config
-  temp: {}
-
-environment:
-  HOME: /drone/garage
+node:
+  nix-daemon: 1
 
 steps:
-  - name: setup nix
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+  - name: check formatting
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - cp nix/nix.conf /etc/nix/nix.conf
-      - nix-build --no-build-output --no-out-link shell.nix --arg release false  -A inputDerivation
-
-  - name: code quality
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
-    commands:
-      - nix-shell --arg release false --run "cargo fmt -- --check"
-      - nix-shell --arg release false --run "cargo clippy -- --deny warnings"
+      - nix-shell --attr rust --run "cargo fmt -- --check"
 
   - name: build
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --option log-lines 100 --argstr target x86_64-unknown-linux-musl --arg release false --argstr git_version $DRONE_COMMIT
+      - nix-build --no-build-output --attr clippy.amd64 --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
 
   - name: unit + func tests
-    image: nixpkgs/nix:nixos-21.05
+    image: nixpkgs/nix:nixos-22.05
     environment:
       GARAGE_TEST_INTEGRATION_EXE: result/bin/garage
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
     commands:
-      - |
-        nix-build \
-          --no-build-output \
-          --option log-lines 100 \
-          --argstr target x86_64-unknown-linux-musl \
-          --argstr compileMode test
+      - nix-build --no-build-output --attr test.amd64
       - ./result/bin/garage_api-*
       - ./result/bin/garage_model-*
       - ./result/bin/garage_rpc-*
@@ -73,16 +31,11 @@ steps:
       - ./result/bin/garage-*
       - ./result/bin/integration-*
 
-  - name: smoke-test
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+  - name: integration tests
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --argstr target x86_64-unknown-linux-musl --arg release false --argstr git_version $DRONE_COMMIT
-      - nix-shell --arg release false --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
+      - nix-build --no-build-output --attr clippy.amd64 --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
+      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
 
 trigger:
   event:
@@ -92,78 +45,39 @@ trigger:
   - tag
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: pipeline
 type: docker
-name: release-linux-x86_64
+name: release-linux-amd64
 
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
-- name: nix_config
-  temp: {}
-
-environment:
-  TARGET: x86_64-unknown-linux-musl
+node:
+  nix-daemon: 1
 
 steps:
-  - name: setup nix
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
-    commands:
-      - cp nix/nix.conf /etc/nix/nix.conf
-      - nix-build --no-build-output --no-out-link shell.nix -A inputDerivation
-
   - name: build
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --argstr target $TARGET --arg release true --argstr git_version $DRONE_COMMIT
+      - nix-build --no-build-output --attr pkgs.amd64.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
+      - nix-shell --attr rust --run "./script/not-dynamic.sh result/bin/garage"
 
   - name: integration
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-shell --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
+      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
 
   - name: push static binary
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
       AWS_SECRET_ACCESS_KEY:
         from_secret: garagehq_aws_secret_access_key
+      TARGET: "x86_64-unknown-linux-musl"
     commands:
-      - nix-shell --arg rust false --arg integration false --run "to_s3"
+      - nix-shell --attr release --run "to_s3"
 
   - name: docker build and publish
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_AUTH:
         from_secret: docker_auth
@@ -174,7 +88,7 @@ steps:
       - mkdir -p /kaniko/.docker
       - echo $DOCKER_AUTH > /kaniko/.docker/config.json
       - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --arg rust false --arg integration false --run "to_docker"
+      - nix-shell --attr release --run "to_docker"
 
 
 trigger:
@@ -182,78 +96,39 @@ trigger:
   - promote
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: pipeline
 type: docker
-name: release-linux-i686
+name: release-linux-i386
 
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
-- name: nix_config
-  temp: {}
-
-environment:
-  TARGET: i686-unknown-linux-musl
+node:
+  nix-daemon: 1
 
 steps:
-  - name: setup nix
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
-    commands:
-      - cp nix/nix.conf /etc/nix/nix.conf
-      - nix-build --no-build-output --no-out-link shell.nix -A inputDerivation
-
   - name: build
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --argstr target $TARGET --arg release true --argstr git_version $DRONE_COMMIT
+      - nix-build --no-build-output --attr pkgs.i386.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
+      - nix-shell --attr rust --run "./script/not-dynamic.sh result/bin/garage"
 
   - name: integration
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-shell --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
+      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
 
   - name: push static binary
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
       AWS_SECRET_ACCESS_KEY:
         from_secret: garagehq_aws_secret_access_key
+      TARGET: "i686-unknown-linux-musl"
     commands:
-      - nix-shell --arg rust false --arg integration false --run "to_s3"
+      - nix-shell --attr release --run "to_s3"
 
   - name: docker build and publish
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_AUTH:
         from_secret: docker_auth
@@ -264,75 +139,41 @@ steps:
       - mkdir -p /kaniko/.docker
       - echo $DOCKER_AUTH > /kaniko/.docker/config.json
       - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --arg rust false --arg integration false --run "to_docker"
+      - nix-shell --attr release --run "to_docker"
 
 trigger:
   event:
   - promote
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: pipeline
 type: docker
-name: release-linux-aarch64
+name: release-linux-arm64
 
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
-- name: nix_config
-  temp: {}
-
-environment:
-  TARGET: aarch64-unknown-linux-musl
+node:
+  nix-daemon: 1
 
 steps:
-  - name: setup nix
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
-    commands:
-      - cp nix/nix.conf /etc/nix/nix.conf
-      - nix-build --no-build-output --no-out-link ./shell.nix --arg rust false --arg integration false -A inputDerivation
-
   - name: build
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --argstr target $TARGET --arg release true --argstr git_version $DRONE_COMMIT
+      - nix-build --no-build-output --attr pkgs.arm64.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
+      - nix-shell --attr rust --run "./script/not-dynamic.sh result/bin/garage"
 
   - name: push static binary
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
       AWS_SECRET_ACCESS_KEY:
         from_secret: garagehq_aws_secret_access_key
+      TARGET: "aarch64-unknown-linux-musl"
     commands:
-      - nix-shell --arg rust false --arg integration false --run "to_s3"
+      - nix-shell --attr release --run "to_s3"
 
   - name: docker build and publish
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_AUTH:
         from_secret: docker_auth
@@ -343,75 +184,41 @@ steps:
       - mkdir -p /kaniko/.docker
       - echo $DOCKER_AUTH > /kaniko/.docker/config.json
       - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --arg rust false --arg integration false --run "to_docker"
+      - nix-shell --attr release --run "to_docker"
 
 trigger:
   event:
   - promote
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: pipeline
 type: docker
-name: release-linux-armv6l
+name: release-linux-arm
 
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
-- name: nix_config
-  temp: {}
-
-environment:
-  TARGET: armv6l-unknown-linux-musleabihf
+node:
+  nix-daemon: 1
 
 steps:
-  - name: setup nix
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
-    commands:
-      - cp nix/nix.conf /etc/nix/nix.conf
-      - nix-build --no-build-output --no-out-link --arg rust false --arg integration false -A inputDerivation
-
   - name: build
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     commands:
-      - nix-build --no-build-output --argstr target $TARGET --arg release true --argstr git_version $DRONE_COMMIT
+      - nix-build --no-build-output --attr pkgs.arm.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
+      - nix-shell --attr rust --run "./script/not-dynamic.sh result/bin/garage"
 
   - name: push static binary
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
       AWS_SECRET_ACCESS_KEY:
         from_secret: garagehq_aws_secret_access_key
+      TARGET: "armv6l-unknown-linux-musleabihf"
     commands:
-      - nix-shell --arg integration false --arg rust false --run "to_s3"
+      - nix-shell --attr release --run "to_s3"
 
   - name: docker build and publish
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
-    - name: nix_config
-      path: /etc/nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       DOCKER_AUTH:
         from_secret: docker_auth
@@ -422,32 +229,24 @@ steps:
       - mkdir -p /kaniko/.docker
       - echo $DOCKER_AUTH > /kaniko/.docker/config.json
       - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --arg rust false --arg integration false --run "to_docker"
+      - nix-shell --attr release --run "to_docker"
 
 trigger:
   event:
   - promote
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: pipeline
 type: docker
 name: refresh-release-page
 
-volumes:
-- name: nix_store
-  host:
-    path: /var/lib/drone/nix
+node:
+  nix-daemon: 1
 
 steps:
   - name: refresh-index
-    image: nixpkgs/nix:nixos-21.05
-    volumes:
-    - name: nix_store
-      path: /nix
+    image: nixpkgs/nix:nixos-22.05
     environment:
       AWS_ACCESS_KEY_ID:
         from_secret: garagehq_aws_access_key_id
@@ -455,24 +254,21 @@ steps:
         from_secret: garagehq_aws_secret_access_key
     commands:
       - mkdir -p /etc/nix && cp nix/nix.conf /etc/nix/nix.conf
-      - nix-shell --arg integration false --arg rust false --run "refresh_index"
+      - nix-shell --attr release --run "refresh_index"
 
 depends_on:
-  - release-linux-x86_64
-  - release-linux-i686
-  - release-linux-aarch64
-  - release-linux-armv6l
+  - release-linux-amd64
+  - release-linux-i386
+  - release-linux-arm64
+  - release-linux-arm
 
 trigger:
   event:
   - promote
   - cron
 
-node:
-  nix: 1
-
 ---
 kind: signature
-hmac: 3fc19d6f9a3555519c8405e3281b2e74289bb802f644740d5481d53df3a01fa4
+hmac: 362639b4c9541ad9bd06ff7f72b5235b2b0216bcb16eececd25285b6fe94ba6f
 
 ...

Cargo.toml

@@ -1,5 +1,7 @@
 [workspace]
+resolver = "2"
 members = [
+	"src/db",
 	"src/util",
 	"src/rpc",
 	"src/table",

Makefile

@@ -1,13 +1,27 @@
-.PHONY: doc all release shell
+.PHONY: doc all release shell run1 run2 run3
 
 all:
-	clear; cargo build --all-features
-
-doc:
-	cd doc/book; mdbook build
+	clear; cargo build
 
 release:
 	nix-build --arg release true
 
 shell:
 	nix-shell
+
+# ----
+
+run1:
+	RUST_LOG=garage=debug ./target/debug/garage -c tmp/config1.toml server
+run1rel:
+	RUST_LOG=garage=debug ./target/release/garage -c tmp/config1.toml server
+
+run2:
+	RUST_LOG=garage=debug ./target/debug/garage -c tmp/config2.toml server
+run2rel:
+	RUST_LOG=garage=debug ./target/release/garage -c tmp/config2.toml server
+
+run3:
+	RUST_LOG=garage=debug ./target/debug/garage -c tmp/config3.toml server
+run3rel:
+	RUST_LOG=garage=debug ./target/release/garage -c tmp/config3.toml server

default.nix

@@ -1,147 +1,33 @@
 {
   system ? builtins.currentSystem,
-  release ? false,
-  target ? "x86_64-unknown-linux-musl",
-  compileMode ? null,
   git_version ? null,
 }:
 
 with import ./nix/common.nix;
 
-let
-  crossSystem = { config = target; };
-in let
-  log = v: builtins.trace v v;
+let 
+  pkgs = import pkgsSrc { };
+  compile = import ./nix/compile.nix;
+  build_debug_and_release = (target: {
+    debug = (compile { inherit target git_version; release = false; }).workspace.garage { compileMode = "build"; }; 
+    release = (compile { inherit target git_version; release = true; }).workspace.garage { compileMode = "build"; }; 
+  });
+  test = (rustPkgs: pkgs.symlinkJoin {
+    name ="garage-tests";
+    paths = builtins.map (key: rustPkgs.workspace.${key} { compileMode = "test"; }) (builtins.attrNames rustPkgs.workspace);
+  });
 
-  pkgs = import pkgsSrc {
-    inherit system crossSystem;
-    overlays = [ cargo2nixOverlay ];
+in {
+  pkgs = {
+    amd64 = build_debug_and_release "x86_64-unknown-linux-musl";
+    i386 = build_debug_and_release "i686-unknown-linux-musl";
+    arm64 = build_debug_and_release "aarch64-unknown-linux-musl";
+    arm = build_debug_and_release "armv6l-unknown-linux-musleabihf";
   };
-
-
-  /*
-   Rust and Nix triples are not the same. Cargo2nix has a dedicated library
-   to convert Nix triples to Rust ones. We need this conversion as we want to
-   set later options linked to our (rust) target in a generic way. Not only
-   the triple terminology is different, but also the "roles" are named differently.
-   Nix uses a build/host/target terminology where Nix's "host" maps to Cargo's "target".
-  */
-  rustTarget = log (pkgs.rustBuilder.rustLib.rustTriple pkgs.stdenv.hostPlatform);
-
-  /*
-   Cargo2nix is built for rustOverlay which installs Rust from Mozilla releases.
-   We want our own Rust to avoid incompatibilities, like we had with musl 1.2.0.
-   rustc was built with musl < 1.2.0 and nix shipped musl >= 1.2.0 which lead to compilation breakage.
-   So we want a Rust release that is bound to our Nix repository to avoid these problems.
-   See here for more info: https://musl.libc.org/time64.html
-   Because Cargo2nix does not support the Rust environment shipped by NixOS,
-   we emulate the structure of the Rust object created by rustOverlay.
-   In practise, rustOverlay ships rustc+cargo in a single derivation while
-   NixOS ships them in separate ones. We reunite them with symlinkJoin.
-   */
-  rustChannel = pkgs.symlinkJoin {
-    name ="rust-channel";
-    paths = [ 
-      pkgs.rustPlatform.rust.rustc
-      pkgs.rustPlatform.rust.cargo
-    ];
+  test = {
+    amd64 = test (compile { inherit git_version; target = "x86_64-unknown-linux-musl"; });
   };
-
-  /*
-   Cargo2nix provides many overrides by default, you can take inspiration from them:
-   https://github.com/cargo2nix/cargo2nix/blob/master/overlay/overrides.nix
-
-   You can have a complete list of the available options by looking at the overriden object, mkcrate:
-   https://github.com/cargo2nix/cargo2nix/blob/master/overlay/mkcrate.nix
-  */
-  overrides = pkgs.rustBuilder.overrides.all ++ [
-    /*
-     [1] We need to alter Nix hardening to be able to statically compile: PIE,
-     Position Independent Executables seems to be supported only on amd64. Having
-     this flags set either make our executables crash or compile as dynamic on many platforms.
-     In the following section codegenOpts, we reactive it for the supported targets
-     (only amd64 curently) through the `-static-pie` flag. PIE is a feature used
-     by ASLR, which helps mitigate security issues.
-     Learn more about Nix Hardening: https://github.com/NixOS/nixpkgs/blob/master/pkgs/build-support/cc-wrapper/add-hardening.sh
-
-     [2] We want to inject the git version while keeping the build deterministic.
-     As we do not want to consider the .git folder as part of the input source,
-     we ask the user (the CI often) to pass the value to Nix.
-    */
-    (pkgs.rustBuilder.rustLib.makeOverride {
-      name = "garage_rpc";
-      overrideAttrs = drv:
-        /* [1] */ { hardeningDisable = [ "pie" ]; }
-        //
-        /* [2] */ (if git_version != null then {
-          preConfigure = ''
-            ${drv.preConfigure or ""}
-            export GIT_VERSION="${git_version}"
-          '';
-        } else {});
-    })
-
-    /*
-     We ship some parts of the code disabled by default by putting them behind a flag.
-     It speeds up the compilation (when the feature is not required) and released crates have less dependency by default (less attack surface, disk space, etc.).
-     But we want to ship these additional features when we release Garage.
-     In the end, we chose to exclude all features from debug builds while putting (all of) them in the release builds.
-     Currently, the only feature of Garage is kubernetes-discovery from the garage_rpc crate.
-    */
-    (pkgs.rustBuilder.rustLib.makeOverride {
-      name = "garage_rpc";
-      overrideArgs = old:
-        {
-          features = if release then [ "kubernetes-discovery" ] else [];
-        };
-    })
-  ];
-
-  packageFun = import ./Cargo.nix;
-
-  /*
-    We compile fully static binaries with musl to simplify deployment on most systems.
-    When possible, we reactivate PIE hardening (see above).
-
-    Also, if you set the RUSTFLAGS environment variable, the following parameters will
-    be ignored.
-
-    For more information on static builds, please refer to Rust's RFC 1721.
-    https://rust-lang.github.io/rfcs/1721-crt-static.html#specifying-dynamicstatic-c-runtime-linkage
-  */
-
-  codegenOpts = {
-   "armv6l-unknown-linux-musleabihf" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* compile as dynamic with static-pie */
-   "aarch64-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* segfault with static-pie */
-   "i686-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* segfault with static-pie */
-   "x86_64-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static-pie" ];
+  clippy = {
+    amd64 = (compile { inherit git_version; compiler = "clippy"; }).workspace.garage { compileMode = "build"; } ;
   };
-
-  /*
-   The following definition is not elegant as we use a low level function of Cargo2nix
-   that enables us to pass our custom rustChannel object. We need this low level definition
-   to pass Nix's Rust toolchains instead of Mozilla's one.
-
-   target is mandatory but must be kept to null to allow cargo2nix to set it to the appropriate value
-   for each crate.
-  */ 
-  rustPkgs = pkgs.rustBuilder.makePackageSet {
-    inherit packageFun rustChannel release codegenOpts;
-    packageOverrides = overrides;
-    target = null;
-
-    buildRustPackages = pkgs.buildPackages.rustBuilder.makePackageSet {
-      inherit rustChannel packageFun codegenOpts;
-      packageOverrides = overrides;
-      target = null;
-    };
-  };
-
-
-in
-  if compileMode == "test"
-    then pkgs.symlinkJoin {
-      name ="garage-tests";
-      paths = builtins.map (key: rustPkgs.workspace.${key} { inherit compileMode; }) (builtins.attrNames rustPkgs.workspace);
-    }
-    else rustPkgs.workspace.garage { inherit compileMode; }
+}

doc/book/connect/websites.md

@@ -3,7 +3,7 @@ title = "Websites (Hugo, Jekyll, Publii...)"
 weight = 10
 +++
 
-Garage is also suitable to host static websites.
+Garage is also suitable [to host static websites](@/documentation/cookbook/exposing-websites.md).
 While they can be deployed with traditional CLI tools, some static website generators have integrated options to ease your workflow.
 
 | Name | Status | Note |

doc/book/cookbook/from-source.md

@@ -20,6 +20,24 @@ sudo apt-get update
 sudo apt-get install build-essential
 ```
 
+## Using source from the Gitea repository (recommended)
+
+The primary location for Garage's source code is the
+[Gitea repository](https://git.deuxfleurs.fr/Deuxfleurs/garage).
+
+Clone the repository and build Garage with the following commands:
+
+```bash
+git clone https://git.deuxfleurs.fr/Deuxfleurs/garage.git
+cd garage
+cargo build
+```
+
+Be careful, as this will make a debug build of Garage, which will be extremely slow!
+To make a release build, invoke `cargo build --release` (this takes much longer).
+
+The binaries built this way are found in `target/{debug,release}/garage`.
+
 ## Using source from `crates.io`
 
 Garage's source code is published on `crates.io`, Rust's official package repository.
@@ -39,21 +57,20 @@ sudo cp $HOME/.cargo/bin/garage /usr/local/bin/garage
 ```
 
 
-## Using source from the Gitea repository
+## Selecting features to activate in your build
 
-The primary location for Garage's source code is the
-[Gitea repository](https://git.deuxfleurs.fr/Deuxfleurs/garage).
-
-Clone the repository and build Garage with the following commands:
-
-```bash
-git clone https://git.deuxfleurs.fr/Deuxfleurs/garage.git
-cd garage
-cargo build
-```
-
-Be careful, as this will make a debug build of Garage, which will be extremely slow!
-To make a release build, invoke `cargo build --release` (this takes much longer).
-
-The binaries built this way are found in `target/{debug,release}/garage`.
+Garage supports a number of compilation options in the form of Cargo features,
+which can be used to provide builds adapted to your system and your use case.
+The following features are available:
 
+| Feature | Enabled | Description |
+| ------- | ------- | ----------- |
+| `bundled-libs` | BY DEFAULT | Use bundled version of sqlite3, zstd, lmdb and libsodium |
+| `system-libs` | optional | Use system version of sqlite3, zstd, lmdb and libsodium if available (exclusive with `bundled-libs`, build using `cargo build --no-default-features --features system-libs`) |
+| `k2v` | optional | Enable the experimental K2V API (if used, all nodes on your Garage cluster must have it enabled as well) |
+| `kubernetes-discovery` | optional | Enable automatic registration and discovery of cluster nodes through the Kubernetes API |
+| `metrics` | BY DEFAULT | Enable collection of metrics in Prometheus format on the admin API |
+| `telemetry-otlp` | optional | Enable collection of execution traces using OpenTelemetry |
+| `sled` | BY DEFAULT | Enable using Sled to store Garage's metadata |
+| `lmdb` | optional | Enable using LMDB to store Garage's metadata |
+| `sqlite` | optional | Enable using Sqlite3 to store Garage's metadata |

doc/book/cookbook/reverse-proxy.md

@@ -100,7 +100,7 @@ server {
 }
 ```
 
-## Exposing the web endpoint
+### Exposing the web endpoint
 
 To better understand the logic involved, you can refer to the [Exposing buckets as websites](/cookbook/exposing_websites.html) section.
 Otherwise, the configuration is very similar to the S3 endpoint.
@@ -140,6 +140,165 @@ server {
 
 @TODO
 
-## Traefik
+## Traefik v2
+
+We will see in this part how to set up a reverse proxy with [Traefik](https://docs.traefik.io/).
+
+Here is [a basic configuration file](https://doc.traefik.io/traefik/https/acme/#configuration-examples):
+
+```toml
+[entryPoints]
+  [entryPoints.web]
+    address = ":80"
+
+  [entryPoints.websecure]
+    address = ":443"
+
+[certificatesResolvers.myresolver.acme]
+  email = "your-email@example.com"
+  storage = "acme.json"
+  [certificatesResolvers.myresolver.acme.httpChallenge]
+    # used during the challenge
+    entryPoint = "web"
+```
+
+### Add Garage service
+
+To add Garage on Traefik you should declare a new service using its IP address (or hostname) and port:
+
+```toml
+[http.services]
+  [http.services.my_garage_service.loadBalancer]
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://xxx.xxx.xxx.xxx"
+      port = 3900
+```
+
+It's possible to declare multiple Garage servers as back-ends:
+
+```toml
+[http.services]
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://xxx.xxx.xxx.xxx"
+      port = 3900
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://yyy.yyy.yyy.yyy"
+      port = 3900
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://zzz.zzz.zzz.zzz"
+      port = 3900
+```
+
+Traefik can remove unhealthy servers automatically with [a health check configuration](https://doc.traefik.io/traefik/routing/services/#health-check):
+
+```
+[http.services]
+  [http.services.my_garage_service.loadBalancer]
+    [http.services.my_garage_service.loadBalancer.healthCheck]
+      path = "/"
+      interval = "60s"
+      timeout = "5s"
+```
+
+### Adding a website
+
+To add a new website, add the following declaration to your Traefik configuration file:
+
+```toml
+[http.routers]
+  [http.routers.my_website]
+    rule = "Host(`yoururl.example.org`)"
+    service = "my_garage_service"
+    entryPoints = ["web"]
+```
+
+Enable HTTPS access to your website with the following configuration section ([documentation](https://doc.traefik.io/traefik/https/overview/)):
+
+```toml
+...
+    entryPoints = ["websecure"]
+    [http.routers.my_website.tls]
+      certResolver = "myresolver"
+...
+```
+
+### Adding gzip compression
+
+Add the following configuration section [to compress response](https://doc.traefik.io/traefik/middlewares/http/compress/) using [gzip](https://developer.mozilla.org/en-US/docs/Glossary/GZip_compression) before sending them to the client:
+
+```toml
+[http.routers]
+  [http.routers.my_website]
+    ...
+    middlewares = ["gzip_compress"]
+    ...
+[http.middlewares]
+  [http.middlewares.gzip_compress.compress]
+```
+
+### Add caching response
+
+Traefik's caching middleware is only available on [entreprise version](https://doc.traefik.io/traefik-enterprise/middlewares/http-cache/), however the freely-available [Souin plugin](https://github.com/darkweak/souin#tr%C3%A6fik-container) can also do the job. (section to be completed)
+
+### Complete example
+
+```toml
+[entryPoints]
+  [entryPoints.web]
+    address = ":80"
+
+  [entryPoints.websecure]
+    address = ":443"
+
+[certificatesResolvers.myresolver.acme]
+  email = "your-email@example.com"
+  storage = "acme.json"
+  [certificatesResolvers.myresolver.acme.httpChallenge]
+    # used during the challenge
+    entryPoint = "web"
+
+[http.routers]
+  [http.routers.my_website]
+    rule = "Host(`yoururl.example.org`)"
+    service = "my_garage_service"
+    middlewares = ["gzip_compress"]
+    entryPoints = ["websecure"]
+
+[http.services]
+  [http.services.my_garage_service.loadBalancer]
+    [http.services.my_garage_service.loadBalancer.healthCheck]
+      path = "/"
+      interval = "60s"
+      timeout = "5s"
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://xxx.xxx.xxx.xxx"
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://yyy.yyy.yyy.yyy"
+    [[http.services.my_garage_service.loadBalancer.servers]]
+      url = "http://zzz.zzz.zzz.zzz"
+
+[http.middlewares]
+  [http.middlewares.gzip_compress.compress]
+```
+
+## Caddy
+
+Your Caddy configuration can be as simple as:
+
+```caddy
+s3.garage.tld, *.s3.garage.tld {
+	reverse_proxy localhost:3900 192.168.1.2:3900 example.tld:3900
+}
+
+*.web.garage.tld {
+	reverse_proxy localhost:3902 192.168.1.2:3900 example.tld:3900
+}
+
+admin.garage.tld {
+	reverse_proxy localhost:3903
+}
+```
+
+But at the same time, the `reverse_proxy` is very flexible.
+For a production deployment, you should [read its documentation](https://caddyserver.com/docs/caddyfile/directives/reverse_proxy) as it supports features like DNS discovery of upstreams, load balancing with checks, streaming parameters, etc.
 
-@TODO

doc/book/quick-start/_index.md

@@ -249,16 +249,6 @@ mc alias set \
   --api S3v4
 ```
 
-You must also add an environment variable to your configuration to
-inform MinIO of our region (`garage` by default, corresponding to the `s3_region` parameter
-in the configuration file).
-The best way is to add the following snippet to your `$HOME/.bash_profile`
-or `$HOME/.bashrc` file:
-
-```bash
-export MC_REGION=garage
-```
-
 ### Use `mc`
 
 You can not list buckets from `mc` currently.

doc/book/reference-manual/admin-api.md

@@ -134,8 +134,8 @@ Example request body:
 
 ```json
 [
-	"ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f@10.0.0.11:3901",
-	"4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff@10.0.0.12:3901"
+    "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f@10.0.0.11:3901",
+    "4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff@10.0.0.12:3901"
 ]
 ```
 
@@ -145,14 +145,14 @@ Example response:
 
 ```json
 [
-	{
-		"success": true,
-		"error": null
-	},
-	{
-		"success": false,
-		"error": "Handshake error"
-	}
+    {
+        "success": true,
+        "error": null
+    },
+    {
+        "success": false,
+        "error": "Handshake error"
+    }
 ]
 ```
 
@@ -301,7 +301,7 @@ Request body format:
 
 ```json
 {
-	"name": "NameOfMyKey"
+    "name": "NameOfMyKey"
 }
 ```
 
@@ -313,9 +313,9 @@ Request body format:
 
 ```json
 {
-	"accessKeyId": "GK31c2f218a2e44f485b94239e",
-	"secretAccessKey": "b892c0665f0ada8a4755dae98baa3b133590e11dae3bcc1f9d769d67f16c3835",
-	"name": "NameOfMyKey"
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "secretAccessKey": "b892c0665f0ada8a4755dae98baa3b133590e11dae3bcc1f9d769d67f16c3835",
+    "name": "NameOfMyKey"
 }
 ```
 
@@ -403,11 +403,11 @@ Request body format:
 
 ```json
 {
-	"name": "NameOfMyKey",
-	"allow": {
-		"createBucket": true,
-	},
-	"deny": {}
+    "name": "NameOfMyKey",
+    "allow": {
+        "createBucket": true,
+    },
+    "deny": {}
 }
 ```
 
@@ -473,24 +473,31 @@ Example response:
 
 ```json
 {
-  "id": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
-  "globalAliases": [
-    "alex"
-  ],
-  "keys": [
-    {
-      "accessKeyId": "GK31c2f218a2e44f485b94239e",
-      "name": "alex",
-      "permissions": {
-        "read": true,
-        "write": true,
-        "owner": true
-      },
-      "bucketLocalAliases": [
-        "test"
-      ]
-    }
-  ]
+    "id": "afa8f0a22b40b1247ccd0affb869b0af5cff980924a20e4b5e0720a44deb8d39",
+        "globalAliases": [],
+        "websiteAccess": false,
+        "websiteConfig": null,
+        "keys": [
+        {
+            "accessKeyId": "GK31c2f218a2e44f485b94239e",
+            "name": "Imported key",
+            "permissions": {
+                "read": true,
+                "write": true,
+                "owner": true
+            },
+            "bucketLocalAliases": [
+                "debug"
+            ]
+        }
+        ],
+        "objects": 14827,
+        "bytes": 13189855625,
+        "unfinshedUploads": 0,
+        "quotas": {
+            "maxSize": null,
+            "maxObjects": null
+        }
 }
 ```
 
@@ -502,7 +509,7 @@ Request body format:
 
 ```json
 {
-	"globalAlias": "NameOfMyBucket"
+    "globalAlias": "NameOfMyBucket"
 }
 ```
 
@@ -510,15 +517,15 @@ OR
 
 ```json
 {
-	"localAlias": {
-		"accessKeyId": "GK31c2f218a2e44f485b94239e",
-		"alias": "NameOfMyBucket",
-		"allow": {
-			"read": true,
-			"write": true,
-			"owner": false
-		}
-	}
+    "localAlias": {
+        "accessKeyId": "GK31c2f218a2e44f485b94239e",
+        "alias": "NameOfMyBucket",
+        "allow": {
+            "read": true,
+            "write": true,
+            "owner": false
+        }
+    }
 }
 ```
 
@@ -540,26 +547,37 @@ Deletes a storage bucket. A bucket cannot be deleted if it is not empty.
 
 Warning: this will delete all aliases associated with the bucket!
 
-#### PutBucketWebsite `PUT /v0/bucket/website?id=<bucket id>`
+#### UpdateBucket `PUT /v0/bucket?id=<bucket id>`
 
-Sets the website configuration for a bucket (this also enables website access for this bucket).
+Updates configuration of the given bucket.
 
 Request body format:
 
 ```json
 {
-	"indexDocument": "index.html",
-	"errorDocument": "404.html"
+    "websiteAccess": {
+        "enabled": true,
+        "indexDocument": "index.html",
+        "errorDocument": "404.html"
+    },
+    "quotas": {
+        "maxSize": 19029801,
+        "maxObjects": null,
+    }
 }
 ```
 
-The field `errorDocument` is optional, if no error document is set a generic error message is displayed when errors happen.
+All fields (`websiteAccess` and `quotas`) are optionnal.
+If they are present, the corresponding modifications are applied to the bucket, otherwise nothing is changed.
 
+In `websiteAccess`: if `enabled` is `true`, `indexDocument` must be specified.
+The field `errorDocument` is optional, if no error document is set a generic
+error message is displayed when errors happen. Conversely, if `enabled` is
+`false`, neither `indexDocument` nor `errorDocument` must be specified.
 
-#### DeleteBucketWebsite `DELETE /v0/bucket/website?id=<bucket id>`
-
-Deletes the website configuration for a bucket (disables website access for this bucket).
-
+In `quotas`: new values of `maxSize` and `maxObjects` must both be specified, or set to `null`
+to remove the quotas. An absent value will be considered the same as a `null`. It is not possible
+to change only one of the two quotas.
 
 ### Operations on permissions for keys on buckets
 
@@ -571,13 +589,13 @@ Request body format:
 
 ```json
 {
-	"bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
-	"accessKeyId": "GK31c2f218a2e44f485b94239e",
-	"permissions": {
-		"read": true,
-		"write": true,
-		"owner": true
-	},
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "permissions": {
+        "read": true,
+        "write": true,
+        "owner": true
+    },
 }
 ```
 
@@ -592,13 +610,13 @@ Request body format:
 
 ```json
 {
-	"bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
-	"accessKeyId": "GK31c2f218a2e44f485b94239e",
-	"permissions": {
-		"read": false,
-		"write": false,
-		"owner": true
-	},
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "permissions": {
+        "read": false,
+        "write": false,
+        "owner": true
+    },
 }
 ```
 

doc/book/reference-manual/s3-compatibility.md

@@ -148,7 +148,7 @@ Please open an issue if you have a use case for replication.
 | [PutBucketReplication](https://docs.aws.amazon.com/AmazonS3/latest/API/API_PutBucketReplication.html) | ❌ Missing | ❌| ⚠ | ❌| ❌|
 
 *Note: Ceph documentation briefly says that Ceph supports
-[replication though the S3 API](https://docs.ceph.com/en/latest/radosgw/multisite-sync-policy/#s3-replication-api)
+[replication through the S3 API](https://docs.ceph.com/en/latest/radosgw/multisite-sync-policy/#s3-replication-api)
 but with some limitations.
 Additionaly, replication endpoints are not documented in the S3 compatibility page so I don't know what kind of support we can expect.*
 

doc/talks/2022-06-23-stack/.gitignore

@@ -0,0 +1,14 @@
+*
+
+!assets
+
+!.gitignore
+!*.svg
+!*.png
+!*.jpg
+!*.tex
+!Makefile
+!.gitignore
+!assets/*.drawio.pdf
+
+!talk.pdf

doc/talks/2022-06-23-stack/Makefile

@@ -0,0 +1,5 @@
+talk.pdf: talk.tex assets/consistent_hashing_1.pdf assets/consistent_hashing_2.pdf assets/consistent_hashing_3.pdf assets/consistent_hashing_4.pdf assets/garage_tables.pdf assets/deuxfleurs.pdf
+	pdflatex talk.tex
+
+assets/%.pdf: assets/%.svg
+	inkscape -D -z --file=$^ --export-pdf=$@

doc/talks/2022-06-23-stack/assets/aerogramme_components1.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9198d0cfc0e04a56f84a353dd660b294ecf0dc68fe429892c8e9e604de016748
+size 31966

doc/talks/2022-06-23-stack/assets/aerogramme_components2.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14dbf4a45545889babb2fdc6580399811366f6cc786cb28e3467cbcedbfe9482
+size 31688

doc/talks/2022-06-23-stack/assets/aerogramme_datatype.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21a3bb004ad35c8b47f14195c720048f8db31f47bd6346c1d747000570149c67
+size 31073

doc/talks/2022-06-23-stack/assets/aerogramme_keys.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4e109078a784c3999ca2ccd1adda8ff5793d1b230a9e27b4ad2e419afa0d37d
+size 25145

doc/talks/2022-06-23-stack/assets/deuxfleurs.svg

@@ -0,0 +1,91 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   viewBox="0 0 70.424515 70.300102"
+   version="1.1"
+   id="svg8"
+   sodipodi:docname="logo.svg"
+   inkscape:version="1.1 (c68e22c387, 2021-05-23)"
+   inkscape:export-filename="/home/quentin/Documents/dev/deuxfleurs/site/src/img/logo.png"
+   inkscape:export-xdpi="699.30194"
+   inkscape:export-ydpi="699.30194"
+   width="70.424515"
+   height="70.300102"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg">
+  <defs
+     id="defs12" />
+  <sodipodi:namedview
+     id="namedview10"
+     pagecolor="#ffffff"
+     bordercolor="#666666"
+     borderopacity="1.0"
+     inkscape:pageshadow="2"
+     inkscape:pageopacity="0.0"
+     inkscape:pagecheckerboard="0"
+     showgrid="false"
+     inkscape:zoom="12.125"
+     inkscape:cx="43.092783"
+     inkscape:cy="48.082474"
+     inkscape:window-width="3072"
+     inkscape:window-height="1659"
+     inkscape:window-x="0"
+     inkscape:window-y="0"
+     inkscape:window-maximized="1"
+     inkscape:current-layer="svg8" />
+  <g
+     id="g79969"
+     transform="translate(-0.827,34.992103)">
+    <path
+       fill="#ffffff"
+       d="m 15.632,34.661 c -0.799,-0.597 -1.498,-1.484 -2.035,-2.592 l -0.228,-0.47 -0.46,0.249 c -0.975,0.528 -1.913,0.858 -2.744,0.969 L 9.963,29.061 6.327,30.029 C 6.17,29.175 6.202,28.142 6.423,27.007 L 6.526,26.482 5.994,26.416 C 4.752,26.262 3.688,25.891 2.89,25.336 L 4.411,22.419 1.423,20.896 C 1.742,19.952 2.371,19.014 3.257,18.161 L 3.634,17.798 3.255,17.438 C 2.452,16.674 1.847,15.884 1.485,15.127 L 4.995,13.774 2.95,10.615 C 3.69,10.213 4.643,9.929 5.739,9.783 L 6.258,9.715 6.167,9.201 C 5.952,7.99 5.995,6.863 6.291,5.913 l 3.308,0.523 0.524,-3.308 c 0.988,0.013 2.08,0.326 3.164,0.907 L 13.749,4.283 13.975,3.81 C 14.454,2.807 15.019,1.986 15.628,1.406 L 18,4.326 20.372,1.406 c 0.609,0.58 1.175,1.401 1.653,2.404 l 0.226,0.473 0.462,-0.247 C 23.798,3.455 24.891,3.142 25.877,3.13 L 26.4,6.438 29.71,5.913 c 0.296,0.951 0.34,2.078 0.124,3.288 l -0.092,0.515 0.518,0.069 c 1.095,0.145 2.048,0.43 2.788,0.832 l -2.046,3.156 3.511,1.355 c -0.361,0.757 -0.966,1.547 -1.77,2.311 l -0.379,0.36 0.377,0.363 c 0.888,0.854 1.516,1.793 1.835,2.736 l -2.984,1.52 1.521,2.984 c -0.812,0.574 -1.871,0.964 -3.094,1.134 l -0.518,0.072 0.096,0.514 c 0.201,1.089 0.226,2.083 0.073,2.909 l -3.634,-0.97 -0.204,3.757 c -0.83,-0.11 -1.768,-0.44 -2.742,-0.968 l -0.459,-0.249 -0.228,0.47 c -0.539,1.107 -1.237,1.994 -2.036,2.591 L 18,32.293 Z"
+       id="path2" />
+    <path
+       d="M 7.092,10.678 C 6.562,9.189 6.394,7.708 6.66,6.478 l 2.368,0.375 0.987,0.156 0.157,-0.988 0.375,-2.368 C 11.808,3.78 13.16,4.396 14.409,5.359 14.527,5.022 14.653,4.696 14.791,4.392 13.24,3.257 11.568,2.629 10.061,2.629 9.938,2.629 9.816,2.633 9.695,2.642 L 9.184,5.865 5.96,5.354 C 5.36,6.841 5.395,8.769 6.045,10.747 6.38,10.71 6.729,10.686 7.092,10.678 Z M 21.593,5.359 c 1.248,-0.962 2.6,-1.578 3.86,-1.705 l 0.376,2.368 0.156,0.988 0.987,-0.157 2.369,-0.376 c 0.266,1.23 0.098,2.71 -0.432,4.2 0.361,0.009 0.711,0.032 1.046,0.07 C 30.606,8.769 30.64,6.841 30.04,5.353 L 26.815,5.865 26.304,2.641 c -0.12,-0.008 -0.242,-0.012 -0.365,-0.012 -1.507,0 -3.179,0.628 -4.73,1.762 0.14,0.306 0.266,0.631 0.384,0.968 z M 7.368,27 h 0.035 c 0.067,0 0.157,-0.604 0.26,-0.947 -0.098,0.004 -0.197,0.046 -0.294,0.046 -1.496,0 -2.826,-0.303 -3.83,-0.89 L 4.628,23.081 5.082,22.194 4.191,21.742 2.055,20.654 C 2.563,19.503 3.57,18.404 4.873,17.511 4.586,17.292 4.312,17.07 4.063,16.842 2.376,18.059 1.217,19.597 0.828,21.152 l 2.908,1.483 -1.482,2.843 C 3.475,26.501 5.303,27 7.368,27 Z m 27.806,-5.846 c -0.39,-1.555 -1.548,-3.093 -3.234,-4.311 -0.25,0.228 -0.523,0.451 -0.81,0.669 1.304,0.893 2.31,1.992 2.817,3.145 l -2.136,1.088 -0.891,0.453 0.454,0.892 1.089,2.137 c -1.004,0.587 -2.332,0.904 -3.828,0.904 -0.099,0 -0.199,-0.01 -0.299,-0.013 0.103,0.344 0.192,0.683 0.26,1.011 l 0.039,0.002 c 2.066,0 3.892,-0.563 5.112,-1.587 l -1.482,-2.908 z m -12.653,9.182 c -0.447,1.517 -1.181,2.812 -2.119,3.651 L 18.707,32.293 18,31.586 l -0.707,0.707 -1.695,1.694 c -0.938,-0.839 -1.673,-2.136 -2.12,-3.652 -0.296,0.206 -0.593,0.397 -0.886,0.563 0.636,1.98 1.741,3.559 3.1,4.409 L 18,33 l 2.308,2.308 c 1.358,-0.851 2.464,-2.428 3.101,-4.408 -0.295,-0.168 -0.591,-0.359 -0.888,-0.564 z"
+       fill="#ea596e"
+       id="path4" />
+    <path
+       fill="#ea596e"
+       d="m 20.118,5.683 c 0.426,1.146 0.748,2.596 0.841,4.284 l 0.2,3.683 3.564,-0.946 c 1.32,-0.351 2.655,-0.536 3.86,-0.536 0.16,0 0.318,0.003 0.474,0.01 l -1.827,2.819 3.139,1.211 c -0.958,0.759 -2.237,1.514 -3.814,2.123 l -3.441,1.328 2.001,3.099 c 0.918,1.42 1.509,2.782 1.838,3.96 L 23.709,25.853 23.527,29.21 C 22.508,28.533 21.395,27.55 20.329,26.237 L 18,23.374 15.672,26.236 c -1.066,1.312 -2.179,2.295 -3.198,2.972 l -0.18,-3.354 -3.248,0.864 c 0.329,-1.178 0.921,-2.54 1.839,-3.961 L 12.889,19.658 9.447,18.33 C 7.87,17.721 6.591,16.967 5.633,16.208 L 8.768,15 6.941,12.177 c 0.155,-0.006 0.313,-0.01 0.473,-0.01 1.206,0 2.541,0.185 3.861,0.536 l 3.564,0.947 0.202,-3.683 c 0.092,-1.688 0.415,-3.138 0.84,-4.284 L 18,8.292 20.118,5.683 M 20.308,0.692 18,3.533 15.692,0.692 C 13.703,2.224 12.271,5.684 12.046,9.804 10.429,9.374 8.854,9.167 7.414,9.167 c -2.11,0 -3.929,0.445 -5.161,1.289 l 1.989,3.073 -3.415,1.316 c 0.842,2.366 3.69,4.797 7.54,6.283 -2.241,3.465 -3.116,7.106 -2.407,9.516 l 3.537,-0.941 0.196,3.654 c 2.512,-0.07 5.703,-2.027 8.307,-5.228 2.603,3.201 5.796,5.158 8.306,5.228 l 0.198,-3.655 3.535,0.943 c 0.71,-2.411 -0.165,-6.05 -2.404,-9.517 3.849,-1.485 6.696,-3.918 7.538,-6.283 l -3.415,-1.318 1.99,-3.07 c -1.233,-0.844 -3.053,-1.29 -5.164,-1.29 -1.438,0 -3.013,0.207 -4.63,0.636 C 23.729,5.684 22.297,2.224 20.308,0.692 Z"
+       id="path6" />
+  </g>
+  <g
+     id="g79964"
+     transform="translate(-1.043816,35.993714)">
+    <path
+       fill="#ffffff"
+       d="m 51.92633,-2.0247139 c -0.799,-0.597 -1.498,-1.484 -2.035,-2.592 l -0.228,-0.47 -0.46,0.249 c -0.975,0.528 -1.913,0.858 -2.744,0.969 l -0.202,-3.7560001 -3.636,0.968 c -0.157,-0.854 -0.125,-1.887 0.096,-3.022 l 0.103,-0.525 -0.532,-0.066 c -1.242,-0.154 -2.306,-0.525 -3.104,-1.08 l 1.521,-2.917 -2.988,-1.523 c 0.319,-0.944 0.948,-1.882 1.834,-2.735 l 0.377,-0.363 -0.379,-0.36 c -0.803,-0.764 -1.408,-1.554 -1.77,-2.311 l 3.51,-1.353 -2.045,-3.159 c 0.74,-0.402 1.693,-0.686 2.789,-0.832 l 0.519,-0.068 -0.091,-0.514 c -0.215,-1.211 -0.172,-2.338 0.124,-3.288 l 3.308,0.523 0.524,-3.308 c 0.988,0.013 2.08,0.326 3.164,0.907 l 0.462,0.248 0.226,-0.473 c 0.479,-1.003 1.044,-1.824 1.653,-2.404 l 2.372,2.92 2.372,-2.92 c 0.609,0.58 1.175,1.401 1.653,2.404 l 0.226,0.473 0.462,-0.247 c 1.085,-0.581 2.178,-0.894 3.164,-0.906 l 0.523,3.308 3.31,-0.525 c 0.296,0.951 0.34,2.078 0.124,3.288 l -0.092,0.515 0.518,0.069 c 1.095,0.145 2.048,0.43 2.788,0.832 l -2.046,3.156 3.511,1.355 c -0.361,0.757 -0.966,1.547 -1.77,2.311 l -0.379,0.36 0.377,0.363 c 0.888,0.854 1.516,1.793 1.835,2.736 l -2.984,1.52 1.521,2.984 c -0.812,0.574 -1.871,0.964 -3.094,1.134 l -0.518,0.072 0.096,0.514 c 0.201,1.089 0.226,2.083 0.073,2.909 l -3.634,-0.97 -0.204,3.7570001 c -0.83,-0.11 -1.768,-0.44 -2.742,-0.968 l -0.459,-0.249 -0.228,0.47 c -0.539,1.107 -1.237,1.994 -2.036,2.591 l -2.367,-2.369 z"
+       id="path2-9" />
+    <path
+       d="m 43.38633,-26.007714 c -0.53,-1.489 -0.698,-2.97 -0.432,-4.2 l 2.368,0.375 0.987,0.156 0.157,-0.988 0.375,-2.368 c 1.261,0.127 2.613,0.743 3.862,1.706 0.118,-0.337 0.244,-0.663 0.382,-0.967 -1.551,-1.135 -3.223,-1.763 -4.73,-1.763 -0.123,0 -0.245,0.004 -0.366,0.013 l -0.511,3.223 -3.224,-0.511 c -0.6,1.487 -0.565,3.415 0.085,5.393 0.335,-0.037 0.684,-0.061 1.047,-0.069 z m 14.501,-5.319 c 1.248,-0.962 2.6,-1.578 3.86,-1.705 l 0.376,2.368 0.156,0.988 0.987,-0.157 2.369,-0.376 c 0.266,1.23 0.098,2.71 -0.432,4.2 0.361,0.009 0.711,0.032 1.046,0.07 0.651,-1.978 0.685,-3.906 0.085,-5.394 l -3.225,0.512 -0.511,-3.224 c -0.12,-0.008 -0.242,-0.012 -0.365,-0.012 -1.507,0 -3.179,0.628 -4.73,1.762 0.14,0.306 0.266,0.631 0.384,0.968 z m -14.225,21.641 h 0.035 c 0.067,0 0.157,-0.604 0.26,-0.947 -0.098,0.004 -0.197,0.046 -0.294,0.046 -1.496,0 -2.826,-0.303 -3.83,-0.89 l 1.089,-2.128 0.454,-0.887 -0.891,-0.452 -2.136,-1.088 c 0.508,-1.151 1.515,-2.25 2.818,-3.143 -0.287,-0.219 -0.561,-0.441 -0.81,-0.669 -1.687,1.217 -2.846,2.755 -3.235,4.31 l 2.908,1.483 -1.482,2.843 c 1.221,1.023 3.049,1.522 5.114,1.522 z m 27.806,-5.846 c -0.39,-1.555 -1.548,-3.093 -3.234,-4.311 -0.25,0.228 -0.523,0.451 -0.81,0.669 1.304,0.893 2.31,1.992 2.817,3.145 l -2.136,1.088 -0.891,0.453 0.454,0.892 1.089,2.137 c -1.004,0.587 -2.332,0.904 -3.828,0.904 -0.099,0 -0.199,-0.01 -0.299,-0.013 0.103,0.344 0.192,0.683 0.26,1.011 l 0.039,0.002 c 2.066,0 3.892,-0.563 5.112,-1.587 l -1.482,-2.908 z m -12.653,9.182 c -0.447,1.5170001 -1.181,2.8120001 -2.119,3.6510001 l -1.695,-1.694 -0.707,-0.707 -0.707,0.707 -1.695,1.694 c -0.938,-0.839 -1.673,-2.136 -2.12,-3.6520001 -0.296,0.2060001 -0.593,0.3970001 -0.886,0.5630001 0.636,1.98 1.741,3.559 3.1,4.409 l 2.308,-2.307 2.308,2.308 c 1.358,-0.851 2.464,-2.428 3.101,-4.408 -0.295,-0.168 -0.591,-0.359 -0.888,-0.5640001 z"
+       fill="#ea596e"
+       id="path4-3" />
+    <path
+       fill="#ea596e"
+       d="m 56.41233,-31.002714 c 0.426,1.146 0.748,2.596 0.841,4.284 l 0.2,3.683 3.564,-0.946 c 1.32,-0.351 2.655,-0.536 3.86,-0.536 0.16,0 0.318,0.003 0.474,0.01 l -1.827,2.819 3.139,1.211 c -0.958,0.759 -2.237,1.514 -3.814,2.123 l -3.441,1.328 2.001,3.099 c 0.918,1.42 1.509,2.782 1.838,3.96 l -3.244,-0.865 -0.182,3.357 c -1.019,-0.677 -2.132,-1.66 -3.198,-2.973 l -2.329,-2.863 -2.328,2.862 c -1.066,1.312 -2.179,2.295 -3.198,2.972 l -0.18,-3.354 -3.248,0.864 c 0.329,-1.178 0.921,-2.54 1.839,-3.961 l 2.004,-3.099 -3.442,-1.328 c -1.577,-0.609 -2.856,-1.363 -3.814,-2.122 l 3.135,-1.208 -1.827,-2.823 c 0.155,-0.006 0.313,-0.01 0.473,-0.01 1.206,0 2.541,0.185 3.861,0.536 l 3.564,0.947 0.202,-3.683 c 0.092,-1.688 0.415,-3.138 0.84,-4.284 l 2.119,2.609 2.118,-2.609 m 0.19,-4.991 -2.308,2.841 -2.308,-2.841 c -1.989,1.532 -3.421,4.992 -3.646,9.112 -1.617,-0.43 -3.192,-0.637 -4.632,-0.637 -2.11,0 -3.929,0.445 -5.161,1.289 l 1.989,3.073 -3.415,1.316 c 0.842,2.366 3.69,4.797 7.54,6.283 -2.241,3.465 -3.116,7.106 -2.407,9.5160001 l 3.537,-0.9410001 0.196,3.6540001 c 2.512,-0.07 5.703,-2.027 8.307,-5.2280001 2.603,3.2010001 5.796,5.1580001 8.306,5.2280001 l 0.198,-3.6550001 3.535,0.9430001 c 0.71,-2.4110001 -0.165,-6.0500001 -2.404,-9.5170001 3.849,-1.485 6.696,-3.918 7.538,-6.283 l -3.415,-1.318 1.99,-3.07 c -1.233,-0.844 -3.053,-1.29 -5.164,-1.29 -1.438,0 -3.013,0.207 -4.63,0.636 -0.225,-4.119 -1.657,-7.579 -3.646,-9.111 z"
+       id="path6-6" />
+  </g>
+  <text
+     xml:space="preserve"
+     style="font-style:normal;font-weight:normal;font-size:42.6667px;line-height:1.25;font-family:sans-serif;fill:#ea596e;fill-opacity:1;stroke:none"
+     x="2.2188232"
+     y="31.430677"
+     id="text46212"><tspan
+       sodipodi:role="line"
+       id="tspan46210"
+       x="2.2188232"
+       y="31.430677"
+       style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:42.6667px;font-family:'TeX Gyre Termes';-inkscape-font-specification:'TeX Gyre Termes'">D</tspan></text>
+  <text
+     xml:space="preserve"
+     style="font-style:normal;font-weight:normal;font-size:42.6667px;line-height:1.25;font-family:sans-serif;fill:#ea596e;fill-opacity:1;stroke:none"
+     x="41.347008"
+     y="67.114784"
+     id="text46212-1"><tspan
+       sodipodi:role="line"
+       id="tspan46210-5"
+       x="41.347008"
+       y="67.114784"
+       style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:42.6667px;font-family:'TeX Gyre Termes';-inkscape-font-specification:'TeX Gyre Termes'">F</tspan></text>
+</svg>

doc/talks/2022-06-23-stack/assets/garage.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d9ca1a04f943664e14d0e1edbde596789e4798f8ad5e1cff68bc9dd0cc1334f
+size 26098

doc/talks/2022-06-23-stack/assets/garage2a.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bffa5d4a72ad25e0b18b43022258a2e4471ce94ff5397aab892fe43ec9d4d7d5
+size 33911

doc/talks/2022-06-23-stack/assets/garage2b.drawio.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c4913b52e84b5af3f8effbcb7e2060845a8c6f43d5b38072b9ee33c0300d49f
+size 31051

doc/talks/2022-06-23-stack/assets/garage_tables.svg

@@ -0,0 +1,537 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   width="850"
+   height="480"
+   viewBox="0 0 224.89584 127"
+   version="1.1"
+   id="svg8"
+   inkscape:version="1.2 (dc2aedaf03, 2022-05-15)"
+   sodipodi:docname="garage_tables.svg"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:dc="http://purl.org/dc/elements/1.1/">
+  <defs
+     id="defs2">
+    <marker
+       style="overflow:visible"
+       id="marker1262"
+       refX="0"
+       refY="0"
+       orient="auto"
+       inkscape:stockid="Arrow1Mend"
+       inkscape:isstock="true">
+      <path
+         transform="matrix(-0.4,0,0,-0.4,-4,0)"
+         style="fill:#000000;fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:1pt;stroke-opacity:1"
+         d="M 0,0 5,-5 -12.5,0 5,5 Z"
+         id="path1260" />
+    </marker>
+    <marker
+       style="overflow:visible"
+       id="Arrow1Mend"
+       refX="0"
+       refY="0"
+       orient="auto"
+       inkscape:stockid="Arrow1Mend"
+       inkscape:isstock="true"
+       inkscape:collect="always">
+      <path
+         transform="matrix(-0.4,0,0,-0.4,-4,0)"
+         style="fill:#000000;fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:1pt;stroke-opacity:1"
+         d="M 0,0 5,-5 -12.5,0 5,5 Z"
+         id="path965" />
+    </marker>
+    <marker
+       style="overflow:visible"
+       id="Arrow1Lend"
+       refX="0"
+       refY="0"
+       orient="auto"
+       inkscape:stockid="Arrow1Lend"
+       inkscape:isstock="true">
+      <path
+         transform="matrix(-0.8,0,0,-0.8,-10,0)"
+         style="fill:#000000;fill-opacity:1;fill-rule:evenodd;stroke:#000000;stroke-width:1pt;stroke-opacity:1"
+         d="M 0,0 5,-5 -12.5,0 5,5 Z"
+         id="path959" />
+    </marker>
+  </defs>
+  <sodipodi:namedview
+     id="base"
+     pagecolor="#ffffff"
+     bordercolor="#666666"
+     borderopacity="1.0"
+     inkscape:pageopacity="0.0"
+     inkscape:pageshadow="2"
+     inkscape:zoom="0.98994949"
+     inkscape:cx="429.31483"
+     inkscape:cy="289.40871"
+     inkscape:document-units="mm"
+     inkscape:current-layer="layer1"
+     inkscape:document-rotation="0"
+     showgrid="false"
+     units="px"
+     inkscape:window-width="1678"
+     inkscape:window-height="993"
+     inkscape:window-x="0"
+     inkscape:window-y="0"
+     inkscape:window-maximized="1"
+     inkscape:showpageshadow="2"
+     inkscape:pagecheckerboard="0"
+     inkscape:deskcolor="#d1d1d1" />
+  <metadata
+     id="metadata5">
+    <rdf:RDF>
+      <cc:Work
+         rdf:about="">
+        <dc:format>image/svg+xml</dc:format>
+        <dc:type
+           rdf:resource="http://purl.org/dc/dcmitype/StillImage" />
+        <dc:title />
+      </cc:Work>
+    </rdf:RDF>
+  </metadata>
+  <g
+     inkscape:label="Layer 1"
+     inkscape:groupmode="layer"
+     id="layer1">
+    <text
+       xml:space="preserve"
+       style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+       x="39.570904"
+       y="38.452755"
+       id="text2025"><tspan
+         sodipodi:role="line"
+         id="tspan2023"
+         x="39.570904"
+         y="38.452755"
+         style="font-size:5.64444px;stroke-width:0.264583" /></text>
+    <text
+       xml:space="preserve"
+       style="font-style:normal;font-weight:normal;font-size:10.5833px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+       x="101.95796"
+       y="92.835831"
+       id="text2139"><tspan
+         sodipodi:role="line"
+         id="tspan2137"
+         x="101.95796"
+         y="92.835831"
+         style="stroke-width:0.264583"> </tspan></text>
+    <g
+       id="g2316"
+       transform="translate(-11.455511,1.5722486)">
+      <g
+         id="g2277">
+        <rect
+           style="fill:none;stroke:#000000;stroke-width:0.8;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+           id="rect833"
+           width="47.419891"
+           height="95.353409"
+           x="18.534418"
+           y="24.42766" />
+        <rect
+           style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+           id="rect833-3"
+           width="47.419891"
+           height="86.973076"
+           x="18.534418"
+           y="32.807987" />
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="32.250839"
+           y="29.894743"
+           id="text852"><tspan
+             sodipodi:role="line"
+             id="tspan850"
+             x="32.250839"
+             y="29.894743"
+             style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Object</tspan></text>
+      </g>
+      <g
+         id="g2066"
+         transform="translate(-2.1807817,-3.0621439)">
+        <g
+           id="g1969"
+           transform="matrix(0.12763631,0,0,0.12763631,0.7215051,24.717273)"
+           style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-opacity:1">
+          <path
+             style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-width:0.264583;stroke-opacity:1"
+             d="m 203.71837,154.80038 c -1.11451,3.75057 -2.45288,5.84095 -5.11132,7.98327 -2.2735,1.83211 -4.66721,2.65982 -8.09339,2.79857 -2.59227,0.10498 -2.92868,0.0577 -5.02863,-0.70611 -3.99215,-1.45212 -7.1627,-4.65496 -8.48408,-8.57046 -1.28374,-3.80398 -0.61478,-8.68216 1.64793,-12.01698 0.87317,-1.28689 3.15089,-3.48326 4.18771,-4.03815 l 0.53332,-28.51234 5.78454,-5.09197 6.95158,6.16704 -3.21112,3.49026 3.17616,3.45499 -3.17616,3.40822 2.98973,3.28645 -3.24843,3.3829 4.49203,4.58395 0.0516,5.69106 c 1.06874,0.64848 3.81974,3.24046 4.69548,4.56257 0.452,0.68241 1.06834,2.0197 1.36962,2.97176 0.62932,1.98864 0.88051,5.785 0.47342,7.15497 z m -10.0406,2.32604 c -0.88184,-3.17515 -4.92402,-3.78864 -6.75297,-1.02492 -0.58328,0.8814 -0.6898,1.28852 -0.58362,2.23056 0.26492,2.35041 2.45434,3.95262 4.60856,3.37255 1.19644,-0.32217 2.39435,-1.44872 2.72875,-2.56621 0.30682,-1.02529 0.30686,-0.9045 -7.9e-4,-2.01198 z"
+             id="path1971"
+             sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+        </g>
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="28.809687"
+           y="44.070885"
+           id="text852-9"><tspan
+             sodipodi:role="line"
+             id="tspan850-4"
+             x="28.809687"
+             y="44.070885"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">bucket </tspan></text>
+      </g>
+      <g
+         id="g2066-7"
+         transform="translate(-2.1807817,6.2627616)">
+        <g
+           id="g1969-8"
+           transform="matrix(0.12763631,0,0,0.12763631,0.7215051,24.717273)"
+           style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-opacity:1">
+          <path
+             style="fill:#4040ff;fill-opacity:1;stroke:none;stroke-width:0.264583;stroke-opacity:1"
+             d="m 203.71837,154.80038 c -1.11451,3.75057 -2.45288,5.84095 -5.11132,7.98327 -2.2735,1.83211 -4.66721,2.65982 -8.09339,2.79857 -2.59227,0.10498 -2.92868,0.0577 -5.02863,-0.70611 -3.99215,-1.45212 -7.1627,-4.65496 -8.48408,-8.57046 -1.28374,-3.80398 -0.61478,-8.68216 1.64793,-12.01698 0.87317,-1.28689 3.15089,-3.48326 4.18771,-4.03815 l 0.53332,-28.51234 5.78454,-5.09197 6.95158,6.16704 -3.21112,3.49026 3.17616,3.45499 -3.17616,3.40822 2.98973,3.28645 -3.24843,3.3829 4.49203,4.58395 0.0516,5.69106 c 1.06874,0.64848 3.81974,3.24046 4.69548,4.56257 0.452,0.68241 1.06834,2.0197 1.36962,2.97176 0.62932,1.98864 0.88051,5.785 0.47342,7.15497 z m -10.0406,2.32604 c -0.88184,-3.17515 -4.92402,-3.78864 -6.75297,-1.02492 -0.58328,0.8814 -0.6898,1.28852 -0.58362,2.23056 0.26492,2.35041 2.45434,3.95262 4.60856,3.37255 1.19644,-0.32217 2.39435,-1.44872 2.72875,-2.56621 0.30682,-1.02529 0.30686,-0.9045 -7.9e-4,-2.01198 z"
+             id="path1971-4"
+             sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+        </g>
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="28.809687"
+           y="44.070885"
+           id="text852-9-5"><tspan
+             sodipodi:role="line"
+             id="tspan850-4-0"
+             x="28.809687"
+             y="44.070885"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">file path </tspan></text>
+        <path
+           style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-width:0.0337704;stroke-opacity:1"
+           d="m 174.20027,104.45585 c -0.14225,0.47871 -0.31308,0.74552 -0.65239,1.01896 -0.29018,0.23384 -0.5957,0.33949 -1.03301,0.3572 -0.33087,0.0134 -0.37381,0.007 -0.64184,-0.0901 -0.50954,-0.18534 -0.91422,-0.59414 -1.08287,-1.0939 -0.16385,-0.48552 -0.0785,-1.10816 0.21033,-1.5338 0.11145,-0.16426 0.40217,-0.44459 0.53451,-0.51542 l 0.0681,-3.639207 0.73832,-0.64992 0.88727,0.787138 -0.40986,0.445484 0.4054,0.440982 -0.4054,0.435013 0.3816,0.41947 -0.41461,0.43178 0.57334,0.58508 0.007,0.72639 c 0.13641,0.0828 0.48753,0.4136 0.59931,0.58235 0.0577,0.0871 0.13636,0.25778 0.17481,0.3793 0.0803,0.25382 0.11239,0.73838 0.0604,0.91323 z m -1.28154,0.29689 c -0.11256,-0.40526 -0.62849,-0.48357 -0.86193,-0.13082 -0.0745,0.1125 -0.088,0.16447 -0.0745,0.2847 0.0338,0.3 0.31326,0.5045 0.58822,0.43046 0.15271,-0.0411 0.30561,-0.1849 0.34829,-0.32754 0.0392,-0.13086 0.0392,-0.11544 -1e-4,-0.2568 z"
+           id="path1971-3"
+           sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="177.8474"
+           y="104.05132"
+           id="text852-9-6"><tspan
+             sodipodi:role="line"
+             id="tspan850-4-7"
+             x="177.8474"
+             y="104.05132"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">= partition key </tspan></text>
+        <path
+           style="fill:#4040ff;fill-opacity:1;stroke:none;stroke-width:0.0337704;stroke-opacity:1"
+           d="m 174.20027,113.78076 c -0.14225,0.47871 -0.31308,0.74552 -0.65239,1.01895 -0.29018,0.23385 -0.5957,0.33949 -1.03301,0.3572 -0.33087,0.0134 -0.37381,0.007 -0.64184,-0.0901 -0.50954,-0.18534 -0.91422,-0.59414 -1.08287,-1.0939 -0.16385,-0.48553 -0.0785,-1.10816 0.21033,-1.53381 0.11145,-0.16425 0.40217,-0.44459 0.53451,-0.51541 l 0.0681,-3.63921 0.73832,-0.64992 0.88727,0.78714 -0.40986,0.44548 0.4054,0.44098 -0.4054,0.43502 0.3816,0.41947 -0.41461,0.43178 0.57334,0.58508 0.007,0.72638 c 0.13641,0.0828 0.48753,0.4136 0.59931,0.58235 0.0577,0.0871 0.13636,0.25779 0.17481,0.37931 0.0803,0.25382 0.11239,0.73837 0.0604,0.91323 z m -1.28154,0.29689 c -0.11256,-0.40527 -0.62849,-0.48357 -0.86193,-0.13082 -0.0745,0.1125 -0.088,0.16446 -0.0745,0.2847 0.0338,0.3 0.31326,0.5045 0.58822,0.43046 0.15271,-0.0411 0.30561,-0.18491 0.34829,-0.32754 0.0392,-0.13087 0.0392,-0.11545 -1e-4,-0.2568 z"
+           id="path1971-4-5"
+           sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="177.8474"
+           y="113.37622"
+           id="text852-9-5-3"><tspan
+             sodipodi:role="line"
+             id="tspan850-4-0-5"
+             x="177.8474"
+             y="113.37622"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">= sort key </tspan></text>
+      </g>
+      <g
+         id="g2161"
+         transform="translate(-62.264403,-59.333115)">
+        <g
+           id="g2271"
+           transform="translate(0,67.042823)">
+          <rect
+             style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+             id="rect833-6"
+             width="39.008453"
+             height="16.775949"
+             x="84.896881"
+             y="90.266838" />
+          <rect
+             style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+             id="rect833-3-1"
+             width="39.008453"
+             height="8.673645"
+             x="84.896881"
+             y="98.369141" />
+          <text
+             xml:space="preserve"
+             style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+             x="89.826942"
+             y="96.212921"
+             id="text852-0"><tspan
+               sodipodi:role="line"
+               id="tspan850-6"
+               x="89.826942"
+               y="96.212921"
+               style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Version 1</tspan></text>
+          <text
+             xml:space="preserve"
+             style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+             x="89.826942"
+             y="104.71013"
+             id="text852-0-3"><tspan
+               sodipodi:role="line"
+               id="tspan850-6-2"
+               x="89.826942"
+               y="104.71013"
+               style="font-style:italic;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';fill:#4d4d4d;stroke-width:0.264583">deleted</tspan></text>
+        </g>
+      </g>
+      <g
+         id="g2263"
+         transform="translate(0,-22.791204)">
+        <g
+           id="g2161-1"
+           transform="translate(-62.264403,-10.910843)">
+          <rect
+             style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+             id="rect833-6-5"
+             width="39.008453"
+             height="36.749603"
+             x="84.896881"
+             y="90.266838" />
+          <rect
+             style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+             id="rect833-3-1-5"
+             width="39.008453"
+             height="28.647301"
+             x="84.896881"
+             y="98.369141" />
+          <text
+             xml:space="preserve"
+             style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+             x="89.826942"
+             y="96.212921"
+             id="text852-0-4"><tspan
+               sodipodi:role="line"
+               id="tspan850-6-7"
+               x="89.826942"
+               y="96.212921"
+               style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Version 2</tspan></text>
+          <text
+             xml:space="preserve"
+             style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+             x="89.826942"
+             y="104.71013"
+             id="text852-0-3-6"><tspan
+               sodipodi:role="line"
+               id="tspan850-6-2-5"
+               x="89.826942"
+               y="104.71013"
+               style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';fill:#000000;stroke-width:0.264583">id</tspan></text>
+        </g>
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="27.56254"
+           y="100.34132"
+           id="text852-0-3-6-6"><tspan
+             sodipodi:role="line"
+             id="tspan850-6-2-5-9"
+             x="27.56254"
+             y="100.34132"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';fill:#000000;stroke-width:0.264583">size</tspan></text>
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="27.56254"
+           y="106.90263"
+           id="text852-0-3-6-6-3"><tspan
+             sodipodi:role="line"
+             id="tspan850-6-2-5-9-7"
+             x="27.56254"
+             y="106.90263"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';fill:#000000;stroke-width:0.264583">MIME type</tspan></text>
+        <text
+           xml:space="preserve"
+           style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+           x="27.56254"
+           y="111.92816"
+           id="text852-0-3-6-6-3-4"><tspan
+             sodipodi:role="line"
+             id="tspan850-6-2-5-9-7-5"
+             x="27.56254"
+             y="111.92816"
+             style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';fill:#000000;stroke-width:0.264583">...</tspan></text>
+      </g>
+    </g>
+    <g
+       id="g898"
+       transform="translate(-6.2484318,29.95006)">
+      <rect
+         style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+         id="rect833-7"
+         width="47.419891"
+         height="44.007515"
+         x="95.443573"
+         y="24.42766" />
+      <rect
+         style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+         id="rect833-3-4"
+         width="47.419891"
+         height="35.627186"
+         x="95.443573"
+         y="32.807987" />
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="107.46638"
+         y="29.894743"
+         id="text852-4"><tspan
+           sodipodi:role="line"
+           id="tspan850-3"
+           x="107.46638"
+           y="29.894743"
+           style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Version</tspan></text>
+      <path
+         style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-width:0.0337704;stroke-opacity:1"
+         d="m 102.90563,41.413279 c -0.14226,0.478709 -0.31308,0.745518 -0.65239,1.018956 -0.29019,0.233843 -0.59571,0.339489 -1.03301,0.357199 -0.33087,0.0134 -0.37381,0.0074 -0.64184,-0.09013 -0.50954,-0.185343 -0.914221,-0.594142 -1.082877,-1.093901 -0.163852,-0.485526 -0.07847,-1.108159 0.210335,-1.533803 0.111448,-0.164254 0.402172,-0.444591 0.534502,-0.515415 l 0.0681,-3.63921 0.73832,-0.64992 0.88727,0.787138 -0.40985,0.445484 0.40539,0.440982 -0.40539,0.435013 0.3816,0.41947 -0.41462,0.431781 0.57335,0.585078 0.007,0.726386 c 0.13641,0.08277 0.48753,0.413601 0.59931,0.58235 0.0577,0.0871 0.13636,0.257787 0.17481,0.379304 0.0803,0.253823 0.11239,0.738377 0.0604,0.913234 z m -1.28155,0.296888 c -0.11255,-0.405265 -0.62848,-0.483569 -0.86192,-0.130817 -0.0744,0.112498 -0.088,0.164461 -0.0745,0.2847 0.0338,0.299998 0.31326,0.504498 0.58822,0.43046 0.15271,-0.04112 0.3056,-0.184909 0.34828,-0.327542 0.0392,-0.130864 0.0392,-0.115447 -1e-4,-0.256801 z"
+         id="path1971-0"
+         sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="41.008743"
+         id="text852-9-7"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8"
+           x="104.99195"
+           y="41.008743"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">id </tspan></text>
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="49.168018"
+         id="text852-9-7-6"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8-8"
+           x="104.99195"
+           y="49.168018"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">h(block 1)</tspan></text>
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="56.583336"
+         id="text852-9-7-6-8"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8-8-4"
+           x="104.99195"
+           y="56.583336"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">h(block 2)</tspan></text>
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="64.265732"
+         id="text852-9-7-6-3"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8-8-1"
+           x="104.99195"
+           y="64.265732"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">...</tspan></text>
+    </g>
+    <g
+       id="g898-3"
+       transform="translate(75.777779,38.888663)">
+      <rect
+         style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+         id="rect833-7-6"
+         width="47.419891"
+         height="29.989157"
+         x="95.443573"
+         y="24.42766" />
+      <rect
+         style="fill:none;stroke:#000000;stroke-width:0.799999;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1"
+         id="rect833-3-4-7"
+         width="47.419891"
+         height="21.608831"
+         x="95.443573"
+         y="32.807987" />
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="102.11134"
+         y="29.894743"
+         id="text852-4-5"><tspan
+           sodipodi:role="line"
+           id="tspan850-3-3"
+           x="102.11134"
+           y="29.894743"
+           style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Data block</tspan></text>
+      <path
+         style="fill:#ff6600;fill-opacity:1;stroke:none;stroke-width:0.0337704;stroke-opacity:1"
+         d="m 102.90563,41.413279 c -0.14226,0.478709 -0.31308,0.745518 -0.65239,1.018956 -0.29019,0.233843 -0.59571,0.339489 -1.03301,0.357199 -0.33087,0.0134 -0.37381,0.0074 -0.64184,-0.09013 -0.50954,-0.185343 -0.914221,-0.594142 -1.082877,-1.093901 -0.163852,-0.485526 -0.07847,-1.108159 0.210335,-1.533803 0.111448,-0.164254 0.402172,-0.444591 0.534502,-0.515415 l 0.0681,-3.63921 0.73832,-0.64992 0.88727,0.787138 -0.40985,0.445484 0.40539,0.440982 -0.40539,0.435013 0.3816,0.41947 -0.41462,0.431781 0.57335,0.585078 0.007,0.726386 c 0.13641,0.08277 0.48753,0.413601 0.59931,0.58235 0.0577,0.0871 0.13636,0.257787 0.17481,0.379304 0.0803,0.253823 0.11239,0.738377 0.0604,0.913234 z m -1.28155,0.296888 c -0.11255,-0.405265 -0.62848,-0.483569 -0.86192,-0.130817 -0.0744,0.112498 -0.088,0.164461 -0.0745,0.2847 0.0338,0.299998 0.31326,0.504498 0.58822,0.43046 0.15271,-0.04112 0.3056,-0.184909 0.34828,-0.327542 0.0392,-0.130864 0.0392,-0.115447 -1e-4,-0.256801 z"
+         id="path1971-0-5"
+         sodipodi:nodetypes="ssscsscccccccccccssscsssscc" />
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="41.008743"
+         id="text852-9-7-62"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8-9"
+           x="104.99195"
+           y="41.008743"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">hash </tspan></text>
+      <text
+         xml:space="preserve"
+         style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+         x="104.99195"
+         y="49.168018"
+         id="text852-9-7-6-1"><tspan
+           sodipodi:role="line"
+           id="tspan850-4-8-8-2"
+           x="104.99195"
+           y="49.168018"
+           style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">data</tspan></text>
+    </g>
+    <path
+       style="fill:none;stroke:#000000;stroke-width:1;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1;marker-end:url(#Arrow1Mend)"
+       d="M 42.105292,69.455903 89.563703,69.317144"
+       id="path954"
+       sodipodi:nodetypes="cc" />
+    <path
+       style="fill:none;stroke:#000000;stroke-width:1;stroke-linecap:butt;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1;marker-end:url(#marker1262)"
+       d="m 134.32612,77.363197 38.12618,0.260865"
+       id="path1258"
+       sodipodi:nodetypes="cc" />
+    <text
+       xml:space="preserve"
+       style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+       x="8.6727352"
+       y="16.687063"
+       id="text852-3"><tspan
+         sodipodi:role="line"
+         id="tspan850-67"
+         x="8.6727352"
+         y="16.687063"
+         style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Objects table </tspan></text>
+    <text
+       xml:space="preserve"
+       style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+       x="89.190445"
+       y="16.687063"
+       id="text852-3-5"><tspan
+         sodipodi:role="line"
+         id="tspan850-67-3"
+         x="89.190445"
+         y="16.687063"
+         style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Versions table </tspan></text>
+    <text
+       xml:space="preserve"
+       style="font-style:normal;font-weight:normal;font-size:5.64444px;line-height:1.25;font-family:sans-serif;fill:#000000;fill-opacity:1;stroke:none;stroke-width:0.264583"
+       x="174.55702"
+       y="16.687063"
+       id="text852-3-56"><tspan
+         sodipodi:role="line"
+         id="tspan850-67-2"
+         x="174.55702"
+         y="16.687063"
+         style="font-style:normal;font-variant:normal;font-weight:bold;font-stretch:normal;font-size:5.64444px;font-family:'Liberation Mono';-inkscape-font-specification:'Liberation Mono Bold';stroke-width:0.264583">Blocks table</tspan></text>
+  </g>
+</svg>

doc/talks/2022-06-23-stack/talk.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f245bb017eb85a227d96e576ae821f165ef478d46de02535c196638aa5fc84b0
+size 2575953

doc/talks/2022-06-23-stack/talk.tex

@@ -0,0 +1,480 @@
+%\nonstopmode
+\documentclass[aspectratio=169]{beamer}
+\usepackage[utf8]{inputenc}
+% \usepackage[frenchb]{babel}
+\usepackage{amsmath}
+\usepackage{mathtools}
+\usepackage{breqn}
+\usepackage{multirow}
+\usetheme{boxes}
+\usepackage{graphicx}
+\usepackage{adjustbox}
+%\useoutertheme[footline=authortitle,subsection=false]{miniframes}
+%\useoutertheme[footline=authorinstitute,subsection=false]{miniframes}
+\useoutertheme{infolines}
+\setbeamertemplate{headline}{}
+
+\beamertemplatenavigationsymbolsempty
+
+\definecolor{TitleOrange}{RGB}{255,137,0}
+\setbeamercolor{title}{fg=TitleOrange}
+\setbeamercolor{frametitle}{fg=TitleOrange}
+
+\definecolor{ListOrange}{RGB}{255,145,5}
+\setbeamertemplate{itemize item}{\color{ListOrange}$\blacktriangleright$}
+
+\definecolor{verygrey}{RGB}{70,70,70}
+\setbeamercolor{normal text}{fg=verygrey}
+
+
+\usepackage{tabu}
+\usepackage{multicol}
+\usepackage{vwcol}
+\usepackage{stmaryrd}
+\usepackage{graphicx}
+
+\usepackage[normalem]{ulem}
+
+\title{Introducing Garage}
+\subtitle{a new storage platform for self-hosted geo-distributed clusters}
+\author{Deuxfleurs Association}
+\date{IMT Atlantique, 2022-06-23}
+
+\begin{document}
+
+\begin{frame}
+	\centering
+	\includegraphics[width=.3\linewidth]{../../sticker/Garage.pdf}
+	\vspace{1em}
+
+	{\large\bf Deuxfleurs Association}
+	\vspace{1em}
+
+	\url{https://garagehq.deuxfleurs.fr/}
+
+	Matrix channel: \texttt{\#garage:deuxfleurs.fr}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Who we are}
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.4\linewidth, valign=t]{assets/alex.jpg}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Alex Auvolat}\\
+			PhD at Inria, team WIDE; co-founder of Deuxfleurs
+		\end{column}
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+	\end{columns}
+	\vspace{1em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Quentin Dufour}\\
+			PhD at Inria, team WIDE; co-founder of Deuxfleurs
+		\end{column}
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.5\linewidth, valign=t]{assets/quentin.jpg}
+		\end{column}
+	\end{columns}
+	\vspace{2em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.5\linewidth, valign=t]{assets/deuxfleurs.pdf}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Deuxfleurs}\\
+			A non-profit self-hosting collective,\\
+			member of the CHATONS network
+		\end{column}
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.7\linewidth, valign=t]{assets/logo_chatons.png}
+		\end{column}
+	\end{columns}
+
+\end{frame}
+
+\begin{frame}
+	\frametitle{Our objective at Deuxfleurs}
+	
+	\begin{center}
+		\textbf{Promote self-hosting and small-scale hosting\\
+			as an alternative to large cloud providers}
+	\end{center}
+	\vspace{2em}
+	\visible<2->{
+		Why is it hard?
+	}
+	\visible<3->{
+		\vspace{2em}
+		\begin{center}
+			\textbf{\underline{Resilience}}\\
+			{\footnotesize (we want good uptime/availability with low supervision)}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to make a \underline{stable} system}
+
+	Enterprise-grade systems typically employ:
+	\vspace{1em}
+	\begin{itemize}
+		\item RAID
+		\item Redundant power grid + UPS
+		\item Redundant Internet connections
+		\item Low-latency links
+		\item ... 
+	\end{itemize}
+	\vspace{1em}
+	$\to$ it's costly and only worth it at DC scale
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to make a \underline{resilient} system}
+
+	\only<1,4-5>{
+		Instead, we use:
+		\vspace{1em}
+		\begin{itemize}
+			\item \textcolor<2->{gray}{Commodity hardware (e.g. old desktop PCs)}
+				\vspace{.5em}
+			\item<4-> \textcolor<5->{gray}{Commodity Internet (e.g. FTTB, FTTH) and power grid}
+				\vspace{.5em}
+			\item<5-> \textcolor<6->{gray}{\textbf{Geographical redundancy} (multi-site replication)}
+		\end{itemize}
+	}
+	\only<2>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{assets/atuin.jpg}
+		\end{center}
+	}
+	\only<3>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{assets/neptune.jpg}
+		\end{center}
+	}
+	\only<6>{
+		\begin{center}
+			\includegraphics[width=.5\linewidth]{assets/inframap.jpg}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to make this happen}
+	\begin{center}
+		\only<1>{\includegraphics[width=.8\linewidth]{assets/slide1.png}}%
+		\only<2>{\includegraphics[width=.8\linewidth]{assets/slide2.png}}%
+		\only<3>{\includegraphics[width=.8\linewidth]{assets/slide3.png}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Distributed file systems are slow}
+	File systems are complex, for example:
+	\vspace{1em}
+	\begin{itemize}
+		\item Concurrent modification by several processes
+			\vspace{1em}
+		\item Folder hierarchies
+			\vspace{1em}
+		\item Other requirements of the POSIX spec
+	\end{itemize}
+	\vspace{1em}
+	Coordination in a distributed system is costly
+
+	\vspace{1em}
+	Costs explode with commodity hardware / Internet connections\\
+	{\small (we experienced this!)}
+\end{frame}
+
+\begin{frame}
+	\frametitle{A simpler solution: object storage}
+	Only two operations:
+	\vspace{1em}
+	\begin{itemize}
+		\item Put an object at a key
+			\vspace{1em}
+		\item Retrieve an object from its key
+	\end{itemize}
+	\vspace{1em}
+	{\footnotesize (and a few others)}
+
+	\vspace{1em}
+	Sufficient for many applications!
+\end{frame}
+
+\begin{frame}
+	\frametitle{A simpler solution: object storage}
+		\begin{center}
+			\includegraphics[width=.2\linewidth]{../2020-12-02_wide-team/img/Amazon-S3.jpg}
+			\hspace{5em}
+			\includegraphics[width=.2\linewidth]{assets/minio.png}
+		\end{center}
+		\vspace{1em}
+	S3: a de-facto standard, many compatible applications
+
+	\vspace{1em}
+
+	MinIO is self-hostable but not suited for geo-distributed deployments
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{But what is Garage, exactly?}
+	\textbf{Garage is a self-hosted drop-in replacement for the Amazon S3 object store}\\
+	\vspace{.5em}
+	that implements resilience through geographical redundancy on commodity hardware
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{assets/garageuses.png}
+		\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Overview}
+	\begin{center}
+		\only<1>{\includegraphics[width=.45\linewidth]{assets/garage2a.drawio.pdf}}%
+		\only<2>{\includegraphics[width=.45\linewidth]{assets/garage2b.drawio.pdf}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Garage is \emph{location-aware}}
+	\begin{center}
+		\includegraphics[width=\linewidth]{assets/location-aware.png}
+	\end{center}
+	\vspace{2em}
+	Garage replicates data on different zones when possible
+\end{frame}
+
+\begin{frame}
+	\frametitle{Garage is \emph{location-aware}}
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{assets/map.png}
+		\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to spread files over different cluster nodes?}
+	\textbf{Consistent hashing (DynamoDB):}
+	\vspace{1em}
+
+	\begin{center}
+	\only<1>{\includegraphics[width=.45\columnwidth]{assets/consistent_hashing_1.pdf}}%
+	\only<2>{\includegraphics[width=.45\columnwidth]{assets/consistent_hashing_2.pdf}}%
+	\only<3>{\includegraphics[width=.45\columnwidth]{assets/consistent_hashing_3.pdf}}%
+	\only<4>{\includegraphics[width=.45\columnwidth]{assets/consistent_hashing_4.pdf}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to spread files over different cluster nodes?}
+	\textbf{Issues with consistent hashing:}
+	\vspace{1em}
+	\begin{itemize}
+		\item Doesn't dispatch data based on geographical location of nodes
+			\vspace{1em}
+		\item<2-> Geographically aware adaptation, try 1:\\
+			data quantities not well balanced between nodes
+			\vspace{1em}
+		\item<3-> Geographically aware adaptation, try 2:\\
+			too many reshuffles when adding/removing nodes
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to spread files over different cluster nodes?}
+	\textbf{Garage's method: build an index table}
+	\vspace{1em}
+
+	Realization: we can actually precompute an optimal solution
+	\vspace{1em}
+
+	\visible<2->{
+		\begin{center}
+		\begin{tabular}{|l|l|l|l|}
+			\hline
+			\textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+			\hline
+			\hline
+			Partition 0 & Io (jupiter) & Drosera (atuin) & Courgette (neptune) \\
+			\hline
+			Partition 1 & Datura (atuin) & Courgette (neptune) & Io (jupiter) \\
+			\hline
+			Partition 2 & Io(jupiter) & Celeri (neptune) & Drosera (atuin) \\
+			\hline
+			\hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ \\
+			\hline
+			Partition 255 & Concombre (neptune) & Io (jupiter) & Drosera (atuin) \\
+			\hline
+		\end{tabular}
+		\end{center}
+	}
+	\vspace{1em}
+	\visible<3->{
+		The index table is built centrally using an optimal* algorithm,\\
+		then propagated to all nodes\\
+		\hfill\footnotesize *not yet optimal but will be soon
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Garage's internal data structures}
+	\centering
+	\includegraphics[width=.75\columnwidth]{assets/garage_tables.pdf}
+\end{frame}
+
+%\begin{frame}
+%	\frametitle{Garage's architecture}
+%	\begin{center}
+%		\includegraphics[width=.35\linewidth]{assets/garage.drawio.pdf}
+%	\end{center}
+%\end{frame}
+
+\begin{frame}
+	\frametitle{Garage is \emph{coordination-free}:}
+	\begin{itemize}
+		\item No Raft or Paxos
+			\vspace{1em}
+		\item Internal data types are CRDTs
+			\vspace{1em}
+		\item All nodes are equivalent (no master/leader/index node)
+	\end{itemize}
+			\vspace{2em}
+	$\to$ less sensitive to higher latencies between nodes
+\end{frame}
+
+\begin{frame}
+	\frametitle{Consistency model}
+	\begin{itemize}
+		\item Not ACID (not required by S3 spec) / not linearizable
+			\vspace{1em}
+		\item \textbf{Read-after-write consistency}\\
+			{\footnotesize (stronger than eventual consistency)}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Impact on performances}
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{assets/endpoint-latency-dc.png}
+		\end{center}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{An ever-increasing compatibility list}
+		\begin{center}
+			\includegraphics[width=.7\linewidth]{assets/compatibility.png}
+		\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Further plans for Garage}
+	\begin{center}
+		\only<1>{\includegraphics[width=.8\linewidth]{assets/slideB1.png}}%
+		\only<2>{\includegraphics[width=.8\linewidth]{assets/slideB2.png}}%
+		\only<3>{\includegraphics[width=.8\linewidth]{assets/slideB3.png}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{K2V Design}
+	\begin{itemize}
+		\item A new, custom, minimal API
+			\vspace{1em}
+		\item<2-> Exposes the partitoning mechanism of Garage\\
+			K2V = partition key / sort key / value (like Dynamo)
+			\vspace{1em}
+		\item<3-> Coordination-free, CRDT-friendly (inspired by Riak)\\
+			\vspace{1em}
+		\item<4-> Cryptography-friendly: values are binary blobs
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Application: an e-mail storage server}
+	\begin{center}
+		\only<1>{\includegraphics[width=.9\linewidth]{assets/aerogramme.png}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Aerogramme data model}
+	\begin{center}
+		\only<1>{\includegraphics[width=.4\linewidth]{assets/aerogramme_datatype.drawio.pdf}}%
+		\only<2->{\includegraphics[width=.9\linewidth]{assets/aerogramme_keys.drawio.pdf}\vspace{1em}}%
+	\end{center}
+	\visible<3->{Aerogramme encrypts all stored values for privacy\\
+	(Garage server administrators can't read your mail)}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Different deployment scenarios}
+	\begin{center}
+		\only<1>{\includegraphics[width=.9\linewidth]{assets/aerogramme_components1.drawio.pdf}}%
+		\only<2>{\includegraphics[width=.9\linewidth]{assets/aerogramme_components2.drawio.pdf}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{A new model for building resilient software}
+	\begin{itemize}
+		\item Design a data model suited to K2V\\
+			{\footnotesize (see Cassandra docs on porting SQL data models to Cassandra)}
+			\vspace{1em}
+			\begin{itemize}
+				\item Use CRDTs or other eventually consistent data types (see e.g. Bayou)
+					\vspace{1em}
+				\item Store opaque binary blobs to provide End-to-End Encryption\\
+			\end{itemize}
+			\vspace{1em}
+		\item Store big blobs (files) in S3
+			\vspace{1em}
+		\item Let Garage manage sharding, replication, failover, etc.
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Research perspectives}
+	\begin{itemize}
+		\item Write about Garage's global architecture \emph{(paper in progress)}
+			\vspace{1em}
+		\item Measure and improve Garage's performances
+			\vspace{1em}
+		\item Discuss the optimal layout algorithm, provide proofs
+			\vspace{1em}
+		\item Write about our proposed architecture for (E2EE) apps over K2V+S3
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Where to find us}
+	\begin{center}
+			\includegraphics[width=.25\linewidth]{../../logo/garage_hires.png}\\
+			\vspace{-1em}
+		\url{https://garagehq.deuxfleurs.fr/}\\
+		\url{mailto:garagehq@deuxfleurs.fr}\\
+		\texttt{\#garage:deuxfleurs.fr} on Matrix
+
+		\vspace{1.5em}
+			\includegraphics[width=.06\linewidth]{assets/rust_logo.png}
+			\includegraphics[width=.13\linewidth]{assets/AGPLv3_Logo.png}
+	\end{center}
+\end{frame}
+
+\end{document}
+
+%% vim: set ts=4 sw=4 tw=0 noet spelllang=en :

nix/common.nix

@@ -4,18 +4,16 @@ rec {
    */
   pkgsSrc = fetchTarball {
     # As of 2021-10-04
-    url ="https://github.com/NixOS/nixpkgs/archive/b27d18a412b071f5d7991d1648cfe78ee7afe68a.tar.gz";
+    url = "https://github.com/NixOS/nixpkgs/archive/b27d18a412b071f5d7991d1648cfe78ee7afe68a.tar.gz";
     sha256 = "1xy9zpypqfxs5gcq5dcla4bfkhxmh5nzn9dyqkr03lqycm9wg5cr";
   };
   cargo2nixSrc = fetchGit {
-    # As of 2022-03-17
-    url = "https://github.com/superboum/cargo2nix";
-    ref = "main";
-    rev = "bcbf3ba99e9e01a61eb83a24624419c2dd9dec64";
+    # As of 2022-08-29, stacking two patches: superboum@dedup_propagate and Alexis211@fix_fetchcrategit
+    url = "https://github.com/Alexis211/cargo2nix";
+    ref = "fix_fetchcrategit";
+    rev = "4b31c0cc05b6394916d46e9289f51263d81973b9";
   };
 
-
-
   /*
    * Shared objects
    */

nix/compile.nix

@@ -0,0 +1,244 @@
+{
+  system ? builtins.currentSystem,
+  target ? null,
+  compiler ? "rustc",
+  release ? false,
+  git_version ? null,
+}:
+
+with import ./common.nix;
+
+let
+  log = v: builtins.trace v v;
+
+  pkgs = import pkgsSrc {
+    inherit system; 
+    ${ if target == null then null else "crossSystem" } = { config = target; };
+    overlays = [ cargo2nixOverlay ];
+  };
+
+  /*
+   Rust and Nix triples are not the same. Cargo2nix has a dedicated library
+   to convert Nix triples to Rust ones. We need this conversion as we want to
+   set later options linked to our (rust) target in a generic way. Not only
+   the triple terminology is different, but also the "roles" are named differently.
+   Nix uses a build/host/target terminology where Nix's "host" maps to Cargo's "target".
+  */
+  rustTarget = log (pkgs.rustBuilder.rustLib.rustTriple pkgs.stdenv.hostPlatform);
+
+  /*
+   Cargo2nix is built for rustOverlay which installs Rust from Mozilla releases.
+   We want our own Rust to avoid incompatibilities, like we had with musl 1.2.0.
+   rustc was built with musl < 1.2.0 and nix shipped musl >= 1.2.0 which lead to compilation breakage.
+   So we want a Rust release that is bound to our Nix repository to avoid these problems.
+   See here for more info: https://musl.libc.org/time64.html
+   Because Cargo2nix does not support the Rust environment shipped by NixOS,
+   we emulate the structure of the Rust object created by rustOverlay.
+   In practise, rustOverlay ships rustc+cargo in a single derivation while
+   NixOS ships them in separate ones. We reunite them with symlinkJoin.
+   */
+  rustChannel = {
+    rustc = pkgs.symlinkJoin {
+      name = "rust-channel";
+      paths = [ 
+        pkgs.rustPlatform.rust.cargo
+        pkgs.rustPlatform.rust.rustc
+      ];
+    };
+    clippy = pkgs.symlinkJoin {
+      name = "clippy-channel";
+      paths = [ 
+        pkgs.rustPlatform.rust.cargo
+        pkgs.rustPlatform.rust.rustc
+        pkgs.clippy
+      ];
+    };
+  }.${compiler};
+
+  clippyBuilder = pkgs.writeScriptBin "clippy" ''
+    #!${pkgs.stdenv.shell}
+    . ${cargo2nixSrc + "/overlay/utils.sh"}
+    isBuildScript=
+    args=("$@")
+    for i in "''${!args[@]}"; do
+      if [ "xmetadata=" = "x''${args[$i]::9}" ]; then
+        args[$i]=metadata=$NIX_RUST_METADATA
+      elif [ "x--crate-name" = "x''${args[$i]}" ] && [ "xbuild_script_" = "x''${args[$i+1]::13}" ]; then
+        isBuildScript=1
+      fi
+    done
+    if [ "$isBuildScript" ]; then
+      args+=($NIX_RUST_BUILD_LINK_FLAGS)
+    else
+      args+=($NIX_RUST_LINK_FLAGS)
+    fi
+    touch invoke.log
+    echo "''${args[@]}" >>invoke.log
+
+    exec ${rustChannel}/bin/clippy-driver --deny warnings "''${args[@]}"
+  '';
+
+  buildEnv = (drv: {
+    rustc = drv.setBuildEnv;
+    clippy = ''
+      ${drv.setBuildEnv or "" }
+      echo
+      echo --- BUILDING WITH CLIPPY ---
+      echo 
+
+      export RUSTC=${clippyBuilder}/bin/clippy
+    '';
+  }.${compiler});
+
+  /*
+   Cargo2nix provides many overrides by default, you can take inspiration from them:
+   https://github.com/cargo2nix/cargo2nix/blob/master/overlay/overrides.nix
+
+   You can have a complete list of the available options by looking at the overriden object, mkcrate:
+   https://github.com/cargo2nix/cargo2nix/blob/master/overlay/mkcrate.nix
+  */
+  overrides = pkgs.rustBuilder.overrides.all ++ [
+    /*
+     [1] We add some logic to compile our crates with clippy, it provides us many additional lints
+
+     [2] We need to alter Nix hardening to make static binaries: PIE,
+     Position Independent Executables seems to be supported only on amd64. Having
+     this flag set either 1. make our executables crash or 2. compile as dynamic on some platforms.
+     Here, we deactivate it. Later (find `codegenOpts`), we reactivate it for supported targets
+     (only amd64 curently) through the `-static-pie` flag.
+     PIE is a feature used by ASLR, which helps mitigate security issues.
+     Learn more about Nix Hardening at: https://github.com/NixOS/nixpkgs/blob/master/pkgs/build-support/cc-wrapper/add-hardening.sh
+
+     [3] We want to inject the git version while keeping the build deterministic.
+     As we do not want to consider the .git folder as part of the input source,
+     we ask the user (the CI often) to pass the value to Nix.
+
+     [4] We ship some parts of the code disabled by default by putting them behind a flag.
+     It speeds up the compilation (when the feature is not required) and released crates have less dependency by default (less attack surface, disk space, etc.).
+     But we want to ship these additional features when we release Garage.
+     In the end, we chose to exclude all features from debug builds while putting (all of) them in the release builds.
+
+     [5] We don't want libsodium-sys and zstd-sys to try to use pkgconfig to build against a system library.
+     However the features to do so get activated for some reason (due to a bug in cargo2nix?),
+     so disable them manually here.
+    */
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage";
+      overrideAttrs = drv:
+        (if git_version != null then {
+          /* [3] */ preConfigure = ''
+            ${drv.preConfigure or ""}
+            export GIT_VERSION="${git_version}"
+          '';
+        } else {})
+        //
+      {
+        /* [1] */ setBuildEnv = (buildEnv drv);
+        /* [2] */ hardeningDisable = [ "pie" ];
+      };
+      overrideArgs = old: {
+        /* [4] */ features = [ "bundled-libs" "sled" "k2v" ]
+          ++ (if release then [ "kubernetes-discovery" "telemetry-otlp" "metrics" "lmdb" "sqlite" ] else []);
+      };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_rpc";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_db";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_util";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_table";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_block";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_model";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_api";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "garage_web";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "k2v-client";
+      overrideAttrs = drv: { /* [1] */ setBuildEnv = (buildEnv drv); };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "libsodium-sys";
+      overrideArgs = old: {
+        features = [ ]; /* [5] */
+      };
+    })
+
+    (pkgs.rustBuilder.rustLib.makeOverride {
+      name = "zstd-sys";
+      overrideArgs = old: {
+        features = [ ]; /* [5] */
+      };
+    })
+  ];
+
+  packageFun = import ../Cargo.nix;
+
+  /*
+    We compile fully static binaries with musl to simplify deployment on most systems.
+    When possible, we reactivate PIE hardening (see above).
+
+    Also, if you set the RUSTFLAGS environment variable, the following parameters will
+    be ignored.
+
+    For more information on static builds, please refer to Rust's RFC 1721.
+    https://rust-lang.github.io/rfcs/1721-crt-static.html#specifying-dynamicstatic-c-runtime-linkage
+  */
+
+  codegenOpts = {
+   "armv6l-unknown-linux-musleabihf" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* compile as dynamic with static-pie */
+   "aarch64-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* segfault with static-pie */
+   "i686-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static" ]; /* segfault with static-pie */
+   "x86_64-unknown-linux-musl" = [ "target-feature=+crt-static" "link-arg=-static-pie" ];
+  };
+
+in
+  /*
+   The following definition is not elegant as we use a low level function of Cargo2nix
+   that enables us to pass our custom rustChannel object. We need this low level definition
+   to pass Nix's Rust toolchains instead of Mozilla's one.
+
+   target is mandatory but must be kept to null to allow cargo2nix to set it to the appropriate value
+   for each crate.
+  */ 
+  pkgs.rustBuilder.makePackageSet {
+    inherit packageFun rustChannel release codegenOpts;
+    packageOverrides = overrides;
+    target = null;
+
+    buildRustPackages = pkgs.buildPackages.rustBuilder.makePackageSet {
+      inherit rustChannel packageFun codegenOpts;
+      packageOverrides = overrides;
+      target = null;
+    };
+  }

nix/nix.conf

@@ -1,4 +1,9 @@
 substituters = https://cache.nixos.org https://nix.web.deuxfleurs.fr
 trusted-public-keys = cache.nixos.org-1:6NCHdD59X431o0gWypbMrAURkbJ16ZPMQFGspcDShjY= nix.web.deuxfleurs.fr:eTGL6kvaQn6cDR/F9lDYUIP9nCVR/kkshYfLDJf1yKs=
 max-jobs = auto
-cores = 4
+cores = 0
+log-lines = 200
+filter-syscalls = false
+sandbox = false
+keep-outputs = true
+keep-derivations = true

script/dev-cluster.sh

@@ -11,7 +11,7 @@ PATH="${GARAGE_DEBUG}:${GARAGE_RELEASE}:${NIX_RELEASE}:$PATH"
 FANCYCOLORS=("41m" "42m" "44m" "45m" "100m" "104m")
 
 export RUST_BACKTRACE=1 
-export RUST_LOG=garage=info,garage_api=debug
+export RUST_LOG=garage=info,garage_api=debug,netapp=trace
 MAIN_LABEL="\e[${FANCYCOLORS[0]}[main]\e[49m"
 
 WHICH_GARAGE=$(which garage || exit 1)

script/not-dynamic.sh

@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+set -e
+
+if [ "$#" -ne 1 ]; then
+    echo "[fail] usage: $0 binary"
+    exit 2
+fi
+
+if file $1 | grep 'dynamically linked' 2>&1; then
+    echo "[fail] $1 is dynamic"
+    exit 1
+fi
+echo "[ok] $1 is probably static"

shell.nix

@@ -1,8 +1,5 @@
 {
   system ? builtins.currentSystem,
-  rust ? true,
-  integration ? true,
-  release ? true,
 }:
 
 with import ./nix/common.nix;
@@ -16,9 +13,59 @@ let
   winscp = (import ./nix/winscp.nix) pkgs;
 
 in
+  {
 
-pkgs.mkShell {
-  shellHook = ''
+    /* --- Rust Shell ---
+     * Use it to compile Garage
+     */
+    rust = pkgs.mkShell {
+      shellHook = ''
+function refresh_toolchain {
+  nix copy \
+    --to 's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/signing-key.sec' \
+    $(nix-store -qR \
+      $(nix-build --quiet --no-build-output --no-out-link nix/toolchain.nix))
+}
+      '';
+
+      nativeBuildInputs = [
+        #pkgs.rustPlatform.rust.rustc
+        pkgs.rustPlatform.rust.cargo
+        #pkgs.clippy
+        pkgs.rustfmt
+        #pkgs.perl
+        #pkgs.protobuf
+        #pkgs.pkg-config
+        #pkgs.openssl
+        pkgs.file
+        #cargo2nix.packages.x86_64-linux.cargo2nix
+      ];
+    };
+
+    /* --- Integration shell ---
+     * Use it to test Garage with common S3 clients
+     */
+    integration = pkgs.mkShell {
+      nativeBuildInputs = [
+        winscp
+        pkgs.s3cmd
+        pkgs.awscli2
+        pkgs.minio-client
+        pkgs.rclone
+        pkgs.socat
+        pkgs.psmisc
+        pkgs.which
+        pkgs.openssl
+        pkgs.curl
+        pkgs.jq
+      ];
+    };
+
+    /* --- Release shell ---
+     * A shell built to make releasing easier
+     */
+    release = pkgs.mkShell {
+      shellHook = ''
 function to_s3 {
   aws \
       --endpoint-url https://garage.deuxfleurs.fr \
@@ -62,45 +109,12 @@ function refresh_index {
       result/share/_releases.html \
       s3://garagehq.deuxfleurs.fr/
 }
+        '';
+      nativeBuildInputs = [
+        pkgs.awscli2
+        kaniko 
+      ];
+    };
+  }
 
-function refresh_toolchain {
-  nix copy \
-    --to 's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/signing-key.sec' \
-    $(nix-store -qR \
-      $(nix-build --quiet --no-build-output --no-out-link nix/toolchain.nix))
-}
-  '';
 
-  nativeBuildInputs = 
-   (if rust then [
-     pkgs.rustPlatform.rust.rustc
-     pkgs.rustPlatform.rust.cargo
-     pkgs.clippy
-     pkgs.rustfmt
-     pkgs.perl
-     pkgs.protobuf
-     pkgs.pkg-config
-     pkgs.openssl
-     cargo2nix.packages.x86_64-linux.cargo2nix
-    ] else [])
-   ++
-   (if integration then [
-     winscp
-     pkgs.s3cmd
-     pkgs.awscli2
-     pkgs.minio-client
-     pkgs.rclone
-     pkgs.socat
-     pkgs.psmisc
-     pkgs.which
-     pkgs.openssl
-     pkgs.curl
-     pkgs.jq
-    ] else [])
-   ++
-   (if release then [ 
-     pkgs.awscli2
-     kaniko 
-    ] else [])
-   ;
-}

src/api/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_api"
-version = "0.7.0"
+version = "0.8.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -14,25 +14,25 @@ path = "lib.rs"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-garage_model = { version = "0.7.0", path = "../model" }
-garage_table = { version = "0.7.0", path = "../table" }
-garage_block = { version = "0.7.0", path = "../block" }
-garage_util = { version = "0.7.0", path = "../util" }
-garage_rpc = { version = "0.7.0", path = "../rpc" }
+garage_model = { version = "0.8.0", path = "../model" }
+garage_table = { version = "0.8.0", path = "../table" }
+garage_block = { version = "0.8.0", path = "../block" }
+garage_util = { version = "0.8.0", path = "../util" }
+garage_rpc = { version = "0.8.0", path = "../rpc" }
 
 async-trait = "0.1.7"
 base64 = "0.13"
 bytes = "1.0"
 chrono = "0.4"
-crypto-mac = "0.10"
+crypto-common = "0.1"
 err-derive = "0.3"
 hex = "0.4"
-hmac = "0.10"
+hmac = "0.12"
 idna = "0.2"
 tracing = "0.1.30"
-md-5 = "0.9"
+md-5 = "0.10"
 nom = "7.1"
-sha2 = "0.9"
+sha2 = "0.10"
 
 futures = "0.3"
 futures-util = "0.3"
@@ -54,9 +54,9 @@ quick-xml = { version = "0.21", features = [ "serialize" ] }
 url = "2.1"
 
 opentelemetry = "0.17"
-opentelemetry-prometheus = "0.10"
-opentelemetry-otlp = "0.10"
-prometheus = "0.13"
+opentelemetry-prometheus = { version = "0.10", optional = true }
+prometheus = { version = "0.13", optional = true }
 
 [features]
 k2v = [ "garage_util/k2v", "garage_model/k2v" ]
+metrics = [ "opentelemetry-prometheus", "prometheus" ]

src/api/admin/api_server.rs

@@ -1,15 +1,17 @@
+use std::net::SocketAddr;
 use std::sync::Arc;
 
 use async_trait::async_trait;
 
 use futures::future::Future;
-use http::header::{
-	ACCESS_CONTROL_ALLOW_METHODS, ACCESS_CONTROL_ALLOW_ORIGIN, ALLOW, CONTENT_TYPE,
-};
+use http::header::{ACCESS_CONTROL_ALLOW_METHODS, ACCESS_CONTROL_ALLOW_ORIGIN, ALLOW};
 use hyper::{Body, Request, Response};
 
-use opentelemetry::trace::{SpanRef, Tracer};
+use opentelemetry::trace::SpanRef;
+
+#[cfg(feature = "metrics")]
 use opentelemetry_prometheus::PrometheusExporter;
+#[cfg(feature = "metrics")]
 use prometheus::{Encoder, TextEncoder};
 
 use garage_model::garage::Garage;
@@ -25,6 +27,7 @@ use crate::admin::router::{Authorization, Endpoint};
 
 pub struct AdminApiServer {
 	garage: Arc<Garage>,
+	#[cfg(feature = "metrics")]
 	exporter: PrometheusExporter,
 	metrics_token: Option<String>,
 	admin_token: Option<String>,
@@ -32,7 +35,6 @@ pub struct AdminApiServer {
 
 impl AdminApiServer {
 	pub fn new(garage: Arc<Garage>) -> Self {
-		let exporter = opentelemetry_prometheus::exporter().init();
 		let cfg = &garage.config.admin;
 		let metrics_token = cfg
 			.metrics_token
@@ -44,21 +46,22 @@ impl AdminApiServer {
 			.map(|tok| format!("Bearer {}", tok));
 		Self {
 			garage,
-			exporter,
+			#[cfg(feature = "metrics")]
+			exporter: opentelemetry_prometheus::exporter().init(),
 			metrics_token,
 			admin_token,
 		}
 	}
 
-	pub async fn run(self, shutdown_signal: impl Future<Output = ()>) -> Result<(), GarageError> {
-		if let Some(bind_addr) = self.garage.config.admin.api_bind_addr {
-			let region = self.garage.config.s3_api.s3_region.clone();
-			ApiServer::new(region, self)
-				.run_server(bind_addr, shutdown_signal)
-				.await
-		} else {
-			Ok(())
-		}
+	pub async fn run(
+		self,
+		bind_addr: SocketAddr,
+		shutdown_signal: impl Future<Output = ()>,
+	) -> Result<(), GarageError> {
+		let region = self.garage.config.s3_api.s3_region.clone();
+		ApiServer::new(region, self)
+			.run_server(bind_addr, shutdown_signal)
+			.await
 	}
 
 	fn handle_options(&self, _req: &Request<Body>) -> Result<Response<Body>, Error> {
@@ -71,22 +74,31 @@ impl AdminApiServer {
 	}
 
 	fn handle_metrics(&self) -> Result<Response<Body>, Error> {
-		let mut buffer = vec![];
-		let encoder = TextEncoder::new();
+		#[cfg(feature = "metrics")]
+		{
+			use opentelemetry::trace::Tracer;
 
-		let tracer = opentelemetry::global::tracer("garage");
-		let metric_families = tracer.in_span("admin/gather_metrics", |_| {
-			self.exporter.registry().gather()
-		});
+			let mut buffer = vec![];
+			let encoder = TextEncoder::new();
 
-		encoder
-			.encode(&metric_families, &mut buffer)
-			.ok_or_internal_error("Could not serialize metrics")?;
+			let tracer = opentelemetry::global::tracer("garage");
+			let metric_families = tracer.in_span("admin/gather_metrics", |_| {
+				self.exporter.registry().gather()
+			});
 
-		Ok(Response::builder()
-			.status(200)
-			.header(CONTENT_TYPE, encoder.format_type())
-			.body(Body::from(buffer))?)
+			encoder
+				.encode(&metric_families, &mut buffer)
+				.ok_or_internal_error("Could not serialize metrics")?;
+
+			Ok(Response::builder()
+				.status(200)
+				.header(http::header::CONTENT_TYPE, encoder.format_type())
+				.body(Body::from(buffer))?)
+		}
+		#[cfg(not(feature = "metrics"))]
+		Err(Error::bad_request(
+			"Garage was built without the metrics feature".to_string(),
+		))
 	}
 }
 
@@ -156,12 +168,7 @@ impl ApiHandler for AdminApiServer {
 			}
 			Endpoint::CreateBucket => handle_create_bucket(&self.garage, req).await,
 			Endpoint::DeleteBucket { id } => handle_delete_bucket(&self.garage, id).await,
-			Endpoint::PutBucketWebsite { id } => {
-				handle_put_bucket_website(&self.garage, id, req).await
-			}
-			Endpoint::DeleteBucketWebsite { id } => {
-				handle_delete_bucket_website(&self.garage, id).await
-			}
+			Endpoint::UpdateBucket { id } => handle_update_bucket(&self.garage, id, req).await,
 			// Bucket-key permissions
 			Endpoint::BucketAllowKey => {
 				handle_bucket_change_key_perm(&self.garage, req, true).await

src/api/admin/bucket.rs

@@ -14,6 +14,7 @@ use garage_model::bucket_alias_table::*;
 use garage_model::bucket_table::*;
 use garage_model::garage::Garage;
 use garage_model::permission::*;
+use garage_model::s3::object_table::*;
 
 use crate::admin::error::*;
 use crate::admin::key::ApiBucketKeyPerm;
@@ -77,6 +78,13 @@ struct BucketLocalAlias {
 	alias: String,
 }
 
+#[derive(Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+struct ApiBucketQuotas {
+	max_size: Option<u64>,
+	max_objects: Option<u64>,
+}
+
 pub async fn handle_get_bucket_info(
 	garage: &Arc<Garage>,
 	id: Option<String>,
@@ -108,6 +116,14 @@ async fn bucket_info_results(
 		.get_existing_bucket(bucket_id)
 		.await?;
 
+	let counters = garage
+		.object_counter_table
+		.table
+		.get(&bucket_id, &EmptyKey)
+		.await?
+		.map(|x| x.filtered_values(&garage.system.ring.borrow()))
+		.unwrap_or_default();
+
 	let mut relevant_keys = HashMap::new();
 	for (k, _) in bucket
 		.state
@@ -148,6 +164,7 @@ async fn bucket_info_results(
 
 	let state = bucket.state.as_option().unwrap();
 
+	let quotas = state.quotas.get();
 	let res =
 		GetBucketInfoResult {
 			id: hex::encode(&bucket.id),
@@ -191,6 +208,16 @@ async fn bucket_info_results(
 					}
 				})
 				.collect::<Vec<_>>(),
+			objects: counters.get(OBJECTS).cloned().unwrap_or_default(),
+			bytes: counters.get(BYTES).cloned().unwrap_or_default(),
+			unfinshed_uploads: counters
+				.get(UNFINISHED_UPLOADS)
+				.cloned()
+				.unwrap_or_default(),
+			quotas: ApiBucketQuotas {
+				max_size: quotas.max_size,
+				max_objects: quotas.max_objects,
+			},
 		};
 
 	Ok(json_ok_response(&res)?)
@@ -205,6 +232,10 @@ struct GetBucketInfoResult {
 	#[serde(default)]
 	website_config: Option<GetBucketInfoWebsiteResult>,
 	keys: Vec<GetBucketInfoKey>,
+	objects: i64,
+	bytes: i64,
+	unfinshed_uploads: i64,
+	quotas: ApiBucketQuotas,
 }
 
 #[derive(Serialize)]
@@ -363,14 +394,12 @@ pub async fn handle_delete_bucket(
 		.body(Body::empty())?)
 }
 
-// ---- BUCKET WEBSITE CONFIGURATION ----
-
-pub async fn handle_put_bucket_website(
+pub async fn handle_update_bucket(
 	garage: &Arc<Garage>,
 	id: String,
 	req: Request<Body>,
 ) -> Result<Response<Body>, Error> {
-	let req = parse_json_body::<PutBucketWebsiteRequest>(req).await?;
+	let req = parse_json_body::<UpdateBucketRequest>(req).await?;
 	let bucket_id = parse_bucket_id(&id)?;
 
 	let mut bucket = garage
@@ -379,10 +408,31 @@ pub async fn handle_put_bucket_website(
 		.await?;
 
 	let state = bucket.state.as_option_mut().unwrap();
-	state.website_config.update(Some(WebsiteConfig {
-		index_document: req.index_document,
-		error_document: req.error_document,
-	}));
+
+	if let Some(wa) = req.website_access {
+		if wa.enabled {
+			state.website_config.update(Some(WebsiteConfig {
+				index_document: wa.index_document.ok_or_bad_request(
+					"Please specify indexDocument when enabling website access.",
+				)?,
+				error_document: wa.error_document,
+			}));
+		} else {
+			if wa.index_document.is_some() || wa.error_document.is_some() {
+				return Err(Error::bad_request(
+					"Cannot specify indexDocument or errorDocument when disabling website access.",
+				));
+			}
+			state.website_config.update(None);
+		}
+	}
+
+	if let Some(q) = req.quotas {
+		state.quotas.update(BucketQuotas {
+			max_size: q.max_size,
+			max_objects: q.max_objects,
+		});
+	}
 
 	garage.bucket_table.insert(&bucket).await?;
 
@@ -391,29 +441,17 @@ pub async fn handle_put_bucket_website(
 
 #[derive(Deserialize)]
 #[serde(rename_all = "camelCase")]
-struct PutBucketWebsiteRequest {
-	index_document: String,
-	#[serde(default)]
-	error_document: Option<String>,
+struct UpdateBucketRequest {
+	website_access: Option<UpdateBucketWebsiteAccess>,
+	quotas: Option<ApiBucketQuotas>,
 }
 
-pub async fn handle_delete_bucket_website(
-	garage: &Arc<Garage>,
-	id: String,
-) -> Result<Response<Body>, Error> {
-	let bucket_id = parse_bucket_id(&id)?;
-
-	let mut bucket = garage
-		.bucket_helper()
-		.get_existing_bucket(bucket_id)
-		.await?;
-
-	let state = bucket.state.as_option_mut().unwrap();
-	state.website_config.update(None);
-
-	garage.bucket_table.insert(&bucket).await?;
-
-	bucket_info_results(garage, bucket_id).await
+#[derive(Deserialize)]
+#[serde(rename_all = "camelCase")]
+struct UpdateBucketWebsiteAccess {
+	enabled: bool,
+	index_document: Option<String>,
+	error_document: Option<String>,
 }
 
 // ---- BUCKET/KEY PERMISSIONS ----

src/api/admin/cluster.rs

@@ -18,7 +18,9 @@ use crate::helpers::{json_ok_response, parse_json_body};
 pub async fn handle_get_cluster_status(garage: &Arc<Garage>) -> Result<Response<Body>, Error> {
 	let res = GetClusterStatusResponse {
 		node: hex::encode(garage.system.id),
-		garage_version: garage.system.garage_version(),
+		garage_version: garage_util::version::garage_version(),
+		garage_features: garage_util::version::garage_features(),
+		db_engine: garage.db.engine(),
 		known_nodes: garage
 			.system
 			.get_known_nodes()
@@ -98,6 +100,8 @@ fn get_cluster_layout(garage: &Arc<Garage>) -> GetClusterLayoutResponse {
 struct GetClusterStatusResponse {
 	node: String,
 	garage_version: &'static str,
+	garage_features: Option<&'static [&'static str]>,
+	db_engine: String,
 	known_nodes: HashMap<String, KnownNodeResp>,
 	layout: GetClusterLayoutResponse,
 }

src/api/admin/router.rs

@@ -48,10 +48,7 @@ pub enum Endpoint {
 	DeleteBucket {
 		id: String,
 	},
-	PutBucketWebsite {
-		id: String,
-	},
-	DeleteBucketWebsite {
+	UpdateBucket {
 		id: String,
 	},
 	// Bucket-Key Permissions
@@ -113,8 +110,7 @@ impl Endpoint {
 			GET "/v0/bucket" => ListBuckets,
 			POST "/v0/bucket" => CreateBucket,
 			DELETE "/v0/bucket" if id => DeleteBucket (query::id),
-			PUT "/v0/bucket/website" if id => PutBucketWebsite (query::id),
-			DELETE "/v0/bucket/website" if id => DeleteBucketWebsite (query::id),
+			PUT "/v0/bucket" if id => UpdateBucket (query::id),
 			// Bucket-key permissions
 			POST "/v0/bucket/allow" => BucketAllowKey,
 			POST "/v0/bucket/deny" => BucketDenyKey,

src/api/k2v/api_server.rs

@@ -1,3 +1,4 @@
+use std::net::SocketAddr;
 use std::sync::Arc;
 
 use async_trait::async_trait;
@@ -36,20 +37,13 @@ pub(crate) struct K2VApiEndpoint {
 impl K2VApiServer {
 	pub async fn run(
 		garage: Arc<Garage>,
+		bind_addr: SocketAddr,
+		s3_region: String,
 		shutdown_signal: impl Future<Output = ()>,
 	) -> Result<(), GarageError> {
-		if let Some(cfg) = &garage.config.k2v_api {
-			let bind_addr = cfg.api_bind_addr;
-
-			ApiServer::new(
-				garage.config.s3_api.s3_region.clone(),
-				K2VApiServer { garage },
-			)
+		ApiServer::new(s3_region, K2VApiServer { garage })
 			.run_server(bind_addr, shutdown_signal)
 			.await
-		} else {
-			Ok(())
-		}
 	}
 }
 

src/api/k2v/index.rs

@@ -10,7 +10,7 @@ use garage_rpc::ring::Ring;
 use garage_table::util::*;
 
 use garage_model::garage::Garage;
-use garage_model::k2v::counter_table::{BYTES, CONFLICTS, ENTRIES, VALUES};
+use garage_model::k2v::item_table::{BYTES, CONFLICTS, ENTRIES, VALUES};
 
 use crate::k2v::error::*;
 use crate::k2v::range::read_range;

src/api/s3/api_server.rs

@@ -1,10 +1,11 @@
+use std::net::SocketAddr;
 use std::sync::Arc;
 
 use async_trait::async_trait;
 
 use futures::future::Future;
 use hyper::header;
-use hyper::{Body, Method, Request, Response};
+use hyper::{Body, Request, Response};
 
 use opentelemetry::{trace::SpanRef, KeyValue};
 
@@ -43,16 +44,13 @@ pub(crate) struct S3ApiEndpoint {
 impl S3ApiServer {
 	pub async fn run(
 		garage: Arc<Garage>,
+		addr: SocketAddr,
+		s3_region: String,
 		shutdown_signal: impl Future<Output = ()>,
 	) -> Result<(), GarageError> {
-		let addr = garage.config.s3_api.api_bind_addr;
-
-		ApiServer::new(
-			garage.config.s3_api.s3_region.clone(),
-			S3ApiServer { garage },
-		)
-		.run_server(addr, shutdown_signal)
-		.await
+		ApiServer::new(s3_region, S3ApiServer { garage })
+			.run_server(addr, shutdown_signal)
+			.await
 	}
 
 	async fn handle_request_without_bucket(
@@ -167,14 +165,7 @@ impl ApiHandler for S3ApiServer {
 			return Err(Error::forbidden("Operation is not allowed for this key."));
 		}
 
-		// Look up what CORS rule might apply to response.
-		// Requests for methods different than GET, HEAD or POST
-		// are always preflighted, i.e. the browser should make
-		// an OPTIONS call before to check it is allowed
-		let matching_cors_rule = match *req.method() {
-			Method::GET | Method::HEAD | Method::POST => find_matching_cors_rule(&bucket, &req)?,
-			_ => None,
-		};
+		let matching_cors_rule = find_matching_cors_rule(&bucket, &req)?;
 
 		let resp = match endpoint {
 			Endpoint::HeadObject {
@@ -219,7 +210,7 @@ impl ApiHandler for S3ApiServer {
 				.await
 			}
 			Endpoint::PutObject { key } => {
-				handle_put(garage, req, bucket_id, &key, content_sha256).await
+				handle_put(garage, req, &bucket, &key, content_sha256).await
 			}
 			Endpoint::AbortMultipartUpload { key, upload_id } => {
 				handle_abort_multipart_upload(garage, bucket_id, &key, &upload_id).await
@@ -233,7 +224,7 @@ impl ApiHandler for S3ApiServer {
 					garage,
 					req,
 					&bucket_name,
-					bucket_id,
+					&bucket,
 					&key,
 					&upload_id,
 					content_sha256,

src/api/s3/bucket.rs

@@ -295,7 +295,6 @@ fn parse_create_bucket_xml(xml_bytes: &[u8]) -> Option<Option<String>> {
 
 	let mut ret = None;
 	for item in cbc.children() {
-		println!("{:?}", item);
 		if item.has_tag_name("LocationConstraint") {
 			if ret != None {
 				return None;

src/api/s3/copy.rs

@@ -5,9 +5,12 @@ use std::time::{Duration, SystemTime, UNIX_EPOCH};
 use futures::{stream, stream::Stream, StreamExt, TryFutureExt};
 use md5::{Digest as Md5Digest, Md5};
 
+use bytes::Bytes;
 use hyper::{Body, Request, Response};
 use serde::Serialize;
 
+use garage_rpc::netapp::bytes_buf::BytesBuf;
+use garage_rpc::rpc_helper::OrderTag;
 use garage_table::*;
 use garage_util::data::*;
 use garage_util::time::*;
@@ -305,13 +308,18 @@ pub async fn handle_upload_part_copy(
 	// if and only if the block returned is a block that already existed
 	// in the Garage data store (thus we don't need to save it again).
 	let garage2 = garage.clone();
+	let order_stream = OrderTag::stream();
 	let source_blocks = stream::iter(blocks_to_copy)
-		.flat_map(|(block_hash, range_to_copy)| {
+		.enumerate()
+		.flat_map(|(i, (block_hash, range_to_copy))| {
 			let garage3 = garage2.clone();
 			stream::once(async move {
-				let data = garage3.block_manager.rpc_get_block(&block_hash).await?;
+				let data = garage3
+					.block_manager
+					.rpc_get_block(&block_hash, Some(order_stream.order(i as u64)))
+					.await?;
 				match range_to_copy {
-					Some(r) => Ok((data[r].to_vec(), None)),
+					Some(r) => Ok((data.slice(r), None)),
 					None => Ok((data, Some(block_hash))),
 				}
 			})
@@ -553,13 +561,13 @@ impl CopyPreconditionHeaders {
 	}
 }
 
-type BlockStreamItemOk = (Vec<u8>, Option<Hash>);
+type BlockStreamItemOk = (Bytes, Option<Hash>);
 type BlockStreamItem = Result<BlockStreamItemOk, garage_util::error::Error>;
 
 struct Defragmenter<S: Stream<Item = BlockStreamItem>> {
 	block_size: usize,
 	block_stream: Pin<Box<stream::Peekable<S>>>,
-	buffer: Vec<u8>,
+	buffer: BytesBuf,
 	hash: Option<Hash>,
 }
 
@@ -568,7 +576,7 @@ impl<S: Stream<Item = BlockStreamItem>> Defragmenter<S> {
 		Self {
 			block_size,
 			block_stream,
-			buffer: vec![],
+			buffer: BytesBuf::new(),
 			hash: None,
 		}
 	}
@@ -586,7 +594,7 @@ impl<S: Stream<Item = BlockStreamItem>> Defragmenter<S> {
 
 			if self.buffer.is_empty() {
 				let (next_block, next_block_hash) = self.block_stream.next().await.unwrap()?;
-				self.buffer = next_block;
+				self.buffer.extend(next_block);
 				self.hash = next_block_hash;
 			} else if self.buffer.len() + peeked_next_block.len() > self.block_size {
 				break;
@@ -597,11 +605,11 @@ impl<S: Stream<Item = BlockStreamItem>> Defragmenter<S> {
 			}
 		}
 
-		Ok((std::mem::take(&mut self.buffer), self.hash.take()))
+		Ok((self.buffer.take_all(), self.hash.take()))
 	}
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct CopyObjectResult {
 	#[serde(rename = "LastModified")]
 	pub last_modified: s3_xml::Value,
@@ -609,7 +617,7 @@ pub struct CopyObjectResult {
 	pub etag: s3_xml::Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct CopyPartResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -654,7 +662,6 @@ mod tests {
 			last_modified: s3_xml::Value("2011-04-11T20:34:56.000Z".into()),
 			etag: s3_xml::Value("\"9b2cf535f27731c974343645a3985328\"".into()),
 		};
-		println!("{}", to_xml_with_header(&v)?);
 
 		assert_eq!(to_xml_with_header(&v)?, expected_retval);
 

src/api/s3/get.rs

@@ -7,9 +7,9 @@ use http::header::{
 	ACCEPT_RANGES, CONTENT_LENGTH, CONTENT_RANGE, CONTENT_TYPE, ETAG, IF_MODIFIED_SINCE,
 	IF_NONE_MATCH, LAST_MODIFIED, RANGE,
 };
-use hyper::body::Bytes;
 use hyper::{Body, Request, Response, StatusCode};
 
+use garage_rpc::rpc_helper::{netapp::stream::ByteStream, OrderTag};
 use garage_table::EmptyKey;
 use garage_util::data::*;
 
@@ -242,10 +242,15 @@ pub async fn handle_get(
 			Ok(resp_builder.body(body)?)
 		}
 		ObjectVersionData::FirstBlock(_, first_block_hash) => {
-			let read_first_block = garage.block_manager.rpc_get_block(first_block_hash);
+			let order_stream = OrderTag::stream();
+
+			let read_first_block = garage
+				.block_manager
+				.rpc_get_block_streaming(first_block_hash, Some(order_stream.order(0)));
 			let get_next_blocks = garage.version_table.get(&last_v.uuid, &EmptyKey);
 
-			let (first_block, version) = futures::try_join!(read_first_block, get_next_blocks)?;
+			let (first_block_stream, version) =
+				futures::try_join!(read_first_block, get_next_blocks)?;
 			let version = version.ok_or(Error::NoSuchKey)?;
 
 			let mut blocks = version
@@ -254,24 +259,26 @@ pub async fn handle_get(
 				.iter()
 				.map(|(_, vb)| (vb.hash, None))
 				.collect::<Vec<_>>();
-			blocks[0].1 = Some(first_block);
+			blocks[0].1 = Some(first_block_stream);
 
 			let body_stream = futures::stream::iter(blocks)
-				.map(move |(hash, data_opt)| {
+				.enumerate()
+				.map(move |(i, (hash, stream_opt))| {
 					let garage = garage.clone();
 					async move {
-						if let Some(data) = data_opt {
-							Ok(Bytes::from(data))
+						if let Some(stream) = stream_opt {
+							stream
 						} else {
 							garage
 								.block_manager
-								.rpc_get_block(&hash)
+								.rpc_get_block_streaming(&hash, Some(order_stream.order(i as u64)))
 								.await
-								.map(Bytes::from)
+								.unwrap_or_else(|e| error_stream(i, e))
 						}
 					}
 				})
-				.buffered(2);
+				.buffered(2)
+				.flatten();
 
 			let body = hyper::body::Body::wrap_stream(body_stream);
 			Ok(resp_builder.body(body)?)
@@ -422,40 +429,79 @@ fn body_from_blocks_range(
 		all_blocks.len(),
 		4 + ((end - begin) / std::cmp::max(all_blocks[0].1.size as u64, 1024)) as usize,
 	));
-	let mut true_offset = 0;
+	let mut block_offset: u64 = 0;
 	for (_, b) in all_blocks.iter() {
-		if true_offset >= end {
+		if block_offset >= end {
 			break;
 		}
 		// Keep only blocks that have an intersection with the requested range
-		if true_offset < end && true_offset + b.size > begin {
-			blocks.push((*b, true_offset));
+		if block_offset < end && block_offset + b.size > begin {
+			blocks.push((*b, block_offset));
 		}
-		true_offset += b.size;
+		block_offset += b.size as u64;
 	}
 
+	let order_stream = OrderTag::stream();
 	let body_stream = futures::stream::iter(blocks)
-		.map(move |(block, true_offset)| {
+		.enumerate()
+		.map(move |(i, (block, block_offset))| {
 			let garage = garage.clone();
 			async move {
-				let data = garage.block_manager.rpc_get_block(&block.hash).await?;
-				let data = Bytes::from(data);
-				let start_in_block = if true_offset > begin {
-					0
-				} else {
-					begin - true_offset
-				};
-				let end_in_block = if true_offset + block.size < end {
-					block.size
-				} else {
-					end - true_offset
-				};
-				Result::<Bytes, Error>::Ok(
-					data.slice(start_in_block as usize..end_in_block as usize),
-				)
+				garage
+					.block_manager
+					.rpc_get_block_streaming(&block.hash, Some(order_stream.order(i as u64)))
+					.await
+					.unwrap_or_else(|e| error_stream(i, e))
+					.scan(block_offset, move |chunk_offset, chunk| {
+						let r = match chunk {
+							Ok(chunk_bytes) => {
+								let chunk_len = chunk_bytes.len() as u64;
+								let r = if *chunk_offset >= end {
+									// The current chunk is after the part we want to read.
+									// Returning None here will stop the scan, the rest of the
+									// stream will be ignored
+									None
+								} else if *chunk_offset + chunk_len <= begin {
+									// The current chunk is before the part we want to read.
+									// We return a None that will be removed by the filter_map
+									// below.
+									Some(None)
+								} else {
+									// The chunk has an intersection with the requested range
+									let start_in_chunk = if *chunk_offset > begin {
+										0
+									} else {
+										begin - *chunk_offset
+									};
+									let end_in_chunk = if *chunk_offset + chunk_len < end {
+										chunk_len
+									} else {
+										end - *chunk_offset
+									};
+									Some(Some(Ok(chunk_bytes
+										.slice(start_in_chunk as usize..end_in_chunk as usize))))
+								};
+								*chunk_offset += chunk_bytes.len() as u64;
+								r
+							}
+							Err(e) => Some(Some(Err(e))),
+						};
+						futures::future::ready(r)
+					})
+					.filter_map(futures::future::ready)
 			}
 		})
-		.buffered(2);
+		.buffered(2)
+		.flatten();
 
 	hyper::body::Body::wrap_stream(body_stream)
 }
+
+fn error_stream(i: usize, e: garage_util::error::Error) -> ByteStream {
+	Box::pin(futures::stream::once(async move {
+		Err(std::io::Error::new(
+			std::io::ErrorKind::Other,
+			format!("Could not get block {}: {}", i, e),
+		))
+	}))
+}

src/api/s3/post_object.rs

@@ -22,7 +22,7 @@ use crate::signature::payload::{parse_date, verify_v4};
 pub async fn handle_post_object(
 	garage: Arc<Garage>,
 	req: Request<Body>,
-	bucket: String,
+	bucket_name: String,
 ) -> Result<Response<Body>, Error> {
 	let boundary = req
 		.headers()
@@ -126,13 +126,18 @@ pub async fn handle_post_object(
 
 	let bucket_id = garage
 		.bucket_helper()
-		.resolve_bucket(&bucket, &api_key)
+		.resolve_bucket(&bucket_name, &api_key)
 		.await?;
 
 	if !api_key.allow_write(&bucket_id) {
 		return Err(Error::forbidden("Operation is not allowed for this key."));
 	}
 
+	let bucket = garage
+		.bucket_helper()
+		.get_existing_bucket(bucket_id)
+		.await?;
+
 	let decoded_policy = base64::decode(&policy).ok_or_bad_request("Invalid policy")?;
 	let decoded_policy: Policy =
 		serde_json::from_slice(&decoded_policy).ok_or_bad_request("Invalid policy")?;
@@ -227,7 +232,7 @@ pub async fn handle_post_object(
 		garage,
 		headers,
 		StreamLimiter::new(stream, conditions.content_length),
-		bucket_id,
+		&bucket,
 		&key,
 		None,
 		None,
@@ -244,7 +249,7 @@ pub async fn handle_post_object(
 	{
 		target
 			.query_pairs_mut()
-			.append_pair("bucket", &bucket)
+			.append_pair("bucket", &bucket_name)
 			.append_pair("key", &key)
 			.append_pair("etag", &etag);
 		let target = target.to_string();
@@ -289,7 +294,7 @@ pub async fn handle_post_object(
 				let xml = s3_xml::PostObject {
 					xmlns: (),
 					location: s3_xml::Value(location),
-					bucket: s3_xml::Value(bucket),
+					bucket: s3_xml::Value(bucket_name),
 					key: s3_xml::Value(key),
 					etag: s3_xml::Value(etag),
 				};

src/api/s3/put.rs

@@ -1,4 +1,4 @@
-use std::collections::{BTreeMap, BTreeSet, VecDeque};
+use std::collections::{BTreeMap, BTreeSet, HashMap};
 use std::sync::Arc;
 
 use futures::prelude::*;
@@ -8,13 +8,22 @@ use hyper::{Request, Response};
 use md5::{digest::generic_array::*, Digest as Md5Digest, Md5};
 use sha2::Sha256;
 
+use opentelemetry::{
+	trace::{FutureExt as OtelFutureExt, TraceContextExt, Tracer},
+	Context,
+};
+
+use garage_rpc::netapp::bytes_buf::BytesBuf;
 use garage_table::*;
+use garage_util::async_hash::*;
 use garage_util::data::*;
 use garage_util::error::Error as GarageError;
 use garage_util::time::*;
 
 use garage_block::manager::INLINE_THRESHOLD;
+use garage_model::bucket_table::Bucket;
 use garage_model::garage::Garage;
+use garage_model::index_counter::CountedItem;
 use garage_model::s3::block_ref_table::*;
 use garage_model::s3::object_table::*;
 use garage_model::s3::version_table::*;
@@ -26,7 +35,7 @@ use crate::signature::verify_signed_content;
 pub async fn handle_put(
 	garage: Arc<Garage>,
 	req: Request<Body>,
-	bucket_id: Uuid,
+	bucket: &Bucket,
 	key: &str,
 	content_sha256: Option<Hash>,
 ) -> Result<Response<Body>, Error> {
@@ -46,7 +55,7 @@ pub async fn handle_put(
 		garage,
 		headers,
 		body,
-		bucket_id,
+		bucket,
 		key,
 		content_md5,
 		content_sha256,
@@ -59,7 +68,7 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 	garage: Arc<Garage>,
 	headers: ObjectVersionHeaders,
 	body: S,
-	bucket_id: Uuid,
+	bucket: &Bucket,
 	key: &str,
 	content_md5: Option<String>,
 	content_sha256: Option<FixedBytes32>,
@@ -80,6 +89,7 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 		let data_md5sum_hex = hex::encode(data_md5sum);
 
 		let data_sha256sum = sha256sum(&first_block[..]);
+		let size = first_block.len() as u64;
 
 		ensure_checksum_matches(
 			data_md5sum.as_slice(),
@@ -88,20 +98,22 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 			content_sha256,
 		)?;
 
+		check_quotas(&garage, bucket, key, size).await?;
+
 		let object_version = ObjectVersion {
 			uuid: version_uuid,
 			timestamp: version_timestamp,
 			state: ObjectVersionState::Complete(ObjectVersionData::Inline(
 				ObjectVersionMeta {
 					headers,
-					size: first_block.len() as u64,
+					size,
 					etag: data_md5sum_hex.clone(),
 				},
-				first_block,
+				first_block.to_vec(),
 			)),
 		};
 
-		let object = Object::new(bucket_id, key.into(), vec![object_version]);
+		let object = Object::new(bucket.id, key.into(), vec![object_version]);
 		garage.object_table.insert(&object).await?;
 
 		return Ok((version_uuid, data_md5sum_hex));
@@ -114,36 +126,42 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 		timestamp: version_timestamp,
 		state: ObjectVersionState::Uploading(headers.clone()),
 	};
-	let object = Object::new(bucket_id, key.into(), vec![object_version.clone()]);
+	let object = Object::new(bucket.id, key.into(), vec![object_version.clone()]);
 	garage.object_table.insert(&object).await?;
 
 	// Initialize corresponding entry in version table
 	// Write this entry now, even with empty block list,
 	// to prevent block_ref entries from being deleted (they can be deleted
 	// if the reference a version that isn't found in the version table)
-	let version = Version::new(version_uuid, bucket_id, key.into(), false);
+	let version = Version::new(version_uuid, bucket.id, key.into(), false);
 	garage.version_table.insert(&version).await?;
 
 	// Transfer data and verify checksum
-	let first_block_hash = blake2sum(&first_block[..]);
-	let tx_result = read_and_put_blocks(
-		&garage,
-		&version,
-		1,
-		first_block,
-		first_block_hash,
-		&mut chunker,
-	)
-	.await
-	.and_then(|(total_size, data_md5sum, data_sha256sum)| {
+	let first_block_hash = async_blake2sum(first_block.clone()).await;
+
+	let tx_result = (|| async {
+		let (total_size, data_md5sum, data_sha256sum) = read_and_put_blocks(
+			&garage,
+			&version,
+			1,
+			first_block,
+			first_block_hash,
+			&mut chunker,
+		)
+		.await?;
+
 		ensure_checksum_matches(
 			data_md5sum.as_slice(),
 			data_sha256sum,
 			content_md5.as_deref(),
 			content_sha256,
-		)
-		.map(|()| (total_size, data_md5sum))
-	});
+		)?;
+
+		check_quotas(&garage, bucket, key, total_size).await?;
+
+		Ok((total_size, data_md5sum))
+	})()
+	.await;
 
 	// If something went wrong, clean up
 	let (total_size, md5sum_arr) = match tx_result {
@@ -151,7 +169,7 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 		Err(e) => {
 			// Mark object as aborted, this will free the blocks further down
 			object_version.state = ObjectVersionState::Aborted;
-			let object = Object::new(bucket_id, key.into(), vec![object_version.clone()]);
+			let object = Object::new(bucket.id, key.into(), vec![object_version.clone()]);
 			garage.object_table.insert(&object).await?;
 			return Err(e);
 		}
@@ -167,7 +185,7 @@ pub(crate) async fn save_stream<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 		},
 		first_block_hash,
 	));
-	let object = Object::new(bucket_id, key.into(), vec![object_version]);
+	let object = Object::new(bucket.id, key.into(), vec![object_version]);
 	garage.object_table.insert(&object).await?;
 
 	Ok((version_uuid, md5sum_hex))
@@ -200,18 +218,85 @@ fn ensure_checksum_matches(
 	Ok(())
 }
 
+/// Check that inserting this object with this size doesn't exceed bucket quotas
+async fn check_quotas(
+	garage: &Arc<Garage>,
+	bucket: &Bucket,
+	key: &str,
+	size: u64,
+) -> Result<(), Error> {
+	let quotas = bucket.state.as_option().unwrap().quotas.get();
+	if quotas.max_objects.is_none() && quotas.max_size.is_none() {
+		return Ok(());
+	};
+
+	let key = key.to_string();
+	let (prev_object, counters) = futures::try_join!(
+		garage.object_table.get(&bucket.id, &key),
+		garage.object_counter_table.table.get(&bucket.id, &EmptyKey),
+	)?;
+
+	let counters = counters
+		.map(|x| x.filtered_values(&garage.system.ring.borrow()))
+		.unwrap_or_default();
+
+	let (prev_cnt_obj, prev_cnt_size) = match prev_object {
+		Some(o) => {
+			let prev_cnt = o.counts().into_iter().collect::<HashMap<_, _>>();
+			(
+				prev_cnt.get(OBJECTS).cloned().unwrap_or_default(),
+				prev_cnt.get(BYTES).cloned().unwrap_or_default(),
+			)
+		}
+		None => (0, 0),
+	};
+	let cnt_obj_diff = 1 - prev_cnt_obj;
+	let cnt_size_diff = size as i64 - prev_cnt_size;
+
+	if let Some(mo) = quotas.max_objects {
+		let current_objects = counters.get(OBJECTS).cloned().unwrap_or_default();
+		if cnt_obj_diff > 0 && current_objects + cnt_obj_diff > mo as i64 {
+			return Err(Error::forbidden(format!(
+				"Object quota is reached, maximum objects for this bucket: {}",
+				mo
+			)));
+		}
+	}
+
+	if let Some(ms) = quotas.max_size {
+		let current_size = counters.get(BYTES).cloned().unwrap_or_default();
+		if cnt_size_diff > 0 && current_size + cnt_size_diff > ms as i64 {
+			return Err(Error::forbidden(format!(
+				"Bucket size quota is reached, maximum total size of objects for this bucket: {}. The bucket is already {} bytes, and this object would add {} bytes.",
+				ms, current_size, size
+			)));
+		}
+	}
+
+	Ok(())
+}
+
 async fn read_and_put_blocks<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 	garage: &Garage,
 	version: &Version,
 	part_number: u64,
-	first_block: Vec<u8>,
+	first_block: Bytes,
 	first_block_hash: Hash,
 	chunker: &mut StreamChunker<S>,
 ) -> Result<(u64, GenericArray<u8, typenum::U16>, Hash), Error> {
-	let mut md5hasher = Md5::new();
-	let mut sha256hasher = Sha256::new();
-	md5hasher.update(&first_block[..]);
-	sha256hasher.update(&first_block[..]);
+	let tracer = opentelemetry::global::tracer("garage");
+
+	let md5hasher = AsyncHasher::<Md5>::new();
+	let sha256hasher = AsyncHasher::<Sha256>::new();
+
+	futures::future::join(
+		md5hasher.update(first_block.clone()),
+		sha256hasher.update(first_block.clone()),
+	)
+	.with_context(Context::current_with_span(
+		tracer.start("Hash first block (md5, sha256)"),
+	))
+	.await;
 
 	let mut next_offset = first_block.len();
 	let mut put_curr_version_block = put_block_meta(
@@ -233,9 +318,15 @@ async fn read_and_put_blocks<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 			chunker.next(),
 		)?;
 		if let Some(block) = next_block {
-			md5hasher.update(&block[..]);
-			sha256hasher.update(&block[..]);
-			let block_hash = blake2sum(&block[..]);
+			let (_, _, block_hash) = futures::future::join3(
+				md5hasher.update(block.clone()),
+				sha256hasher.update(block.clone()),
+				async_blake2sum(block.clone()),
+			)
+			.with_context(Context::current_with_span(
+				tracer.start("Hash block (md5, sha256, blake2)"),
+			))
+			.await;
 			let block_len = block.len();
 			put_curr_version_block = put_block_meta(
 				garage,
@@ -253,9 +344,9 @@ async fn read_and_put_blocks<S: Stream<Item = Result<Bytes, Error>> + Unpin>(
 	}
 
 	let total_size = next_offset as u64;
-	let data_md5sum = md5hasher.finalize();
+	let data_md5sum = md5hasher.finalize().await;
 
-	let data_sha256sum = sha256hasher.finalize();
+	let data_sha256sum = sha256hasher.finalize().await;
 	let data_sha256sum = Hash::try_from(&data_sha256sum[..]).unwrap();
 
 	Ok((total_size, data_md5sum, data_sha256sum))
@@ -295,7 +386,7 @@ struct StreamChunker<S: Stream<Item = Result<Bytes, Error>>> {
 	stream: S,
 	read_all: bool,
 	block_size: usize,
-	buf: VecDeque<u8>,
+	buf: BytesBuf,
 }
 
 impl<S: Stream<Item = Result<Bytes, Error>> + Unpin> StreamChunker<S> {
@@ -304,11 +395,11 @@ impl<S: Stream<Item = Result<Bytes, Error>> + Unpin> StreamChunker<S> {
 			stream,
 			read_all: false,
 			block_size,
-			buf: VecDeque::with_capacity(2 * block_size),
+			buf: BytesBuf::new(),
 		}
 	}
 
-	async fn next(&mut self) -> Result<Option<Vec<u8>>, Error> {
+	async fn next(&mut self) -> Result<Option<Bytes>, Error> {
 		while !self.read_all && self.buf.len() < self.block_size {
 			if let Some(block) = self.stream.next().await {
 				let bytes = block?;
@@ -321,12 +412,8 @@ impl<S: Stream<Item = Result<Bytes, Error>> + Unpin> StreamChunker<S> {
 
 		if self.buf.is_empty() {
 			Ok(None)
-		} else if self.buf.len() <= self.block_size {
-			let block = self.buf.drain(..).collect::<Vec<u8>>();
-			Ok(Some(block))
 		} else {
-			let block = self.buf.drain(..self.block_size).collect::<Vec<u8>>();
-			Ok(Some(block))
+			Ok(Some(self.buf.take_max(self.block_size)))
 		}
 	}
 }
@@ -435,7 +522,9 @@ pub async fn handle_put_part(
 
 	// Copy block to store
 	let version = Version::new(version_uuid, bucket_id, key, false);
-	let first_block_hash = blake2sum(&first_block[..]);
+
+	let first_block_hash = async_blake2sum(first_block.clone()).await;
+
 	let (_, data_md5sum, data_sha256sum) = read_and_put_blocks(
 		&garage,
 		&version,
@@ -473,7 +562,7 @@ pub async fn handle_complete_multipart_upload(
 	garage: Arc<Garage>,
 	req: Request<Body>,
 	bucket_name: &str,
-	bucket_id: Uuid,
+	bucket: &Bucket,
 	key: &str,
 	upload_id: &str,
 	content_sha256: Option<Hash>,
@@ -497,7 +586,7 @@ pub async fn handle_complete_multipart_upload(
 	// Get object and version
 	let key = key.to_string();
 	let (object, version) = futures::try_join!(
-		garage.object_table.get(&bucket_id, &key),
+		garage.object_table.get(&bucket.id, &key),
 		garage.version_table.get(&version_uuid, &EmptyKey),
 	)?;
 
@@ -590,6 +679,14 @@ pub async fn handle_complete_multipart_upload(
 	// Calculate total size of final object
 	let total_size = version.blocks.items().iter().map(|x| x.1.size).sum();
 
+	if let Err(e) = check_quotas(&garage, bucket, &key, total_size).await {
+		object_version.state = ObjectVersionState::Aborted;
+		let final_object = Object::new(bucket.id, key.clone(), vec![object_version]);
+		garage.object_table.insert(&final_object).await?;
+
+		return Err(e);
+	}
+
 	// Write final object version
 	object_version.state = ObjectVersionState::Complete(ObjectVersionData::FirstBlock(
 		ObjectVersionMeta {
@@ -600,7 +697,7 @@ pub async fn handle_complete_multipart_upload(
 		version.blocks.items()[0].1.hash,
 	));
 
-	let final_object = Object::new(bucket_id, key.clone(), vec![object_version]);
+	let final_object = Object::new(bucket.id, key.clone(), vec![object_version]);
 	garage.object_table.insert(&final_object).await?;
 
 	// Send response saying ok we're done

src/api/s3/xml.rs

@@ -25,7 +25,7 @@ impl From<&str> for Value {
 #[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
 pub struct IntValue(#[serde(rename = "$value")] pub i64);
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct Bucket {
 	#[serde(rename = "CreationDate")]
 	pub creation_date: Value,
@@ -33,7 +33,7 @@ pub struct Bucket {
 	pub name: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct Owner {
 	#[serde(rename = "DisplayName")]
 	pub display_name: Value,
@@ -41,13 +41,13 @@ pub struct Owner {
 	pub id: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct BucketList {
 	#[serde(rename = "Bucket")]
 	pub entries: Vec<Bucket>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListAllMyBucketsResult {
 	#[serde(rename = "Buckets")]
 	pub buckets: BucketList,
@@ -55,7 +55,7 @@ pub struct ListAllMyBucketsResult {
 	pub owner: Owner,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct LocationConstraint {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -63,7 +63,7 @@ pub struct LocationConstraint {
 	pub region: String,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct Deleted {
 	#[serde(rename = "Key")]
 	pub key: Value,
@@ -73,7 +73,7 @@ pub struct Deleted {
 	pub delete_marker_version_id: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct Error {
 	#[serde(rename = "Code")]
 	pub code: Value,
@@ -85,7 +85,7 @@ pub struct Error {
 	pub region: Option<Value>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct DeleteError {
 	#[serde(rename = "Code")]
 	pub code: Value,
@@ -97,7 +97,7 @@ pub struct DeleteError {
 	pub version_id: Option<Value>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct DeleteResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -107,7 +107,7 @@ pub struct DeleteResult {
 	pub errors: Vec<DeleteError>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct InitiateMultipartUploadResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -119,7 +119,7 @@ pub struct InitiateMultipartUploadResult {
 	pub upload_id: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct CompleteMultipartUploadResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -133,7 +133,7 @@ pub struct CompleteMultipartUploadResult {
 	pub etag: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct Initiator {
 	#[serde(rename = "DisplayName")]
 	pub display_name: Value,
@@ -141,7 +141,7 @@ pub struct Initiator {
 	pub id: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListMultipartItem {
 	#[serde(rename = "Initiated")]
 	pub initiated: Value,
@@ -157,7 +157,7 @@ pub struct ListMultipartItem {
 	pub storage_class: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListMultipartUploadsResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -187,7 +187,7 @@ pub struct ListMultipartUploadsResult {
 	pub encoding_type: Option<Value>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct PartItem {
 	#[serde(rename = "ETag")]
 	pub etag: Value,
@@ -199,7 +199,7 @@ pub struct PartItem {
 	pub size: IntValue,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListPartsResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -227,7 +227,7 @@ pub struct ListPartsResult {
 	pub storage_class: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListBucketItem {
 	#[serde(rename = "Key")]
 	pub key: Value,
@@ -241,13 +241,13 @@ pub struct ListBucketItem {
 	pub storage_class: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct CommonPrefix {
 	#[serde(rename = "Prefix")]
 	pub prefix: Value,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct ListBucketResult {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -281,7 +281,7 @@ pub struct ListBucketResult {
 	pub common_prefixes: Vec<CommonPrefix>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct VersioningConfiguration {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),
@@ -289,7 +289,7 @@ pub struct VersioningConfiguration {
 	pub status: Option<Value>,
 }
 
-#[derive(Debug, Serialize, PartialEq)]
+#[derive(Debug, Serialize, PartialEq, Eq)]
 pub struct PostObject {
 	#[serde(serialize_with = "xmlns_tag")]
 	pub xmlns: (),

src/api/signature/mod.rs

@@ -1,5 +1,5 @@
 use chrono::{DateTime, Utc};
-use hmac::{Hmac, Mac, NewMac};
+use hmac::{Hmac, Mac};
 use sha2::Sha256;
 
 use garage_util::data::{sha256sum, Hash};
@@ -29,17 +29,17 @@ pub fn signing_hmac(
 	secret_key: &str,
 	region: &str,
 	service: &str,
-) -> Result<HmacSha256, crypto_mac::InvalidKeyLength> {
+) -> Result<HmacSha256, crypto_common::InvalidLength> {
 	let secret = String::from("AWS4") + secret_key;
-	let mut date_hmac = HmacSha256::new_varkey(secret.as_bytes())?;
+	let mut date_hmac = HmacSha256::new_from_slice(secret.as_bytes())?;
 	date_hmac.update(datetime.format(SHORT_DATE).to_string().as_bytes());
-	let mut region_hmac = HmacSha256::new_varkey(&date_hmac.finalize().into_bytes())?;
+	let mut region_hmac = HmacSha256::new_from_slice(&date_hmac.finalize().into_bytes())?;
 	region_hmac.update(region.as_bytes());
-	let mut service_hmac = HmacSha256::new_varkey(&region_hmac.finalize().into_bytes())?;
+	let mut service_hmac = HmacSha256::new_from_slice(&region_hmac.finalize().into_bytes())?;
 	service_hmac.update(service.as_bytes());
-	let mut signing_hmac = HmacSha256::new_varkey(&service_hmac.finalize().into_bytes())?;
+	let mut signing_hmac = HmacSha256::new_from_slice(&service_hmac.finalize().into_bytes())?;
 	signing_hmac.update(b"aws4_request");
-	let hmac = HmacSha256::new_varkey(&signing_hmac.finalize().into_bytes())?;
+	let hmac = HmacSha256::new_from_slice(&signing_hmac.finalize().into_bytes())?;
 	Ok(hmac)
 }
 

src/block/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage_block"
-version = "0.7.0"
+version = "0.8.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -14,20 +14,22 @@ path = "lib.rs"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-garage_rpc = { version = "0.7.0", path = "../rpc" }
-garage_util = { version = "0.7.0", path = "../util" }
-garage_table = { version = "0.7.0", path = "../table" }
+garage_db = { version = "0.8.0", path = "../db" }
+garage_rpc = { version = "0.8.0", path = "../rpc" }
+garage_util = { version = "0.8.0", path = "../util" }
+garage_table = { version = "0.8.0", path = "../table" }
 
 opentelemetry = "0.17"
 
+arc-swap = "1.5"
 async-trait = "0.1.7"
 bytes = "1.0"
 hex = "0.4"
 tracing = "0.1.30"
 rand = "0.8"
-zstd = { version = "0.9", default-features = false }
 
-sled = "0.34"
+async-compression = { version = "0.3", features = ["tokio", "zstd"] }
+zstd = { version = "0.9", default-features = false }
 
 rmp-serde = "0.15"
 serde = { version = "1.0", default-features = false, features = ["derive", "rc"] }
@@ -36,3 +38,7 @@ serde_bytes = "0.11"
 futures = "0.3"
 futures-util = "0.3"
 tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
+tokio-util = { version = "0.6", features = ["io"] }
+
+[features]
+system-libs = [ "zstd/pkg-config" ]

src/block/block.rs

@@ -1,16 +1,22 @@
+use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 use zstd::stream::{decode_all as zstd_decode, Encoder};
 
 use garage_util::data::*;
 use garage_util::error::*;
 
+#[derive(Debug, Serialize, Deserialize, Copy, Clone)]
+pub enum DataBlockHeader {
+	Plain,
+	Compressed,
+}
+
 /// A possibly compressed block of data
-#[derive(Debug, Serialize, Deserialize)]
 pub enum DataBlock {
 	/// Uncompressed data
-	Plain(#[serde(with = "serde_bytes")] Vec<u8>),
+	Plain(Bytes),
 	/// Data compressed with zstd
-	Compressed(#[serde(with = "serde_bytes")] Vec<u8>),
+	Compressed(Bytes),
 }
 
 impl DataBlock {
@@ -30,7 +36,7 @@ impl DataBlock {
 	/// Get the buffer, possibly decompressing it, and verify it's integrity.
 	/// For Plain block, data is compared to hash, for Compressed block, zstd checksumming system
 	/// is used instead.
-	pub fn verify_get(self, hash: Hash) -> Result<Vec<u8>, Error> {
+	pub fn verify_get(self, hash: Hash) -> Result<Bytes, Error> {
 		match self {
 			DataBlock::Plain(data) => {
 				if blake2sum(&data) == hash {
@@ -39,9 +45,9 @@ impl DataBlock {
 					Err(Error::CorruptData(hash))
 				}
 			}
-			DataBlock::Compressed(data) => {
-				zstd_decode(&data[..]).map_err(|_| Error::CorruptData(hash))
-			}
+			DataBlock::Compressed(data) => zstd_decode(&data[..])
+				.map_err(|_| Error::CorruptData(hash))
+				.map(Bytes::from),
 		}
 	}
 
@@ -61,13 +67,31 @@ impl DataBlock {
 		}
 	}
 
-	pub fn from_buffer(data: Vec<u8>, level: Option<i32>) -> DataBlock {
-		if let Some(level) = level {
-			if let Ok(data) = zstd_encode(&data[..], level) {
-				return DataBlock::Compressed(data);
+	pub async fn from_buffer(data: Bytes, level: Option<i32>) -> DataBlock {
+		tokio::task::spawn_blocking(move || {
+			if let Some(level) = level {
+				if let Ok(data) = zstd_encode(&data[..], level) {
+					return DataBlock::Compressed(data.into());
+				}
 			}
+			DataBlock::Plain(data)
+		})
+		.await
+		.unwrap()
+	}
+
+	pub fn into_parts(self) -> (DataBlockHeader, Bytes) {
+		match self {
+			DataBlock::Plain(data) => (DataBlockHeader::Plain, data),
+			DataBlock::Compressed(data) => (DataBlockHeader::Compressed, data),
+		}
+	}
+
+	pub fn from_parts(h: DataBlockHeader, bytes: Bytes) -> Self {
+		match h {
+			DataBlockHeader::Plain => DataBlock::Plain(bytes),
+			DataBlockHeader::Compressed => DataBlock::Compressed(bytes),
 		}
-		DataBlock::Plain(data)
 	}
 }
 

src/block/lib.rs

@@ -2,6 +2,8 @@
 extern crate tracing;
 
 pub mod manager;
+pub mod repair;
+pub mod resync;
 
 mod block;
 mod metrics;

src/block/metrics.rs

@@ -1,6 +1,6 @@
 use opentelemetry::{global, metrics::*};
 
-use garage_util::sled_counter::SledCountedTree;
+use garage_db::counted_tree_hack::CountedTree;
 
 /// TableMetrics reference all counter used for metrics
 pub struct BlockManagerMetrics {
@@ -23,7 +23,7 @@ pub struct BlockManagerMetrics {
 }
 
 impl BlockManagerMetrics {
-	pub fn new(resync_queue: SledCountedTree, resync_errors: SledCountedTree) -> Self {
+	pub fn new(resync_queue: CountedTree, resync_errors: CountedTree) -> Self {
 		let meter = global::meter("garage_model/block");
 		Self {
 			_resync_queue_len: meter

src/block/rc.rs

@@ -1,5 +1,7 @@
 use std::convert::TryInto;
 
+use garage_db as db;
+
 use garage_util::data::*;
 use garage_util::error::*;
 use garage_util::time::*;
@@ -7,31 +9,41 @@ use garage_util::time::*;
 use crate::manager::BLOCK_GC_DELAY;
 
 pub struct BlockRc {
-	pub(crate) rc: sled::Tree,
+	pub(crate) rc: db::Tree,
 }
 
 impl BlockRc {
-	pub(crate) fn new(rc: sled::Tree) -> Self {
+	pub(crate) fn new(rc: db::Tree) -> Self {
 		Self { rc }
 	}
 
 	/// Increment the reference counter associated to a hash.
 	/// Returns true if the RC goes from zero to nonzero.
-	pub(crate) fn block_incref(&self, hash: &Hash) -> Result<bool, Error> {
-		let old_rc = self
-			.rc
-			.fetch_and_update(&hash, |old| RcEntry::parse_opt(old).increment().serialize())?;
-		let old_rc = RcEntry::parse_opt(old_rc);
+	pub(crate) fn block_incref(
+		&self,
+		tx: &mut db::Transaction,
+		hash: &Hash,
+	) -> db::TxOpResult<bool> {
+		let old_rc = RcEntry::parse_opt(tx.get(&self.rc, &hash)?);
+		match old_rc.increment().serialize() {
+			Some(x) => tx.insert(&self.rc, &hash, x)?,
+			None => unreachable!(),
+		};
 		Ok(old_rc.is_zero())
 	}
 
 	/// Decrement the reference counter associated to a hash.
 	/// Returns true if the RC is now zero.
-	pub(crate) fn block_decref(&self, hash: &Hash) -> Result<bool, Error> {
-		let new_rc = self
-			.rc
-			.update_and_fetch(&hash, |old| RcEntry::parse_opt(old).decrement().serialize())?;
-		let new_rc = RcEntry::parse_opt(new_rc);
+	pub(crate) fn block_decref(
+		&self,
+		tx: &mut db::Transaction,
+		hash: &Hash,
+	) -> db::TxOpResult<bool> {
+		let new_rc = RcEntry::parse_opt(tx.get(&self.rc, &hash)?).decrement();
+		match new_rc.serialize() {
+			Some(x) => tx.insert(&self.rc, &hash, x)?,
+			None => tx.remove(&self.rc, &hash)?,
+		};
 		Ok(matches!(new_rc, RcEntry::Deletable { .. }))
 	}
 
@@ -44,12 +56,15 @@ impl BlockRc {
 	/// deletion time has passed
 	pub(crate) fn clear_deleted_block_rc(&self, hash: &Hash) -> Result<(), Error> {
 		let now = now_msec();
-		self.rc.update_and_fetch(&hash, |rcval| {
-			let updated = match RcEntry::parse_opt(rcval) {
-				RcEntry::Deletable { at_time } if now > at_time => RcEntry::Absent,
-				v => v,
+		self.rc.db().transaction(|mut tx| {
+			let rcval = RcEntry::parse_opt(tx.get(&self.rc, &hash)?);
+			match rcval {
+				RcEntry::Deletable { at_time } if now > at_time => {
+					tx.remove(&self.rc, &hash)?;
+				}
+				_ => (),
 			};
-			updated.serialize()
+			tx.commit(())
 		})?;
 		Ok(())
 	}

src/block/repair.rs

@@ -0,0 +1,466 @@
+use core::ops::Bound;
+use std::path::PathBuf;
+use std::sync::Arc;
+use std::time::Duration;
+
+use async_trait::async_trait;
+use serde::{Deserialize, Serialize};
+use tokio::fs;
+use tokio::select;
+use tokio::sync::mpsc;
+use tokio::sync::watch;
+
+use garage_util::background::*;
+use garage_util::data::*;
+use garage_util::error::*;
+use garage_util::persister::Persister;
+use garage_util::time::*;
+use garage_util::tranquilizer::Tranquilizer;
+
+use crate::manager::*;
+
+// Full scrub every 30 days
+const SCRUB_INTERVAL: Duration = Duration::from_secs(3600 * 24 * 30);
+// Scrub tranquility is initially set to 4, but can be changed in the CLI
+// and the updated version is persisted over Garage restarts
+const INITIAL_SCRUB_TRANQUILITY: u32 = 4;
+
+// ---- ---- ----
+// FIRST KIND OF REPAIR: FINDING MISSING BLOCKS/USELESS BLOCKS
+// This is a one-shot repair operation that can be launched,
+// checks everything, and then exits.
+// ---- ---- ----
+
+pub struct RepairWorker {
+	manager: Arc<BlockManager>,
+	next_start: Option<Hash>,
+	block_iter: Option<BlockStoreIterator>,
+}
+
+impl RepairWorker {
+	pub fn new(manager: Arc<BlockManager>) -> Self {
+		Self {
+			manager,
+			next_start: None,
+			block_iter: None,
+		}
+	}
+}
+
+#[async_trait]
+impl Worker for RepairWorker {
+	fn name(&self) -> String {
+		"Block repair worker".into()
+	}
+
+	fn info(&self) -> Option<String> {
+		match self.block_iter.as_ref() {
+			None => {
+				let idx_bytes = self
+					.next_start
+					.as_ref()
+					.map(|x| x.as_slice())
+					.unwrap_or(&[]);
+				let idx_bytes = if idx_bytes.len() > 4 {
+					&idx_bytes[..4]
+				} else {
+					idx_bytes
+				};
+				Some(format!("Phase 1: {}", hex::encode(idx_bytes)))
+			}
+			Some(bi) => Some(format!("Phase 2: {:.2}% done", bi.progress() * 100.)),
+		}
+	}
+
+	async fn work(&mut self, _must_exit: &mut watch::Receiver<bool>) -> Result<WorkerState, Error> {
+		match self.block_iter.as_mut() {
+			None => {
+				// Phase 1: Repair blocks from RC table.
+
+				// We have to do this complicated two-step process where we first read a bunch
+				// of hashes from the RC table, and then insert them in the to-resync queue,
+				// because of SQLite. Basically, as long as we have an iterator on a DB table,
+				// we can't do anything else on the DB. The naive approach (which we had previously)
+				// of just iterating on the RC table and inserting items one to one in the resync
+				// queue can't work here, it would just provoke a deadlock in the SQLite adapter code.
+				// This is mostly because the Rust bindings for SQLite assume a worst-case scenario
+				// where SQLite is not compiled in thread-safe mode, so we have to wrap everything
+				// in a mutex (see db/sqlite_adapter.rs and discussion in PR #322).
+				// TODO: maybe do this with tokio::task::spawn_blocking ?
+				let mut batch_of_hashes = vec![];
+				let start_bound = match self.next_start.as_ref() {
+					None => Bound::Unbounded,
+					Some(x) => Bound::Excluded(x.as_slice()),
+				};
+				for entry in self
+					.manager
+					.rc
+					.rc
+					.range::<&[u8], _>((start_bound, Bound::Unbounded))?
+				{
+					let (hash, _) = entry?;
+					let hash = Hash::try_from(&hash[..]).unwrap();
+					batch_of_hashes.push(hash);
+					if batch_of_hashes.len() >= 1000 {
+						break;
+					}
+				}
+				if batch_of_hashes.is_empty() {
+					// move on to phase 2
+					self.block_iter = Some(BlockStoreIterator::new(&self.manager));
+					return Ok(WorkerState::Busy);
+				}
+
+				for hash in batch_of_hashes.into_iter() {
+					self.manager
+						.resync
+						.put_to_resync(&hash, Duration::from_secs(0))?;
+					self.next_start = Some(hash)
+				}
+
+				Ok(WorkerState::Busy)
+			}
+			Some(bi) => {
+				// Phase 2: Repair blocks actually on disk
+				// Lists all blocks on disk and adds them to the resync queue.
+				// This allows us to find blocks we are storing but don't actually need,
+				// so that we can offload them if necessary and then delete them locally.
+				if let Some(hash) = bi.next().await? {
+					self.manager
+						.resync
+						.put_to_resync(&hash, Duration::from_secs(0))?;
+					Ok(WorkerState::Busy)
+				} else {
+					Ok(WorkerState::Done)
+				}
+			}
+		}
+	}
+
+	async fn wait_for_work(&mut self, _must_exit: &watch::Receiver<bool>) -> WorkerState {
+		unreachable!()
+	}
+}
+
+// ---- ---- ----
+// SECOND KIND OF REPAIR: SCRUBBING THE DATASTORE
+// This is significantly more complex than the process above,
+// as it is a continuously-running task that triggers automatically
+// every SCRUB_INTERVAL, but can also be triggered manually
+// and whose parameter (esp. speed) can be controlled at runtime.
+// ---- ---- ----
+
+pub struct ScrubWorker {
+	manager: Arc<BlockManager>,
+	rx_cmd: mpsc::Receiver<ScrubWorkerCommand>,
+
+	work: ScrubWorkerState,
+	tranquilizer: Tranquilizer,
+
+	persister: Persister<ScrubWorkerPersisted>,
+	persisted: ScrubWorkerPersisted,
+}
+
+#[derive(Serialize, Deserialize)]
+struct ScrubWorkerPersisted {
+	tranquility: u32,
+	time_last_complete_scrub: u64,
+	corruptions_detected: u64,
+}
+
+enum ScrubWorkerState {
+	Running(BlockStoreIterator),
+	Paused(BlockStoreIterator, u64), // u64 = time when to resume scrub
+	Finished,
+}
+
+impl Default for ScrubWorkerState {
+	fn default() -> Self {
+		ScrubWorkerState::Finished
+	}
+}
+
+#[derive(Debug)]
+pub enum ScrubWorkerCommand {
+	Start,
+	Pause(Duration),
+	Resume,
+	Cancel,
+	SetTranquility(u32),
+}
+
+impl ScrubWorker {
+	pub fn new(manager: Arc<BlockManager>, rx_cmd: mpsc::Receiver<ScrubWorkerCommand>) -> Self {
+		let persister = Persister::new(&manager.system.metadata_dir, "scrub_info");
+		let persisted = match persister.load() {
+			Ok(v) => v,
+			Err(_) => ScrubWorkerPersisted {
+				time_last_complete_scrub: 0,
+				tranquility: INITIAL_SCRUB_TRANQUILITY,
+				corruptions_detected: 0,
+			},
+		};
+		Self {
+			manager,
+			rx_cmd,
+			work: ScrubWorkerState::Finished,
+			tranquilizer: Tranquilizer::new(30),
+			persister,
+			persisted,
+		}
+	}
+
+	async fn handle_cmd(&mut self, cmd: ScrubWorkerCommand) {
+		match cmd {
+			ScrubWorkerCommand::Start => {
+				self.work = match std::mem::take(&mut self.work) {
+					ScrubWorkerState::Finished => {
+						let iterator = BlockStoreIterator::new(&self.manager);
+						ScrubWorkerState::Running(iterator)
+					}
+					work => {
+						error!("Cannot start scrub worker: already running!");
+						work
+					}
+				};
+			}
+			ScrubWorkerCommand::Pause(dur) => {
+				self.work = match std::mem::take(&mut self.work) {
+					ScrubWorkerState::Running(it) | ScrubWorkerState::Paused(it, _) => {
+						ScrubWorkerState::Paused(it, now_msec() + dur.as_millis() as u64)
+					}
+					work => {
+						error!("Cannot pause scrub worker: not running!");
+						work
+					}
+				};
+			}
+			ScrubWorkerCommand::Resume => {
+				self.work = match std::mem::take(&mut self.work) {
+					ScrubWorkerState::Paused(it, _) => ScrubWorkerState::Running(it),
+					work => {
+						error!("Cannot resume scrub worker: not paused!");
+						work
+					}
+				};
+			}
+			ScrubWorkerCommand::Cancel => {
+				self.work = match std::mem::take(&mut self.work) {
+					ScrubWorkerState::Running(_) | ScrubWorkerState::Paused(_, _) => {
+						ScrubWorkerState::Finished
+					}
+					work => {
+						error!("Cannot cancel scrub worker: not running!");
+						work
+					}
+				}
+			}
+			ScrubWorkerCommand::SetTranquility(t) => {
+				self.persisted.tranquility = t;
+				if let Err(e) = self.persister.save_async(&self.persisted).await {
+					error!("Could not save new tranquilitiy value: {}", e);
+				}
+			}
+		}
+	}
+}
+
+#[async_trait]
+impl Worker for ScrubWorker {
+	fn name(&self) -> String {
+		"Block scrub worker".into()
+	}
+
+	fn info(&self) -> Option<String> {
+		let s = match &self.work {
+			ScrubWorkerState::Running(bsi) => format!(
+				"{:.2}% done (tranquility = {})",
+				bsi.progress() * 100.,
+				self.persisted.tranquility
+			),
+			ScrubWorkerState::Paused(bsi, rt) => {
+				format!(
+					"Paused, {:.2}% done, resumes at {}",
+					bsi.progress() * 100.,
+					msec_to_rfc3339(*rt)
+				)
+			}
+			ScrubWorkerState::Finished => format!(
+				"Last completed scrub: {}",
+				msec_to_rfc3339(self.persisted.time_last_complete_scrub)
+			),
+		};
+		Some(format!(
+			"{} ; corruptions detected: {}",
+			s, self.persisted.corruptions_detected
+		))
+	}
+
+	async fn work(&mut self, _must_exit: &mut watch::Receiver<bool>) -> Result<WorkerState, Error> {
+		match self.rx_cmd.try_recv() {
+			Ok(cmd) => self.handle_cmd(cmd).await,
+			Err(mpsc::error::TryRecvError::Disconnected) => return Ok(WorkerState::Done),
+			Err(mpsc::error::TryRecvError::Empty) => (),
+		};
+
+		match &mut self.work {
+			ScrubWorkerState::Running(bsi) => {
+				self.tranquilizer.reset();
+				if let Some(hash) = bsi.next().await? {
+					match self.manager.read_block(&hash).await {
+						Err(Error::CorruptData(_)) => {
+							error!("Found corrupt data block during scrub: {:?}", hash);
+							self.persisted.corruptions_detected += 1;
+							self.persister.save_async(&self.persisted).await?;
+						}
+						Err(e) => return Err(e),
+						_ => (),
+					};
+					Ok(self
+						.tranquilizer
+						.tranquilize_worker(self.persisted.tranquility))
+				} else {
+					self.persisted.time_last_complete_scrub = now_msec();
+					self.persister.save_async(&self.persisted).await?;
+					self.work = ScrubWorkerState::Finished;
+					self.tranquilizer.clear();
+					Ok(WorkerState::Idle)
+				}
+			}
+			_ => Ok(WorkerState::Idle),
+		}
+	}
+
+	async fn wait_for_work(&mut self, _must_exit: &watch::Receiver<bool>) -> WorkerState {
+		let (wait_until, command) = match &self.work {
+			ScrubWorkerState::Running(_) => return WorkerState::Busy,
+			ScrubWorkerState::Paused(_, resume_time) => (*resume_time, ScrubWorkerCommand::Resume),
+			ScrubWorkerState::Finished => (
+				self.persisted.time_last_complete_scrub + SCRUB_INTERVAL.as_millis() as u64,
+				ScrubWorkerCommand::Start,
+			),
+		};
+
+		let now = now_msec();
+		if now >= wait_until {
+			self.handle_cmd(command).await;
+			return WorkerState::Busy;
+		}
+		let delay = Duration::from_millis(wait_until - now);
+		select! {
+			_ = tokio::time::sleep(delay) => self.handle_cmd(command).await,
+			cmd = self.rx_cmd.recv() => if let Some(cmd) = cmd {
+				self.handle_cmd(cmd).await;
+			} else {
+				return WorkerState::Done;
+			}
+		}
+
+		match &self.work {
+			ScrubWorkerState::Running(_) => WorkerState::Busy,
+			_ => WorkerState::Idle,
+		}
+	}
+}
+
+// ---- ---- ----
+// UTILITY FOR ENUMERATING THE BLOCK STORE
+// ---- ---- ----
+
+struct BlockStoreIterator {
+	path: Vec<ReadingDir>,
+}
+
+enum ReadingDir {
+	Pending(PathBuf),
+	Read {
+		subpaths: Vec<fs::DirEntry>,
+		pos: usize,
+	},
+}
+
+impl BlockStoreIterator {
+	fn new(manager: &BlockManager) -> Self {
+		let root_dir = manager.data_dir.clone();
+		Self {
+			path: vec![ReadingDir::Pending(root_dir)],
+		}
+	}
+
+	/// Returns progress done, between 0 and 1
+	fn progress(&self) -> f32 {
+		if self.path.is_empty() {
+			1.0
+		} else {
+			let mut ret = 0.0;
+			let mut next_div = 1;
+			for p in self.path.iter() {
+				match p {
+					ReadingDir::Pending(_) => break,
+					ReadingDir::Read { subpaths, pos } => {
+						next_div *= subpaths.len();
+						ret += ((*pos - 1) as f32) / (next_div as f32);
+					}
+				}
+			}
+			ret
+		}
+	}
+
+	async fn next(&mut self) -> Result<Option<Hash>, Error> {
+		loop {
+			let last_path = match self.path.last_mut() {
+				None => return Ok(None),
+				Some(lp) => lp,
+			};
+
+			if let ReadingDir::Pending(path) = last_path {
+				let mut reader = fs::read_dir(&path).await?;
+				let mut subpaths = vec![];
+				while let Some(ent) = reader.next_entry().await? {
+					subpaths.push(ent);
+				}
+				*last_path = ReadingDir::Read { subpaths, pos: 0 };
+			}
+
+			let (subpaths, pos) = match *last_path {
+				ReadingDir::Read {
+					ref subpaths,
+					ref mut pos,
+				} => (subpaths, pos),
+				ReadingDir::Pending(_) => unreachable!(),
+			};
+
+			let data_dir_ent = match subpaths.get(*pos) {
+				None => {
+					self.path.pop();
+					continue;
+				}
+				Some(ent) => {
+					*pos += 1;
+					ent
+				}
+			};
+
+			let name = data_dir_ent.file_name();
+			let name = if let Ok(n) = name.into_string() {
+				n
+			} else {
+				continue;
+			};
+			let ent_type = data_dir_ent.file_type().await?;
+
+			let name = name.strip_suffix(".zst").unwrap_or(&name);
+			if name.len() == 2 && hex::decode(&name).is_ok() && ent_type.is_dir() {
+				let path = data_dir_ent.path();
+				self.path.push(ReadingDir::Pending(path));
+			} else if name.len() == 64 {
+				if let Ok(h) = hex::decode(&name) {
+					let mut hash = [0u8; 32];
+					hash.copy_from_slice(&h);
+					return Ok(Some(hash.into()));
+				}
+			}
+		}
+	}
+}

src/block/resync.rs

@@ -0,0 +1,599 @@
+use std::collections::HashSet;
+use std::convert::TryInto;
+use std::sync::{Arc, Mutex};
+use std::time::Duration;
+
+use arc_swap::ArcSwap;
+use async_trait::async_trait;
+use serde::{Deserialize, Serialize};
+
+use tokio::select;
+use tokio::sync::{watch, Notify};
+
+use opentelemetry::{
+	trace::{FutureExt as OtelFutureExt, TraceContextExt, Tracer},
+	Context, KeyValue,
+};
+
+use garage_db as db;
+use garage_db::counted_tree_hack::CountedTree;
+
+use garage_util::background::*;
+use garage_util::data::*;
+use garage_util::error::*;
+use garage_util::metrics::RecordDuration;
+use garage_util::persister::Persister;
+use garage_util::time::*;
+use garage_util::tranquilizer::Tranquilizer;
+
+use garage_rpc::system::System;
+use garage_rpc::*;
+
+use garage_table::replication::TableReplication;
+
+use crate::manager::*;
+
+// Timeout for RPCs that ask other nodes whether they need a copy
+// of a given block before we delete it locally
+// The timeout here is relatively low because we don't want to block
+// the entire resync loop when some nodes are not responding.
+// Nothing will be deleted if the nodes don't answer the queries,
+// we will just retry later.
+const NEED_BLOCK_QUERY_TIMEOUT: Duration = Duration::from_secs(15);
+
+// The delay between the time where a resync operation fails
+// and the time when it is retried, with exponential backoff
+// (multiplied by 2, 4, 8, 16, etc. for every consecutive failure).
+pub(crate) const RESYNC_RETRY_DELAY: Duration = Duration::from_secs(60);
+// The minimum retry delay is 60 seconds = 1 minute
+// The maximum retry delay is 60 seconds * 2^6 = 60 seconds << 6 = 64 minutes (~1 hour)
+pub(crate) const RESYNC_RETRY_DELAY_MAX_BACKOFF_POWER: u64 = 6;
+
+// No more than 4 resync workers can be running in the system
+pub(crate) const MAX_RESYNC_WORKERS: usize = 4;
+// Resync tranquility is initially set to 2, but can be changed in the CLI
+// and the updated version is persisted over Garage restarts
+const INITIAL_RESYNC_TRANQUILITY: u32 = 2;
+
+pub struct BlockResyncManager {
+	pub(crate) queue: CountedTree,
+	pub(crate) notify: Notify,
+	pub(crate) errors: CountedTree,
+
+	busy_set: BusySet,
+
+	persister: Persister<ResyncPersistedConfig>,
+	persisted: ArcSwap<ResyncPersistedConfig>,
+}
+
+#[derive(Serialize, Deserialize, Clone, Copy)]
+struct ResyncPersistedConfig {
+	n_workers: usize,
+	tranquility: u32,
+}
+
+enum ResyncIterResult {
+	BusyDidSomething,
+	BusyDidNothing,
+	IdleFor(Duration),
+}
+
+type BusySet = Arc<Mutex<HashSet<Vec<u8>>>>;
+
+struct BusyBlock {
+	time_bytes: Vec<u8>,
+	hash_bytes: Vec<u8>,
+	busy_set: BusySet,
+}
+
+impl BlockResyncManager {
+	pub(crate) fn new(db: &db::Db, system: &System) -> Self {
+		let queue = db
+			.open_tree("block_local_resync_queue")
+			.expect("Unable to open block_local_resync_queue tree");
+		let queue = CountedTree::new(queue).expect("Could not count block_local_resync_queue");
+
+		let errors = db
+			.open_tree("block_local_resync_errors")
+			.expect("Unable to open block_local_resync_errors tree");
+		let errors = CountedTree::new(errors).expect("Could not count block_local_resync_errors");
+
+		let persister = Persister::new(&system.metadata_dir, "resync_cfg");
+		let persisted = match persister.load() {
+			Ok(v) => v,
+			Err(_) => ResyncPersistedConfig {
+				n_workers: 1,
+				tranquility: INITIAL_RESYNC_TRANQUILITY,
+			},
+		};
+
+		Self {
+			queue,
+			notify: Notify::new(),
+			errors,
+			busy_set: Arc::new(Mutex::new(HashSet::new())),
+			persister,
+			persisted: ArcSwap::new(Arc::new(persisted)),
+		}
+	}
+
+	/// Get lenght of resync queue
+	pub fn queue_len(&self) -> Result<usize, Error> {
+		// This currently can't return an error because the CountedTree hack
+		// doesn't error on .len(), but this will change when we remove the hack
+		// (hopefully someday!)
+		Ok(self.queue.len())
+	}
+
+	/// Get number of blocks that have an error
+	pub fn errors_len(&self) -> Result<usize, Error> {
+		// (see queue_len comment)
+		Ok(self.errors.len())
+	}
+
+	// ---- Resync loop ----
+
+	// This part manages a queue of blocks that need to be
+	// "resynchronized", i.e. that need to have a check that
+	// they are at present if we need them, or that they are
+	// deleted once the garbage collection delay has passed.
+	//
+	// Here are some explanations on how the resync queue works.
+	// There are two Sled trees that are used to have information
+	// about the status of blocks that need to be resynchronized:
+	//
+	// - resync.queue: a tree that is ordered first by a timestamp
+	//   (in milliseconds since Unix epoch) that is the time at which
+	//   the resync must be done, and second by block hash.
+	//   The key in this tree is just:
+	//       concat(timestamp (8 bytes), hash (32 bytes))
+	//   The value is the same 32-byte hash.
+	//
+	// - resync.errors: a tree that indicates for each block
+	//   if the last resync resulted in an error, and if so,
+	//   the following two informations (see the ErrorCounter struct):
+	//   - how many consecutive resync errors for this block?
+	//   - when was the last try?
+	//   These two informations are used to implement an
+	//   exponential backoff retry strategy.
+	//   The key in this tree is the 32-byte hash of the block,
+	//   and the value is the encoded ErrorCounter value.
+	//
+	// We need to have these two trees, because the resync queue
+	// is not just a queue of items to process, but a set of items
+	// that are waiting a specific delay until we can process them
+	// (the delay being necessary both internally for the exponential
+	// backoff strategy, and exposed as a parameter when adding items
+	// to the queue, e.g. to wait until the GC delay has passed).
+	// This is why we need one tree ordered by time, and one
+	// ordered by identifier of item to be processed (block hash).
+	//
+	// When the worker wants to process an item it takes from
+	// resync.queue, it checks in resync.errors that if there is an
+	// exponential back-off delay to await, it has passed before we
+	// process the item. If not, the item in the queue is skipped
+	// (but added back for later processing after the time of the
+	// delay).
+	//
+	// An alternative that would have seemed natural is to
+	// only add items to resync.queue with a processing time that is
+	// after the delay, but there are several issues with this:
+	// - This requires to synchronize updates to resync.queue and
+	//   resync.errors (with the current model, there is only one thread,
+	//   the worker thread, that accesses resync.errors,
+	//   so no need to synchronize) by putting them both in a lock.
+	//   This would mean that block_incref might need to take a lock
+	//   before doing its thing, meaning it has much more chances of
+	//   not completing successfully if something bad happens to Garage.
+	//   Currently Garage is not able to recover from block_incref that
+	//   doesn't complete successfully, because it is necessary to ensure
+	//   the consistency between the state of the block manager and
+	//   information in the BlockRef table.
+	// - If a resync fails, we put that block in the resync.errors table,
+	//   and also add it back to resync.queue to be processed after
+	//   the exponential back-off delay,
+	//   but maybe the block is already scheduled to be resynced again
+	//   at another time that is before the exponential back-off delay,
+	//   and we have no way to check that easily. This means that
+	//   in all cases, we need to check the resync.errors table
+	//   in the resync loop at the time when a block is popped from
+	//   the resync.queue.
+	// Overall, the current design is therefore simpler and more robust
+	// because it tolerates inconsistencies between the resync.queue
+	// and resync.errors table (items being scheduled in resync.queue
+	// for times that are earlier than the exponential back-off delay
+	// is a natural condition that is handled properly).
+
+	pub(crate) fn put_to_resync(&self, hash: &Hash, delay: Duration) -> db::Result<()> {
+		let when = now_msec() + delay.as_millis() as u64;
+		self.put_to_resync_at(hash, when)
+	}
+
+	pub(crate) fn put_to_resync_at(&self, hash: &Hash, when: u64) -> db::Result<()> {
+		trace!("Put resync_queue: {} {:?}", when, hash);
+		let mut key = u64::to_be_bytes(when).to_vec();
+		key.extend(hash.as_ref());
+		self.queue.insert(key, hash.as_ref())?;
+		self.notify.notify_waiters();
+		Ok(())
+	}
+
+	async fn resync_iter(&self, manager: &BlockManager) -> Result<ResyncIterResult, db::Error> {
+		if let Some(block) = self.get_block_to_resync()? {
+			let time_msec = u64::from_be_bytes(block.time_bytes[0..8].try_into().unwrap());
+			let now = now_msec();
+
+			if now >= time_msec {
+				let hash = Hash::try_from(&block.hash_bytes[..]).unwrap();
+
+				if let Some(ec) = self.errors.get(hash.as_slice())? {
+					let ec = ErrorCounter::decode(&ec);
+					if now < ec.next_try() {
+						// if next retry after an error is not yet,
+						// don't do resync and return early, but still
+						// make sure the item is still in queue at expected time
+						self.put_to_resync_at(&hash, ec.next_try())?;
+						// ec.next_try() > now >= time_msec, so this remove
+						// is not removing the one we added just above
+						// (we want to do the remove after the insert to ensure
+						// that the item is not lost if we crash in-between)
+						self.queue.remove(&block.time_bytes)?;
+						return Ok(ResyncIterResult::BusyDidNothing);
+					}
+				}
+
+				let tracer = opentelemetry::global::tracer("garage");
+				let trace_id = gen_uuid();
+				let span = tracer
+					.span_builder("Resync block")
+					.with_trace_id(
+						opentelemetry::trace::TraceId::from_hex(&hex::encode(
+							&trace_id.as_slice()[..16],
+						))
+						.unwrap(),
+					)
+					.with_attributes(vec![KeyValue::new("block", format!("{:?}", hash))])
+					.start(&tracer);
+
+				let res = self
+					.resync_block(manager, &hash)
+					.with_context(Context::current_with_span(span))
+					.bound_record_duration(&manager.metrics.resync_duration)
+					.await;
+
+				manager.metrics.resync_counter.add(1);
+
+				if let Err(e) = &res {
+					manager.metrics.resync_error_counter.add(1);
+					warn!("Error when resyncing {:?}: {}", hash, e);
+
+					let err_counter = match self.errors.get(hash.as_slice())? {
+						Some(ec) => ErrorCounter::decode(&ec).add1(now + 1),
+						None => ErrorCounter::new(now + 1),
+					};
+
+					self.errors.insert(hash.as_slice(), err_counter.encode())?;
+
+					self.put_to_resync_at(&hash, err_counter.next_try())?;
+					// err_counter.next_try() >= now + 1 > now,
+					// the entry we remove from the queue is not
+					// the entry we inserted with put_to_resync_at
+					self.queue.remove(&block.time_bytes)?;
+				} else {
+					self.errors.remove(hash.as_slice())?;
+					self.queue.remove(&block.time_bytes)?;
+				}
+
+				Ok(ResyncIterResult::BusyDidSomething)
+			} else {
+				Ok(ResyncIterResult::IdleFor(Duration::from_millis(
+					time_msec - now,
+				)))
+			}
+		} else {
+			// Here we wait either for a notification that an item has been
+			// added to the queue, or for a constant delay of 10 secs to expire.
+			// The delay avoids a race condition where the notification happens
+			// between the time we checked the queue and the first poll
+			// to resync_notify.notified(): if that happens, we'll just loop
+			// back 10 seconds later, which is fine.
+			Ok(ResyncIterResult::IdleFor(Duration::from_secs(10)))
+		}
+	}
+
+	fn get_block_to_resync(&self) -> Result<Option<BusyBlock>, db::Error> {
+		let mut busy = self.busy_set.lock().unwrap();
+		for it in self.queue.iter()? {
+			let (time_bytes, hash_bytes) = it?;
+			if !busy.contains(&time_bytes) {
+				busy.insert(time_bytes.clone());
+				return Ok(Some(BusyBlock {
+					time_bytes,
+					hash_bytes,
+					busy_set: self.busy_set.clone(),
+				}));
+			}
+		}
+		Ok(None)
+	}
+
+	async fn resync_block(&self, manager: &BlockManager, hash: &Hash) -> Result<(), Error> {
+		let BlockStatus { exists, needed } = manager.check_block_status(hash).await?;
+
+		if exists != needed.is_needed() || exists != needed.is_nonzero() {
+			debug!(
+				"Resync block {:?}: exists {}, nonzero rc {}, deletable {}",
+				hash,
+				exists,
+				needed.is_nonzero(),
+				needed.is_deletable(),
+			);
+		}
+
+		if exists && needed.is_deletable() {
+			info!("Resync block {:?}: offloading and deleting", hash);
+
+			let mut who = manager.replication.write_nodes(hash);
+			if who.len() < manager.replication.write_quorum() {
+				return Err(Error::Message("Not trying to offload block because we don't have a quorum of nodes to write to".to_string()));
+			}
+			who.retain(|id| *id != manager.system.id);
+
+			let who_needs_resps = manager
+				.system
+				.rpc
+				.call_many(
+					&manager.endpoint,
+					&who,
+					BlockRpc::NeedBlockQuery(*hash),
+					RequestStrategy::with_priority(PRIO_BACKGROUND)
+						.with_timeout(NEED_BLOCK_QUERY_TIMEOUT),
+				)
+				.await?;
+
+			let mut need_nodes = vec![];
+			for (node, needed) in who_needs_resps {
+				match needed.err_context("NeedBlockQuery RPC")? {
+					BlockRpc::NeedBlockReply(needed) => {
+						if needed {
+							need_nodes.push(node);
+						}
+					}
+					m => {
+						return Err(Error::unexpected_rpc_message(m));
+					}
+				}
+			}
+
+			if !need_nodes.is_empty() {
+				trace!(
+					"Block {:?} needed by {} nodes, sending",
+					hash,
+					need_nodes.len()
+				);
+
+				for node in need_nodes.iter() {
+					manager
+						.metrics
+						.resync_send_counter
+						.add(1, &[KeyValue::new("to", format!("{:?}", node))]);
+				}
+
+				let block = manager.read_block(hash).await?;
+				let (header, bytes) = block.into_parts();
+				let put_block_message = Req::new(BlockRpc::PutBlock {
+					hash: *hash,
+					header,
+				})?
+				.with_stream_from_buffer(bytes);
+				manager
+					.system
+					.rpc
+					.try_call_many(
+						&manager.endpoint,
+						&need_nodes[..],
+						put_block_message,
+						RequestStrategy::with_priority(PRIO_BACKGROUND)
+							.with_quorum(need_nodes.len())
+							.with_timeout(BLOCK_RW_TIMEOUT),
+					)
+					.await
+					.err_context("PutBlock RPC")?;
+			}
+			info!(
+				"Deleting unneeded block {:?}, offload finished ({} / {})",
+				hash,
+				need_nodes.len(),
+				who.len()
+			);
+
+			manager.delete_if_unneeded(hash).await?;
+
+			manager.rc.clear_deleted_block_rc(hash)?;
+		}
+
+		if needed.is_nonzero() && !exists {
+			info!(
+				"Resync block {:?}: fetching absent but needed block (refcount > 0)",
+				hash
+			);
+
+			let block_data = manager.rpc_get_raw_block(hash, None).await?;
+
+			manager.metrics.resync_recv_counter.add(1);
+
+			manager.write_block(hash, &block_data).await?;
+		}
+
+		Ok(())
+	}
+
+	async fn update_persisted(
+		&self,
+		update: impl Fn(&mut ResyncPersistedConfig),
+	) -> Result<(), Error> {
+		let mut cfg: ResyncPersistedConfig = *self.persisted.load().as_ref();
+		update(&mut cfg);
+		self.persister.save_async(&cfg).await?;
+		self.persisted.store(Arc::new(cfg));
+		self.notify.notify_waiters();
+		Ok(())
+	}
+
+	pub async fn set_n_workers(&self, n_workers: usize) -> Result<(), Error> {
+		if !(1..=MAX_RESYNC_WORKERS).contains(&n_workers) {
+			return Err(Error::Message(format!(
+				"Invalid number of resync workers, must be between 1 and {}",
+				MAX_RESYNC_WORKERS
+			)));
+		}
+		self.update_persisted(|cfg| cfg.n_workers = n_workers).await
+	}
+
+	pub async fn set_tranquility(&self, tranquility: u32) -> Result<(), Error> {
+		self.update_persisted(|cfg| cfg.tranquility = tranquility)
+			.await
+	}
+}
+
+impl Drop for BusyBlock {
+	fn drop(&mut self) {
+		let mut busy = self.busy_set.lock().unwrap();
+		busy.remove(&self.time_bytes);
+	}
+}
+
+pub(crate) struct ResyncWorker {
+	index: usize,
+	manager: Arc<BlockManager>,
+	tranquilizer: Tranquilizer,
+	next_delay: Duration,
+}
+
+impl ResyncWorker {
+	pub(crate) fn new(index: usize, manager: Arc<BlockManager>) -> Self {
+		Self {
+			index,
+			manager,
+			tranquilizer: Tranquilizer::new(30),
+			next_delay: Duration::from_secs(10),
+		}
+	}
+}
+
+#[async_trait]
+impl Worker for ResyncWorker {
+	fn name(&self) -> String {
+		format!("Block resync worker #{}", self.index + 1)
+	}
+
+	fn info(&self) -> Option<String> {
+		let persisted = self.manager.resync.persisted.load();
+
+		if self.index >= persisted.n_workers {
+			return Some("(unused)".into());
+		}
+
+		let mut ret = vec![];
+		ret.push(format!("tranquility = {}", persisted.tranquility));
+
+		let qlen = self.manager.resync.queue_len().unwrap_or(0);
+		if qlen > 0 {
+			ret.push(format!("{} blocks in queue", qlen));
+		}
+
+		let elen = self.manager.resync.errors_len().unwrap_or(0);
+		if elen > 0 {
+			ret.push(format!("{} blocks in error state", elen));
+		}
+
+		Some(ret.join(", "))
+	}
+
+	async fn work(&mut self, _must_exit: &mut watch::Receiver<bool>) -> Result<WorkerState, Error> {
+		if self.index >= self.manager.resync.persisted.load().n_workers {
+			return Ok(WorkerState::Idle);
+		}
+
+		self.tranquilizer.reset();
+		match self.manager.resync.resync_iter(&self.manager).await {
+			Ok(ResyncIterResult::BusyDidSomething) => Ok(self
+				.tranquilizer
+				.tranquilize_worker(self.manager.resync.persisted.load().tranquility)),
+			Ok(ResyncIterResult::BusyDidNothing) => Ok(WorkerState::Busy),
+			Ok(ResyncIterResult::IdleFor(delay)) => {
+				self.next_delay = delay;
+				Ok(WorkerState::Idle)
+			}
+			Err(e) => {
+				// The errors that we have here are only Sled errors
+				// We don't really know how to handle them so just ¯\_(ツ)_/¯
+				// (there is kind of an assumption that Sled won't error on us,
+				// if it does there is not much we can do -- TODO should we just panic?)
+				// Here we just give the error to the worker manager,
+				// it will print it to the logs and increment a counter
+				Err(e.into())
+			}
+		}
+	}
+
+	async fn wait_for_work(&mut self, _must_exit: &watch::Receiver<bool>) -> WorkerState {
+		while self.index >= self.manager.resync.persisted.load().n_workers {
+			self.manager.resync.notify.notified().await
+		}
+
+		select! {
+			_ = tokio::time::sleep(self.next_delay) => (),
+			_ = self.manager.resync.notify.notified() => (),
+		};
+
+		WorkerState::Busy
+	}
+}
+
+/// Counts the number of errors when resyncing a block,
+/// and the time of the last try.
+/// Used to implement exponential backoff.
+#[derive(Clone, Copy, Debug)]
+struct ErrorCounter {
+	errors: u64,
+	last_try: u64,
+}
+
+impl ErrorCounter {
+	fn new(now: u64) -> Self {
+		Self {
+			errors: 1,
+			last_try: now,
+		}
+	}
+
+	fn decode(data: &[u8]) -> Self {
+		Self {
+			errors: u64::from_be_bytes(data[0..8].try_into().unwrap()),
+			last_try: u64::from_be_bytes(data[8..16].try_into().unwrap()),
+		}
+	}
+	fn encode(&self) -> Vec<u8> {
+		[
+			u64::to_be_bytes(self.errors),
+			u64::to_be_bytes(self.last_try),
+		]
+		.concat()
+	}
+
+	fn add1(self, now: u64) -> Self {
+		Self {
+			errors: self.errors + 1,
+			last_try: now,
+		}
+	}
+
+	fn delay_msec(&self) -> u64 {
+		(RESYNC_RETRY_DELAY.as_millis() as u64)
+			<< std::cmp::min(self.errors - 1, RESYNC_RETRY_DELAY_MAX_BACKOFF_POWER)
+	}
+	fn next_try(&self) -> u64 {
+		self.last_try + self.delay_msec()
+	}
+}

src/db/Cargo.toml

@@ -0,0 +1,39 @@
+[package]
+name = "garage_db"
+version = "0.8.0"
+authors = ["Alex Auvolat <alex@adnab.me>"]
+edition = "2018"
+license = "AGPL-3.0"
+description = "Abstraction over multiple key/value storage engines that supports transactions"
+repository = "https://git.deuxfleurs.fr/Deuxfleurs/garage"
+readme = "../../README.md"
+
+[lib]
+path = "lib.rs"
+
+[[bin]]
+name = "convert"
+path = "bin/convert.rs"
+required-features = ["cli"]
+
+[dependencies]
+err-derive = "0.3"
+hexdump = "0.1"
+tracing = "0.1.30"
+
+heed = { version = "0.11", default-features = false, features = ["lmdb"], optional = true }
+rusqlite = { version = "0.27", optional = true }
+sled = { version = "0.34", optional = true }
+
+# cli deps
+clap = { version = "3.1.18", optional = true, features = ["derive", "env"] }
+pretty_env_logger = { version = "0.4", optional = true }
+
+[dev-dependencies]
+mktemp = "0.4"
+
+[features]
+bundled-libs = [ "rusqlite/bundled" ]
+cli = ["clap", "pretty_env_logger"]
+lmdb = [ "heed" ]
+sqlite = [ "rusqlite" ]

src/db/bin/convert.rs

@@ -0,0 +1,69 @@
+use std::path::PathBuf;
+
+use garage_db::*;
+
+use clap::Parser;
+
+/// K2V command line interface
+#[derive(Parser, Debug)]
+#[clap(author, version, about, long_about = None)]
+struct Args {
+	/// Input DB path
+	#[clap(short = 'i')]
+	input_path: PathBuf,
+	/// Input DB engine
+	#[clap(short = 'a')]
+	input_engine: String,
+
+	/// Output DB path
+	#[clap(short = 'o')]
+	output_path: PathBuf,
+	/// Output DB engine
+	#[clap(short = 'b')]
+	output_engine: String,
+}
+
+fn main() {
+	let args = Args::parse();
+	pretty_env_logger::init();
+
+	match do_conversion(args) {
+		Ok(()) => println!("Success!"),
+		Err(e) => eprintln!("Error: {}", e),
+	}
+}
+
+fn do_conversion(args: Args) -> Result<()> {
+	let input = open_db(args.input_path, args.input_engine)?;
+	let output = open_db(args.output_path, args.output_engine)?;
+	output.import(&input)?;
+	Ok(())
+}
+
+fn open_db(path: PathBuf, engine: String) -> Result<Db> {
+	match engine.as_str() {
+		"sled" => {
+			let db = sled_adapter::sled::Config::default().path(&path).open()?;
+			Ok(sled_adapter::SledDb::init(db))
+		}
+		"sqlite" | "sqlite3" | "rusqlite" => {
+			let db = sqlite_adapter::rusqlite::Connection::open(&path)?;
+			Ok(sqlite_adapter::SqliteDb::init(db))
+		}
+		"lmdb" | "heed" => {
+			std::fs::create_dir_all(&path).map_err(|e| {
+				Error(format!("Unable to create LMDB data directory: {}", e).into())
+			})?;
+
+			let map_size = lmdb_adapter::recommended_map_size();
+
+			let db = lmdb_adapter::heed::EnvOpenOptions::new()
+				.max_dbs(100)
+				.map_size(map_size)
+				.open(&path)
+				.unwrap();
+			Ok(lmdb_adapter::LmdbDb::init(db))
+		}
+		e => Err(Error(format!("Invalid DB engine: {}", e).into())),
+	}
+}

src/db/counted_tree_hack.rs

@@ -0,0 +1,127 @@
+//! This hack allows a db tree to keep in RAM a counter of the number of entries
+//! it contains, which is used to call .len() on it.  This is usefull only for
+//! the sled backend where .len() otherwise would have to traverse the whole
+//! tree to count items.  For sqlite and lmdb, this is mostly useless (but
+//! hopefully not harmfull!). Note that a CountedTree cannot be part of a
+//! transaction.
+
+use std::sync::{
+	atomic::{AtomicUsize, Ordering},
+	Arc,
+};
+
+use crate::{Result, Tree, TxError, Value, ValueIter};
+
+#[derive(Clone)]
+pub struct CountedTree(Arc<CountedTreeInternal>);
+
+struct CountedTreeInternal {
+	tree: Tree,
+	len: AtomicUsize,
+}
+
+impl CountedTree {
+	pub fn new(tree: Tree) -> Result<Self> {
+		let len = tree.len()?;
+		Ok(Self(Arc::new(CountedTreeInternal {
+			tree,
+			len: AtomicUsize::new(len),
+		})))
+	}
+
+	pub fn len(&self) -> usize {
+		self.0.len.load(Ordering::SeqCst)
+	}
+
+	pub fn is_empty(&self) -> bool {
+		self.len() == 0
+	}
+
+	pub fn get<K: AsRef<[u8]>>(&self, key: K) -> Result<Option<Value>> {
+		self.0.tree.get(key)
+	}
+
+	pub fn first(&self) -> Result<Option<(Value, Value)>> {
+		self.0.tree.first()
+	}
+
+	pub fn iter(&self) -> Result<ValueIter<'_>> {
+		self.0.tree.iter()
+	}
+
+	// ---- writing functions ----
+
+	pub fn insert<K, V>(&self, key: K, value: V) -> Result<Option<Value>>
+	where
+		K: AsRef<[u8]>,
+		V: AsRef<[u8]>,
+	{
+		let old_val = self.0.tree.insert(key, value)?;
+		if old_val.is_none() {
+			self.0.len.fetch_add(1, Ordering::SeqCst);
+		}
+		Ok(old_val)
+	}
+
+	pub fn remove<K: AsRef<[u8]>>(&self, key: K) -> Result<Option<Value>> {
+		let old_val = self.0.tree.remove(key)?;
+		if old_val.is_some() {
+			self.0.len.fetch_sub(1, Ordering::SeqCst);
+		}
+		Ok(old_val)
+	}
+
+	pub fn compare_and_swap<K, OV, NV>(
+		&self,
+		key: K,
+		expected_old: Option<OV>,
+		new: Option<NV>,
+	) -> Result<bool>
+	where
+		K: AsRef<[u8]>,
+		OV: AsRef<[u8]>,
+		NV: AsRef<[u8]>,
+	{
+		let old_some = expected_old.is_some();
+		let new_some = new.is_some();
+
+		let tx_res = self.0.tree.db().transaction(|mut tx| {
+			let old_val = tx.get(&self.0.tree, &key)?;
+			let is_same = match (&old_val, &expected_old) {
+				(None, None) => true,
+				(Some(x), Some(y)) if x == y.as_ref() => true,
+				_ => false,
+			};
+			if is_same {
+				match &new {
+					Some(v) => {
+						tx.insert(&self.0.tree, &key, v)?;
+					}
+					None => {
+						tx.remove(&self.0.tree, &key)?;
+					}
+				}
+				tx.commit(())
+			} else {
+				tx.abort(())
+			}
+		});
+
+		match tx_res {
+			Ok(()) => {
+				match (old_some, new_some) {
+					(false, true) => {
+						self.0.len.fetch_add(1, Ordering::SeqCst);
+					}
+					(true, false) => {
+						self.0.len.fetch_sub(1, Ordering::SeqCst);
+					}
+					_ => (),
+				}
+				Ok(true)
+			}
+			Err(TxError::Abort(())) => Ok(false),
+			Err(TxError::Db(e)) => Err(e),
+		}
+	}
+}

src/db/lib.rs

@@ -0,0 +1,416 @@
+#[macro_use]
+#[cfg(feature = "sqlite")]
+extern crate tracing;
+
+#[cfg(not(any(feature = "lmdb", feature = "sled", feature = "sqlite")))]
+compile_error!("Must activate the Cargo feature for at least one DB engine: lmdb, sled or sqlite.");
+
+#[cfg(feature = "lmdb")]
+pub mod lmdb_adapter;
+#[cfg(feature = "sled")]
+pub mod sled_adapter;
+#[cfg(feature = "sqlite")]
+pub mod sqlite_adapter;
+
+pub mod counted_tree_hack;
+
+#[cfg(test)]
+pub mod test;
+
+use core::ops::{Bound, RangeBounds};
+
+use std::borrow::Cow;
+use std::cell::Cell;
+use std::sync::Arc;
+
+use err_derive::Error;
+
+#[derive(Clone)]
+pub struct Db(pub(crate) Arc<dyn IDb>);
+
+pub struct Transaction<'a>(&'a mut dyn ITx);
+
+#[derive(Clone)]
+pub struct Tree(Arc<dyn IDb>, usize);
+
+pub type Value = Vec<u8>;
+pub type ValueIter<'a> = Box<dyn std::iter::Iterator<Item = Result<(Value, Value)>> + 'a>;
+pub type TxValueIter<'a> = Box<dyn std::iter::Iterator<Item = TxOpResult<(Value, Value)>> + 'a>;
+
+// ----
+
+#[derive(Debug, Error)]
+#[error(display = "{}", _0)]
+pub struct Error(pub Cow<'static, str>);
+
+pub type Result<T> = std::result::Result<T, Error>;
+
+#[derive(Debug, Error)]
+#[error(display = "{}", _0)]
+pub struct TxOpError(pub(crate) Error);
+pub type TxOpResult<T> = std::result::Result<T, TxOpError>;
+
+pub enum TxError<E> {
+	Abort(E),
+	Db(Error),
+}
+pub type TxResult<R, E> = std::result::Result<R, TxError<E>>;
+
+impl<E> From<TxOpError> for TxError<E> {
+	fn from(e: TxOpError) -> TxError<E> {
+		TxError::Db(e.0)
+	}
+}
+
+pub fn unabort<R, E>(res: TxResult<R, E>) -> TxOpResult<std::result::Result<R, E>> {
+	match res {
+		Ok(v) => Ok(Ok(v)),
+		Err(TxError::Abort(e)) => Ok(Err(e)),
+		Err(TxError::Db(e)) => Err(TxOpError(e)),
+	}
+}
+
+// ----
+
+impl Db {
+	pub fn engine(&self) -> String {
+		self.0.engine()
+	}
+
+	pub fn open_tree<S: AsRef<str>>(&self, name: S) -> Result<Tree> {
+		let tree_id = self.0.open_tree(name.as_ref())?;
+		Ok(Tree(self.0.clone(), tree_id))
+	}
+
+	pub fn list_trees(&self) -> Result<Vec<String>> {
+		self.0.list_trees()
+	}
+
+	pub fn transaction<R, E, F>(&self, fun: F) -> TxResult<R, E>
+	where
+		F: Fn(Transaction<'_>) -> TxResult<R, E>,
+	{
+		let f = TxFn {
+			function: fun,
+			result: Cell::new(None),
+		};
+		let tx_res = self.0.transaction(&f);
+		let ret = f
+			.result
+			.into_inner()
+			.expect("Transaction did not store result");
+
+		match tx_res {
+			Ok(()) => {
+				assert!(matches!(ret, Ok(_)));
+				ret
+			}
+			Err(TxError::Abort(())) => {
+				assert!(matches!(ret, Err(TxError::Abort(_))));
+				ret
+			}
+			Err(TxError::Db(e2)) => match ret {
+				// Ok was stored -> the error occured when finalizing
+				// transaction
+				Ok(_) => Err(TxError::Db(e2)),
+				// An error was already stored: that's the one we want to
+				// return
+				Err(TxError::Db(e)) => Err(TxError::Db(e)),
+				_ => unreachable!(),
+			},
+		}
+	}
+
+	pub fn import(&self, other: &Db) -> Result<()> {
+		let existing_trees = self.list_trees()?;
+		if !existing_trees.is_empty() {
+			return Err(Error(
+				format!(
+					"destination database already contains data: {:?}",
+					existing_trees
+				)
+				.into(),
+			));
+		}
+
+		let tree_names = other.list_trees()?;
+		for name in tree_names {
+			let tree = self.open_tree(&name)?;
+			if tree.len()? > 0 {
+				return Err(Error(format!("tree {} already contains data", name).into()));
+			}
+
+			let ex_tree = other.open_tree(&name)?;
+
+			let tx_res = self.transaction(|mut tx| {
+				let mut i = 0;
+				for item in ex_tree.iter().map_err(TxError::Abort)? {
+					let (k, v) = item.map_err(TxError::Abort)?;
+					tx.insert(&tree, k, v)?;
+					i += 1;
+					if i % 1000 == 0 {
+						println!("{}: imported {}", name, i);
+					}
+				}
+				tx.commit(i)
+			});
+			let total = match tx_res {
+				Err(TxError::Db(e)) => return Err(e),
+				Err(TxError::Abort(e)) => return Err(e),
+				Ok(x) => x,
+			};
+
+			println!("{}: finished importing, {} items", name, total);
+		}
+		Ok(())
+	}
+}
+
+#[allow(clippy::len_without_is_empty)]
+impl Tree {
+	#[inline]
+	pub fn db(&self) -> Db {
+		Db(self.0.clone())
+	}
+
+	#[inline]
+	pub fn get<T: AsRef<[u8]>>(&self, key: T) -> Result<Option<Value>> {
+		self.0.get(self.1, key.as_ref())
+	}
+	#[inline]
+	pub fn len(&self) -> Result<usize> {
+		self.0.len(self.1)
+	}
+
+	#[inline]
+	pub fn first(&self) -> Result<Option<(Value, Value)>> {
+		self.iter()?.next().transpose()
+	}
+	#[inline]
+	pub fn get_gt<T: AsRef<[u8]>>(&self, from: T) -> Result<Option<(Value, Value)>> {
+		self.range((Bound::Excluded(from), Bound::Unbounded))?
+			.next()
+			.transpose()
+	}
+
+	/// Returns the old value if there was one
+	#[inline]
+	pub fn insert<T: AsRef<[u8]>, U: AsRef<[u8]>>(
+		&self,
+		key: T,
+		value: U,
+	) -> Result<Option<Value>> {
+		self.0.insert(self.1, key.as_ref(), value.as_ref())
+	}
+	/// Returns the old value if there was one
+	#[inline]
+	pub fn remove<T: AsRef<[u8]>>(&self, key: T) -> Result<Option<Value>> {
+		self.0.remove(self.1, key.as_ref())
+	}
+	/// Clears all values from the tree
+	#[inline]
+	pub fn clear(&self) -> Result<()> {
+		self.0.clear(self.1)
+	}
+
+	#[inline]
+	pub fn iter(&self) -> Result<ValueIter<'_>> {
+		self.0.iter(self.1)
+	}
+	#[inline]
+	pub fn iter_rev(&self) -> Result<ValueIter<'_>> {
+		self.0.iter_rev(self.1)
+	}
+
+	#[inline]
+	pub fn range<K, R>(&self, range: R) -> Result<ValueIter<'_>>
+	where
+		K: AsRef<[u8]>,
+		R: RangeBounds<K>,
+	{
+		let sb = range.start_bound();
+		let eb = range.end_bound();
+		self.0.range(self.1, get_bound(sb), get_bound(eb))
+	}
+	#[inline]
+	pub fn range_rev<K, R>(&self, range: R) -> Result<ValueIter<'_>>
+	where
+		K: AsRef<[u8]>,
+		R: RangeBounds<K>,
+	{
+		let sb = range.start_bound();
+		let eb = range.end_bound();
+		self.0.range_rev(self.1, get_bound(sb), get_bound(eb))
+	}
+}
+
+#[allow(clippy::len_without_is_empty)]
+impl<'a> Transaction<'a> {
+	#[inline]
+	pub fn get<T: AsRef<[u8]>>(&self, tree: &Tree, key: T) -> TxOpResult<Option<Value>> {
+		self.0.get(tree.1, key.as_ref())
+	}
+	#[inline]
+	pub fn len(&self, tree: &Tree) -> TxOpResult<usize> {
+		self.0.len(tree.1)
+	}
+
+	/// Returns the old value if there was one
+	#[inline]
+	pub fn insert<T: AsRef<[u8]>, U: AsRef<[u8]>>(
+		&mut self,
+		tree: &Tree,
+		key: T,
+		value: U,
+	) -> TxOpResult<Option<Value>> {
+		self.0.insert(tree.1, key.as_ref(), value.as_ref())
+	}
+	/// Returns the old value if there was one
+	#[inline]
+	pub fn remove<T: AsRef<[u8]>>(&mut self, tree: &Tree, key: T) -> TxOpResult<Option<Value>> {
+		self.0.remove(tree.1, key.as_ref())
+	}
+
+	#[inline]
+	pub fn iter(&self, tree: &Tree) -> TxOpResult<TxValueIter<'_>> {
+		self.0.iter(tree.1)
+	}
+	#[inline]
+	pub fn iter_rev(&self, tree: &Tree) -> TxOpResult<TxValueIter<'_>> {
+		self.0.iter_rev(tree.1)
+	}
+
+	#[inline]
+	pub fn range<K, R>(&self, tree: &Tree, range: R) -> TxOpResult<TxValueIter<'_>>
+	where
+		K: AsRef<[u8]>,
+		R: RangeBounds<K>,
+	{
+		let sb = range.start_bound();
+		let eb = range.end_bound();
+		self.0.range(tree.1, get_bound(sb), get_bound(eb))
+	}
+	#[inline]
+	pub fn range_rev<K, R>(&self, tree: &Tree, range: R) -> TxOpResult<TxValueIter<'_>>
+	where
+		K: AsRef<[u8]>,
+		R: RangeBounds<K>,
+	{
+		let sb = range.start_bound();
+		let eb = range.end_bound();
+		self.0.range_rev(tree.1, get_bound(sb), get_bound(eb))
+	}
+
+	// ----
+
+	#[inline]
+	pub fn abort<R, E>(self, e: E) -> TxResult<R, E> {
+		Err(TxError::Abort(e))
+	}
+
+	#[inline]
+	pub fn commit<R, E>(self, r: R) -> TxResult<R, E> {
+		Ok(r)
+	}
+}
+
+// ---- Internal interfaces
+
+pub(crate) trait IDb: Send + Sync {
+	fn engine(&self) -> String;
+	fn open_tree(&self, name: &str) -> Result<usize>;
+	fn list_trees(&self) -> Result<Vec<String>>;
+
+	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>>;
+	fn len(&self, tree: usize) -> Result<usize>;
+
+	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<Option<Value>>;
+	fn remove(&self, tree: usize, key: &[u8]) -> Result<Option<Value>>;
+	fn clear(&self, tree: usize) -> Result<()>;
+
+	fn iter(&self, tree: usize) -> Result<ValueIter<'_>>;
+	fn iter_rev(&self, tree: usize) -> Result<ValueIter<'_>>;
+
+	fn range<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>>;
+	fn range_rev<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>>;
+
+	fn transaction(&self, f: &dyn ITxFn) -> TxResult<(), ()>;
+}
+
+pub(crate) trait ITx {
+	fn get(&self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>>;
+	fn len(&self, tree: usize) -> TxOpResult<usize>;
+
+	fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult<Option<Value>>;
+	fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>>;
+
+	fn iter(&self, tree: usize) -> TxOpResult<TxValueIter<'_>>;
+	fn iter_rev(&self, tree: usize) -> TxOpResult<TxValueIter<'_>>;
+
+	fn range<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>>;
+	fn range_rev<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>>;
+}
+
+pub(crate) trait ITxFn {
+	fn try_on(&self, tx: &mut dyn ITx) -> TxFnResult;
+}
+
+pub(crate) enum TxFnResult {
+	Ok,
+	Abort,
+	DbErr,
+}
+
+struct TxFn<F, R, E>
+where
+	F: Fn(Transaction<'_>) -> TxResult<R, E>,
+{
+	function: F,
+	result: Cell<Option<TxResult<R, E>>>,
+}
+
+impl<F, R, E> ITxFn for TxFn<F, R, E>
+where
+	F: Fn(Transaction<'_>) -> TxResult<R, E>,
+{
+	fn try_on(&self, tx: &mut dyn ITx) -> TxFnResult {
+		let res = (self.function)(Transaction(tx));
+		let res2 = match &res {
+			Ok(_) => TxFnResult::Ok,
+			Err(TxError::Abort(_)) => TxFnResult::Abort,
+			Err(TxError::Db(_)) => TxFnResult::DbErr,
+		};
+		self.result.set(Some(res));
+		res2
+	}
+}
+
+// ----
+
+fn get_bound<K: AsRef<[u8]>>(b: Bound<&K>) -> Bound<&[u8]> {
+	match b {
+		Bound::Included(v) => Bound::Included(v.as_ref()),
+		Bound::Excluded(v) => Bound::Excluded(v.as_ref()),
+		Bound::Unbounded => Bound::Unbounded,
+	}
+}

src/db/lmdb_adapter.rs

@@ -0,0 +1,350 @@
+use core::ops::Bound;
+use core::ptr::NonNull;
+
+use std::collections::HashMap;
+use std::convert::TryInto;
+use std::sync::{Arc, RwLock};
+
+use heed::types::ByteSlice;
+use heed::{BytesDecode, Env, RoTxn, RwTxn, UntypedDatabase as Database};
+
+use crate::{
+	Db, Error, IDb, ITx, ITxFn, Result, TxError, TxFnResult, TxOpError, TxOpResult, TxResult,
+	TxValueIter, Value, ValueIter,
+};
+
+pub use heed;
+
+// -- err
+
+impl From<heed::Error> for Error {
+	fn from(e: heed::Error) -> Error {
+		Error(format!("LMDB: {}", e).into())
+	}
+}
+
+impl From<heed::Error> for TxOpError {
+	fn from(e: heed::Error) -> TxOpError {
+		TxOpError(e.into())
+	}
+}
+
+// -- db
+
+pub struct LmdbDb {
+	db: heed::Env,
+	trees: RwLock<(Vec<Database>, HashMap<String, usize>)>,
+}
+
+impl LmdbDb {
+	pub fn init(db: Env) -> Db {
+		let s = Self {
+			db,
+			trees: RwLock::new((Vec::new(), HashMap::new())),
+		};
+		Db(Arc::new(s))
+	}
+
+	fn get_tree(&self, i: usize) -> Result<Database> {
+		self.trees
+			.read()
+			.unwrap()
+			.0
+			.get(i)
+			.cloned()
+			.ok_or_else(|| Error("invalid tree id".into()))
+	}
+}
+
+impl IDb for LmdbDb {
+	fn engine(&self) -> String {
+		"LMDB (using Heed crate)".into()
+	}
+
+	fn open_tree(&self, name: &str) -> Result<usize> {
+		let mut trees = self.trees.write().unwrap();
+		if let Some(i) = trees.1.get(name) {
+			Ok(*i)
+		} else {
+			let tree = self.db.create_database(Some(name))?;
+			let i = trees.0.len();
+			trees.0.push(tree);
+			trees.1.insert(name.to_string(), i);
+			Ok(i)
+		}
+	}
+
+	fn list_trees(&self) -> Result<Vec<String>> {
+		let tree0 = match self.db.open_database::<heed::types::Str, ByteSlice>(None)? {
+			Some(x) => x,
+			None => return Ok(vec![]),
+		};
+
+		let mut ret = vec![];
+		let tx = self.db.read_txn()?;
+		for item in tree0.iter(&tx)? {
+			let (tree_name, _) = item?;
+			ret.push(tree_name.to_string());
+		}
+		drop(tx);
+
+		let mut ret2 = vec![];
+		for tree_name in ret {
+			if self
+				.db
+				.open_database::<ByteSlice, ByteSlice>(Some(&tree_name))?
+				.is_some()
+			{
+				ret2.push(tree_name);
+			}
+		}
+
+		Ok(ret2)
+	}
+
+	// ----
+
+	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+
+		let tx = self.db.read_txn()?;
+		let val = tree.get(&tx, key)?;
+		match val {
+			None => Ok(None),
+			Some(v) => Ok(Some(v.to_vec())),
+		}
+	}
+
+	fn len(&self, tree: usize) -> Result<usize> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		Ok(tree.len(&tx)?.try_into().unwrap())
+	}
+
+	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let mut tx = self.db.write_txn()?;
+		let old_val = tree.get(&tx, key)?.map(Vec::from);
+		tree.put(&mut tx, key, value)?;
+		tx.commit()?;
+		Ok(old_val)
+	}
+
+	fn remove(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let mut tx = self.db.write_txn()?;
+		let old_val = tree.get(&tx, key)?.map(Vec::from);
+		tree.delete(&mut tx, key)?;
+		tx.commit()?;
+		Ok(old_val)
+	}
+
+	fn clear(&self, tree: usize) -> Result<()> {
+		let tree = self.get_tree(tree)?;
+		let mut tx = self.db.write_txn()?;
+		tree.clear(&mut tx)?;
+		tx.commit()?;
+		Ok(())
+	}
+
+	fn iter(&self, tree: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		TxAndIterator::make(tx, |tx| Ok(tree.iter(tx)?))
+	}
+
+	fn iter_rev(&self, tree: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		TxAndIterator::make(tx, |tx| Ok(tree.rev_iter(tx)?))
+	}
+
+	fn range<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		TxAndIterator::make(tx, |tx| Ok(tree.range(tx, &(low, high))?))
+	}
+	fn range_rev<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		TxAndIterator::make(tx, |tx| Ok(tree.rev_range(tx, &(low, high))?))
+	}
+
+	// ----
+
+	fn transaction(&self, f: &dyn ITxFn) -> TxResult<(), ()> {
+		let trees = self.trees.read().unwrap();
+		let mut tx = LmdbTx {
+			trees: &trees.0[..],
+			tx: self
+				.db
+				.write_txn()
+				.map_err(Error::from)
+				.map_err(TxError::Db)?,
+		};
+
+		let res = f.try_on(&mut tx);
+		match res {
+			TxFnResult::Ok => {
+				tx.tx.commit().map_err(Error::from).map_err(TxError::Db)?;
+				Ok(())
+			}
+			TxFnResult::Abort => {
+				tx.tx.abort().map_err(Error::from).map_err(TxError::Db)?;
+				Err(TxError::Abort(()))
+			}
+			TxFnResult::DbErr => {
+				tx.tx.abort().map_err(Error::from).map_err(TxError::Db)?;
+				Err(TxError::Db(Error(
+					"(this message will be discarded)".into(),
+				)))
+			}
+		}
+	}
+}
+
+// ----
+
+struct LmdbTx<'a> {
+	trees: &'a [Database],
+	tx: RwTxn<'a, 'a>,
+}
+
+impl<'a> LmdbTx<'a> {
+	fn get_tree(&self, i: usize) -> TxOpResult<&Database> {
+		self.trees.get(i).ok_or_else(|| {
+			TxOpError(Error(
+				"invalid tree id (it might have been openned after the transaction started)".into(),
+			))
+		})
+	}
+}
+
+impl<'a> ITx for LmdbTx<'a> {
+	fn get(&self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		match tree.get(&self.tx, key)? {
+			Some(v) => Ok(Some(v.to_vec())),
+			None => Ok(None),
+		}
+	}
+	fn len(&self, _tree: usize) -> TxOpResult<usize> {
+		unimplemented!(".len() in transaction not supported with LMDB backend")
+	}
+
+	fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = *self.get_tree(tree)?;
+		let old_val = tree.get(&self.tx, key)?.map(Vec::from);
+		tree.put(&mut self.tx, key, value)?;
+		Ok(old_val)
+	}
+	fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = *self.get_tree(tree)?;
+		let old_val = tree.get(&self.tx, key)?.map(Vec::from);
+		tree.delete(&mut self.tx, key)?;
+		Ok(old_val)
+	}
+
+	fn iter(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with LMDB backend");
+	}
+	fn iter_rev(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with LMDB backend");
+	}
+
+	fn range<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with LMDB backend");
+	}
+	fn range_rev<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with LMDB backend");
+	}
+}
+
+// ----
+
+type IteratorItem<'a> = heed::Result<(
+	<ByteSlice as BytesDecode<'a>>::DItem,
+	<ByteSlice as BytesDecode<'a>>::DItem,
+)>;
+
+struct TxAndIterator<'a, I>
+where
+	I: Iterator<Item = IteratorItem<'a>> + 'a,
+{
+	tx: RoTxn<'a>,
+	iter: Option<I>,
+}
+
+impl<'a, I> TxAndIterator<'a, I>
+where
+	I: Iterator<Item = IteratorItem<'a>> + 'a,
+{
+	fn make<F>(tx: RoTxn<'a>, iterfun: F) -> Result<ValueIter<'a>>
+	where
+		F: FnOnce(&'a RoTxn<'a>) -> Result<I>,
+	{
+		let mut res = TxAndIterator { tx, iter: None };
+
+		let tx = unsafe { NonNull::from(&res.tx).as_ref() };
+		res.iter = Some(iterfun(tx)?);
+
+		Ok(Box::new(res))
+	}
+}
+
+impl<'a, I> Drop for TxAndIterator<'a, I>
+where
+	I: Iterator<Item = IteratorItem<'a>> + 'a,
+{
+	fn drop(&mut self) {
+		drop(self.iter.take());
+	}
+}
+
+impl<'a, I> Iterator for TxAndIterator<'a, I>
+where
+	I: Iterator<Item = IteratorItem<'a>> + 'a,
+{
+	type Item = Result<(Value, Value)>;
+
+	fn next(&mut self) -> Option<Self::Item> {
+		match self.iter.as_mut().unwrap().next() {
+			None => None,
+			Some(Err(e)) => Some(Err(e.into())),
+			Some(Ok((k, v))) => Some(Ok((k.to_vec(), v.to_vec()))),
+		}
+	}
+}
+
+// ----
+
+#[cfg(target_pointer_width = "64")]
+pub fn recommended_map_size() -> usize {
+	1usize << 40
+}
+
+#[cfg(target_pointer_width = "32")]
+pub fn recommended_map_size() -> usize {
+	warn!("LMDB is not recommended on 32-bit systems, database size will be limited");
+	1usize << 30
+}

src/db/sled_adapter.rs

@@ -0,0 +1,266 @@
+use core::ops::Bound;
+
+use std::cell::Cell;
+use std::collections::HashMap;
+use std::sync::{Arc, RwLock};
+
+use sled::transaction::{
+	ConflictableTransactionError, TransactionError, Transactional, TransactionalTree,
+	UnabortableTransactionError,
+};
+
+use crate::{
+	Db, Error, IDb, ITx, ITxFn, Result, TxError, TxFnResult, TxOpError, TxOpResult, TxResult,
+	TxValueIter, Value, ValueIter,
+};
+
+pub use sled;
+
+// -- err
+
+impl From<sled::Error> for Error {
+	fn from(e: sled::Error) -> Error {
+		Error(format!("Sled: {}", e).into())
+	}
+}
+
+impl From<sled::Error> for TxOpError {
+	fn from(e: sled::Error) -> TxOpError {
+		TxOpError(e.into())
+	}
+}
+
+// -- db
+
+pub struct SledDb {
+	db: sled::Db,
+	trees: RwLock<(Vec<sled::Tree>, HashMap<String, usize>)>,
+}
+
+impl SledDb {
+	pub fn init(db: sled::Db) -> Db {
+		let s = Self {
+			db,
+			trees: RwLock::new((Vec::new(), HashMap::new())),
+		};
+		Db(Arc::new(s))
+	}
+
+	fn get_tree(&self, i: usize) -> Result<sled::Tree> {
+		self.trees
+			.read()
+			.unwrap()
+			.0
+			.get(i)
+			.cloned()
+			.ok_or_else(|| Error("invalid tree id".into()))
+	}
+}
+
+impl IDb for SledDb {
+	fn engine(&self) -> String {
+		"Sled".into()
+	}
+
+	fn open_tree(&self, name: &str) -> Result<usize> {
+		let mut trees = self.trees.write().unwrap();
+		if let Some(i) = trees.1.get(name) {
+			Ok(*i)
+		} else {
+			let tree = self.db.open_tree(name)?;
+			let i = trees.0.len();
+			trees.0.push(tree);
+			trees.1.insert(name.to_string(), i);
+			Ok(i)
+		}
+	}
+
+	fn list_trees(&self) -> Result<Vec<String>> {
+		let mut trees = vec![];
+		for name in self.db.tree_names() {
+			let name = std::str::from_utf8(&name)
+				.map_err(|e| Error(format!("{}", e).into()))?
+				.to_string();
+			if name != "__sled__default" {
+				trees.push(name);
+			}
+		}
+		Ok(trees)
+	}
+
+	// ----
+
+	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let val = tree.get(key)?;
+		Ok(val.map(|x| x.to_vec()))
+	}
+
+	fn len(&self, tree: usize) -> Result<usize> {
+		let tree = self.get_tree(tree)?;
+		Ok(tree.len())
+	}
+
+	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = tree.insert(key, value)?;
+		Ok(old_val.map(|x| x.to_vec()))
+	}
+
+	fn remove(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = tree.remove(key)?;
+		Ok(old_val.map(|x| x.to_vec()))
+	}
+
+	fn clear(&self, tree: usize) -> Result<()> {
+		let tree = self.get_tree(tree)?;
+		tree.clear()?;
+		Ok(())
+	}
+
+	fn iter(&self, tree: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		Ok(Box::new(tree.iter().map(|v| {
+			v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into)
+		})))
+	}
+
+	fn iter_rev(&self, tree: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		Ok(Box::new(tree.iter().rev().map(|v| {
+			v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into)
+		})))
+	}
+
+	fn range<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		Ok(Box::new(tree.range::<&'r [u8], _>((low, high)).map(|v| {
+			v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into)
+		})))
+	}
+	fn range_rev<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree)?;
+		Ok(Box::new(tree.range::<&'r [u8], _>((low, high)).rev().map(
+			|v| v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into),
+		)))
+	}
+
+	// ----
+
+	fn transaction(&self, f: &dyn ITxFn) -> TxResult<(), ()> {
+		let trees = self.trees.read().unwrap();
+		let res = trees.0.transaction(|txtrees| {
+			let mut tx = SledTx {
+				trees: txtrees,
+				err: Cell::new(None),
+			};
+			match f.try_on(&mut tx) {
+				TxFnResult::Ok => {
+					assert!(tx.err.into_inner().is_none());
+					Ok(())
+				}
+				TxFnResult::Abort => {
+					assert!(tx.err.into_inner().is_none());
+					Err(ConflictableTransactionError::Abort(()))
+				}
+				TxFnResult::DbErr => {
+					let e = tx.err.into_inner().expect("No DB error");
+					Err(e.into())
+				}
+			}
+		});
+		match res {
+			Ok(()) => Ok(()),
+			Err(TransactionError::Abort(())) => Err(TxError::Abort(())),
+			Err(TransactionError::Storage(s)) => Err(TxError::Db(s.into())),
+		}
+	}
+}
+
+// ----
+
+struct SledTx<'a> {
+	trees: &'a [TransactionalTree],
+	err: Cell<Option<UnabortableTransactionError>>,
+}
+
+impl<'a> SledTx<'a> {
+	fn get_tree(&self, i: usize) -> TxOpResult<&TransactionalTree> {
+		self.trees.get(i).ok_or_else(|| {
+			TxOpError(Error(
+				"invalid tree id (it might have been openned after the transaction started)".into(),
+			))
+		})
+	}
+
+	fn save_error<R>(
+		&self,
+		v: std::result::Result<R, UnabortableTransactionError>,
+	) -> TxOpResult<R> {
+		match v {
+			Ok(x) => Ok(x),
+			Err(e) => {
+				let txt = format!("{}", e);
+				self.err.set(Some(e));
+				Err(TxOpError(Error(txt.into())))
+			}
+		}
+	}
+}
+
+impl<'a> ITx for SledTx<'a> {
+	fn get(&self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let tmp = self.save_error(tree.get(key))?;
+		Ok(tmp.map(|x| x.to_vec()))
+	}
+	fn len(&self, _tree: usize) -> TxOpResult<usize> {
+		unimplemented!(".len() in transaction not supported with Sled backend")
+	}
+
+	fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = self.save_error(tree.insert(key, value))?;
+		Ok(old_val.map(|x| x.to_vec()))
+	}
+	fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = self.save_error(tree.remove(key))?;
+		Ok(old_val.map(|x| x.to_vec()))
+	}
+
+	fn iter(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with Sled backend");
+	}
+	fn iter_rev(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with Sled backend");
+	}
+
+	fn range<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with Sled backend");
+	}
+	fn range_rev<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!("Iterators in transactions not supported with Sled backend");
+	}
+}

src/db/sqlite_adapter.rs

@@ -0,0 +1,508 @@
+use core::ops::Bound;
+
+use std::borrow::BorrowMut;
+use std::marker::PhantomPinned;
+use std::pin::Pin;
+use std::ptr::NonNull;
+use std::sync::{Arc, Mutex, MutexGuard};
+
+use rusqlite::{params, Connection, Rows, Statement, Transaction};
+
+use crate::{
+	Db, Error, IDb, ITx, ITxFn, Result, TxError, TxFnResult, TxOpError, TxOpResult, TxResult,
+	TxValueIter, Value, ValueIter,
+};
+
+pub use rusqlite;
+
+// --- err
+
+impl From<rusqlite::Error> for Error {
+	fn from(e: rusqlite::Error) -> Error {
+		Error(format!("Sqlite: {}", e).into())
+	}
+}
+
+impl From<rusqlite::Error> for TxOpError {
+	fn from(e: rusqlite::Error) -> TxOpError {
+		TxOpError(e.into())
+	}
+}
+
+// -- db
+
+pub struct SqliteDb(Mutex<SqliteDbInner>);
+
+struct SqliteDbInner {
+	db: Connection,
+	trees: Vec<String>,
+}
+
+impl SqliteDb {
+	pub fn init(db: rusqlite::Connection) -> Db {
+		let s = Self(Mutex::new(SqliteDbInner {
+			db,
+			trees: Vec::new(),
+		}));
+		Db(Arc::new(s))
+	}
+}
+
+impl SqliteDbInner {
+	fn get_tree(&self, i: usize) -> Result<&'_ str> {
+		self.trees
+			.get(i)
+			.map(String::as_str)
+			.ok_or_else(|| Error("invalid tree id".into()))
+	}
+
+	fn internal_get(&self, tree: &str, key: &[u8]) -> Result<Option<Value>> {
+		let mut stmt = self
+			.db
+			.prepare(&format!("SELECT v FROM {} WHERE k = ?1", tree))?;
+		let mut res_iter = stmt.query([key])?;
+		match res_iter.next()? {
+			None => Ok(None),
+			Some(v) => Ok(Some(v.get::<_, Vec<u8>>(0)?)),
+		}
+	}
+}
+
+impl IDb for SqliteDb {
+	fn engine(&self) -> String {
+		format!("sqlite3 v{} (using rusqlite crate)", rusqlite::version())
+	}
+
+	fn open_tree(&self, name: &str) -> Result<usize> {
+		let name = format!("tree_{}", name.replace(':', "_COLON_"));
+		let mut this = self.0.lock().unwrap();
+
+		if let Some(i) = this.trees.iter().position(|x| x == &name) {
+			Ok(i)
+		} else {
+			trace!("create table {}", name);
+			this.db.execute(
+				&format!(
+					"CREATE TABLE IF NOT EXISTS {} (
+						k BLOB PRIMARY KEY,
+						v BLOB
+					)",
+					name
+				),
+				[],
+			)?;
+			trace!("table created: {}, unlocking", name);
+
+			let i = this.trees.len();
+			this.trees.push(name.to_string());
+			Ok(i)
+		}
+	}
+
+	fn list_trees(&self) -> Result<Vec<String>> {
+		let mut trees = vec![];
+
+		trace!("list_trees: lock db");
+		let this = self.0.lock().unwrap();
+		trace!("list_trees: lock acquired");
+
+		let mut stmt = this.db.prepare(
+			"SELECT name FROM sqlite_schema WHERE type = 'table' AND name LIKE 'tree_%'",
+		)?;
+		let mut rows = stmt.query([])?;
+		while let Some(row) = rows.next()? {
+			let name = row.get::<_, String>(0)?;
+			let name = name.replace("_COLON_", ":");
+			let name = name.strip_prefix("tree_").unwrap().to_string();
+			trees.push(name);
+		}
+		Ok(trees)
+	}
+
+	// ----
+
+	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		trace!("get {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("get {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		this.internal_get(tree, key)
+	}
+
+	fn len(&self, tree: usize) -> Result<usize> {
+		trace!("len {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("len {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		let mut stmt = this.db.prepare(&format!("SELECT COUNT(*) FROM {}", tree))?;
+		let mut res_iter = stmt.query([])?;
+		match res_iter.next()? {
+			None => Ok(0),
+			Some(v) => Ok(v.get::<_, usize>(0)?),
+		}
+	}
+
+	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<Option<Value>> {
+		trace!("insert {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("insert {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		let old_val = this.internal_get(tree, key)?;
+
+		let sql = match &old_val {
+			Some(_) => format!("UPDATE {} SET v = ?2 WHERE k = ?1", tree),
+			None => format!("INSERT INTO {} (k, v) VALUES (?1, ?2)", tree),
+		};
+		let n = this.db.execute(&sql, params![key, value])?;
+		assert_eq!(n, 1);
+
+		Ok(old_val)
+	}
+
+	fn remove(&self, tree: usize, key: &[u8]) -> Result<Option<Value>> {
+		trace!("remove {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("remove {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		let old_val = this.internal_get(tree, key)?;
+
+		if old_val.is_some() {
+			let n = this
+				.db
+				.execute(&format!("DELETE FROM {} WHERE k = ?1", tree), params![key])?;
+			assert_eq!(n, 1);
+		}
+
+		Ok(old_val)
+	}
+
+	fn clear(&self, tree: usize) -> Result<()> {
+		trace!("clear {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("clear {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		this.db.execute(&format!("DELETE FROM {}", tree), [])?;
+		Ok(())
+	}
+
+	fn iter(&self, tree: usize) -> Result<ValueIter<'_>> {
+		trace!("iter {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("iter {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		let sql = format!("SELECT k, v FROM {} ORDER BY k ASC", tree);
+		DbValueIterator::make(this, &sql, [])
+	}
+
+	fn iter_rev(&self, tree: usize) -> Result<ValueIter<'_>> {
+		trace!("iter_rev {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("iter_rev {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+		let sql = format!("SELECT k, v FROM {} ORDER BY k DESC", tree);
+		DbValueIterator::make(this, &sql, [])
+	}
+
+	fn range<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		trace!("range {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("range {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+
+		let (bounds_sql, params) = bounds_sql(low, high);
+		let sql = format!("SELECT k, v FROM {} {} ORDER BY k ASC", tree, bounds_sql);
+
+		let params = params
+			.iter()
+			.map(|x| x as &dyn rusqlite::ToSql)
+			.collect::<Vec<_>>();
+
+		DbValueIterator::make::<&[&dyn rusqlite::ToSql]>(this, &sql, params.as_ref())
+	}
+	fn range_rev<'r>(
+		&self,
+		tree: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		trace!("range_rev {}: lock db", tree);
+		let this = self.0.lock().unwrap();
+		trace!("range_rev {}: lock acquired", tree);
+
+		let tree = this.get_tree(tree)?;
+
+		let (bounds_sql, params) = bounds_sql(low, high);
+		let sql = format!("SELECT k, v FROM {} {} ORDER BY k DESC", tree, bounds_sql);
+
+		let params = params
+			.iter()
+			.map(|x| x as &dyn rusqlite::ToSql)
+			.collect::<Vec<_>>();
+
+		DbValueIterator::make::<&[&dyn rusqlite::ToSql]>(this, &sql, params.as_ref())
+	}
+
+	// ----
+
+	fn transaction(&self, f: &dyn ITxFn) -> TxResult<(), ()> {
+		trace!("transaction: lock db");
+		let mut this = self.0.lock().unwrap();
+		trace!("transaction: lock acquired");
+
+		let this_mut_ref: &mut SqliteDbInner = this.borrow_mut();
+
+		let mut tx = SqliteTx {
+			tx: this_mut_ref
+				.db
+				.transaction()
+				.map_err(Error::from)
+				.map_err(TxError::Db)?,
+			trees: &this_mut_ref.trees,
+		};
+		let res = match f.try_on(&mut tx) {
+			TxFnResult::Ok => {
+				tx.tx.commit().map_err(Error::from).map_err(TxError::Db)?;
+				Ok(())
+			}
+			TxFnResult::Abort => {
+				tx.tx.rollback().map_err(Error::from).map_err(TxError::Db)?;
+				Err(TxError::Abort(()))
+			}
+			TxFnResult::DbErr => {
+				tx.tx.rollback().map_err(Error::from).map_err(TxError::Db)?;
+				Err(TxError::Db(Error(
+					"(this message will be discarded)".into(),
+				)))
+			}
+		};
+
+		trace!("transaction done");
+		res
+	}
+}
+
+// ----
+
+struct SqliteTx<'a> {
+	tx: Transaction<'a>,
+	trees: &'a [String],
+}
+
+impl<'a> SqliteTx<'a> {
+	fn get_tree(&self, i: usize) -> TxOpResult<&'_ str> {
+		self.trees.get(i).map(String::as_ref).ok_or_else(|| {
+			TxOpError(Error(
+				"invalid tree id (it might have been openned after the transaction started)".into(),
+			))
+		})
+	}
+
+	fn internal_get(&self, tree: &str, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let mut stmt = self
+			.tx
+			.prepare(&format!("SELECT v FROM {} WHERE k = ?1", tree))?;
+		let mut res_iter = stmt.query([key])?;
+		match res_iter.next()? {
+			None => Ok(None),
+			Some(v) => Ok(Some(v.get::<_, Vec<u8>>(0)?)),
+		}
+	}
+}
+
+impl<'a> ITx for SqliteTx<'a> {
+	fn get(&self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		self.internal_get(tree, key)
+	}
+	fn len(&self, tree: usize) -> TxOpResult<usize> {
+		let tree = self.get_tree(tree)?;
+		let mut stmt = self.tx.prepare(&format!("SELECT COUNT(*) FROM {}", tree))?;
+		let mut res_iter = stmt.query([])?;
+		match res_iter.next()? {
+			None => Ok(0),
+			Some(v) => Ok(v.get::<_, usize>(0)?),
+		}
+	}
+
+	fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = self.internal_get(tree, key)?;
+
+		let sql = match &old_val {
+			Some(_) => format!("UPDATE {} SET v = ?2 WHERE k = ?1", tree),
+			None => format!("INSERT INTO {} (k, v) VALUES (?1, ?2)", tree),
+		};
+		let n = self.tx.execute(&sql, params![key, value])?;
+		assert_eq!(n, 1);
+
+		Ok(old_val)
+	}
+	fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree)?;
+		let old_val = self.internal_get(tree, key)?;
+
+		if old_val.is_some() {
+			let n = self
+				.tx
+				.execute(&format!("DELETE FROM {} WHERE k = ?1", tree), params![key])?;
+			assert_eq!(n, 1);
+		}
+
+		Ok(old_val)
+	}
+
+	fn iter(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!();
+	}
+	fn iter_rev(&self, _tree: usize) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!();
+	}
+
+	fn range<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!();
+	}
+	fn range_rev<'r>(
+		&self,
+		_tree: usize,
+		_low: Bound<&'r [u8]>,
+		_high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		unimplemented!();
+	}
+}
+
+// ----
+
+struct DbValueIterator<'a> {
+	db: MutexGuard<'a, SqliteDbInner>,
+	stmt: Option<Statement<'a>>,
+	iter: Option<Rows<'a>>,
+	_pin: PhantomPinned,
+}
+
+impl<'a> DbValueIterator<'a> {
+	fn make<P: rusqlite::Params>(
+		db: MutexGuard<'a, SqliteDbInner>,
+		sql: &str,
+		args: P,
+	) -> Result<ValueIter<'a>> {
+		let res = DbValueIterator {
+			db,
+			stmt: None,
+			iter: None,
+			_pin: PhantomPinned,
+		};
+		let mut boxed = Box::pin(res);
+		trace!("make iterator with sql: {}", sql);
+
+		unsafe {
+			let db = NonNull::from(&boxed.db);
+			let stmt = db.as_ref().db.prepare(sql)?;
+
+			let mut_ref: Pin<&mut DbValueIterator<'a>> = Pin::as_mut(&mut boxed);
+			Pin::get_unchecked_mut(mut_ref).stmt = Some(stmt);
+
+			let mut stmt = NonNull::from(&boxed.stmt);
+			let iter = stmt.as_mut().as_mut().unwrap().query(args)?;
+
+			let mut_ref: Pin<&mut DbValueIterator<'a>> = Pin::as_mut(&mut boxed);
+			Pin::get_unchecked_mut(mut_ref).iter = Some(iter);
+		}
+
+		Ok(Box::new(DbValueIteratorPin(boxed)))
+	}
+}
+
+impl<'a> Drop for DbValueIterator<'a> {
+	fn drop(&mut self) {
+		trace!("drop iter");
+		drop(self.iter.take());
+		drop(self.stmt.take());
+	}
+}
+
+struct DbValueIteratorPin<'a>(Pin<Box<DbValueIterator<'a>>>);
+
+impl<'a> Iterator for DbValueIteratorPin<'a> {
+	type Item = Result<(Value, Value)>;
+
+	fn next(&mut self) -> Option<Self::Item> {
+		let next = unsafe {
+			let mut_ref: Pin<&mut DbValueIterator<'a>> = Pin::as_mut(&mut self.0);
+			Pin::get_unchecked_mut(mut_ref).iter.as_mut()?.next()
+		};
+		let row = match next {
+			Err(e) => return Some(Err(e.into())),
+			Ok(None) => return None,
+			Ok(Some(r)) => r,
+		};
+		let k = match row.get::<_, Vec<u8>>(0) {
+			Err(e) => return Some(Err(e.into())),
+			Ok(x) => x,
+		};
+		let v = match row.get::<_, Vec<u8>>(1) {
+			Err(e) => return Some(Err(e.into())),
+			Ok(y) => y,
+		};
+		Some(Ok((k, v)))
+	}
+}
+
+// ----
+
+fn bounds_sql<'r>(low: Bound<&'r [u8]>, high: Bound<&'r [u8]>) -> (String, Vec<Vec<u8>>) {
+	let mut sql = String::new();
+	let mut params: Vec<Vec<u8>> = vec![];
+
+	match low {
+		Bound::Included(b) => {
+			sql.push_str(" WHERE k >= ?1");
+			params.push(b.to_vec());
+		}
+		Bound::Excluded(b) => {
+			sql.push_str(" WHERE k > ?1");
+			params.push(b.to_vec());
+		}
+		Bound::Unbounded => (),
+	};
+
+	match high {
+		Bound::Included(b) => {
+			if !params.is_empty() {
+				sql.push_str(" AND k <= ?2");
+			} else {
+				sql.push_str(" WHERE k <= ?1");
+			}
+			params.push(b.to_vec());
+		}
+		Bound::Excluded(b) => {
+			if !params.is_empty() {
+				sql.push_str(" AND k < ?2");
+			} else {
+				sql.push_str(" WHERE k < ?1");
+			}
+			params.push(b.to_vec());
+		}
+		Bound::Unbounded => (),
+	}
+
+	(sql, params)
+}

src/db/test.rs

@@ -0,0 +1,106 @@
+use crate::*;
+
+use crate::lmdb_adapter::LmdbDb;
+use crate::sled_adapter::SledDb;
+use crate::sqlite_adapter::SqliteDb;
+
+fn test_suite(db: Db) {
+	let tree = db.open_tree("tree").unwrap();
+
+	let ka: &[u8] = &b"test"[..];
+	let kb: &[u8] = &b"zwello"[..];
+	let kint: &[u8] = &b"tz"[..];
+	let va: &[u8] = &b"plop"[..];
+	let vb: &[u8] = &b"plip"[..];
+	let vc: &[u8] = &b"plup"[..];
+
+	assert!(tree.insert(ka, va).unwrap().is_none());
+	assert_eq!(tree.get(ka).unwrap().unwrap(), va);
+
+	let res = db.transaction::<_, (), _>(|mut tx| {
+		assert_eq!(tx.get(&tree, ka).unwrap().unwrap(), va);
+
+		assert_eq!(tx.insert(&tree, ka, vb).unwrap().unwrap(), va);
+
+		assert_eq!(tx.get(&tree, ka).unwrap().unwrap(), vb);
+
+		tx.commit(12)
+	});
+	assert!(matches!(res, Ok(12)));
+	assert_eq!(tree.get(ka).unwrap().unwrap(), vb);
+
+	let res = db.transaction::<(), _, _>(|mut tx| {
+		assert_eq!(tx.get(&tree, ka).unwrap().unwrap(), vb);
+
+		assert_eq!(tx.insert(&tree, ka, vc).unwrap().unwrap(), vb);
+
+		assert_eq!(tx.get(&tree, ka).unwrap().unwrap(), vc);
+
+		tx.abort(42)
+	});
+	assert!(matches!(res, Err(TxError::Abort(42))));
+	assert_eq!(tree.get(ka).unwrap().unwrap(), vb);
+
+	let mut iter = tree.iter().unwrap();
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (ka, vb));
+	assert!(iter.next().is_none());
+	drop(iter);
+
+	assert!(tree.insert(kb, vc).unwrap().is_none());
+	assert_eq!(tree.get(kb).unwrap().unwrap(), vc);
+
+	let mut iter = tree.iter().unwrap();
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (ka, vb));
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (kb, vc));
+	assert!(iter.next().is_none());
+	drop(iter);
+
+	let mut iter = tree.range(kint..).unwrap();
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (kb, vc));
+	assert!(iter.next().is_none());
+	drop(iter);
+
+	let mut iter = tree.range_rev(..kint).unwrap();
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (ka, vb));
+	assert!(iter.next().is_none());
+	drop(iter);
+
+	let mut iter = tree.iter_rev().unwrap();
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (kb, vc));
+	let next = iter.next().unwrap().unwrap();
+	assert_eq!((next.0.as_ref(), next.1.as_ref()), (ka, vb));
+	assert!(iter.next().is_none());
+	drop(iter);
+}
+
+#[test]
+fn test_lmdb_db() {
+	let path = mktemp::Temp::new_dir().unwrap();
+	let db = heed::EnvOpenOptions::new()
+		.max_dbs(100)
+		.open(&path)
+		.unwrap();
+	let db = LmdbDb::init(db);
+	test_suite(db);
+	drop(path);
+}
+
+#[test]
+fn test_sled_db() {
+	let path = mktemp::Temp::new_dir().unwrap();
+	let db = SledDb::init(sled::open(path.to_path_buf()).unwrap());
+	test_suite(db);
+	drop(path);
+}
+
+#[test]
+fn test_sqlite_db() {
+	let db = SqliteDb::init(rusqlite::Connection::open_in_memory().unwrap());
+	test_suite(db);
+}

src/garage/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "garage"
-version = "0.7.0"
+version = "0.8.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -21,23 +21,25 @@ path = "tests/lib.rs"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-garage_api = { version = "0.7.0", path = "../api" }
-garage_model = { version = "0.7.0", path = "../model" }
-garage_rpc = { version = "0.7.0", path = "../rpc" }
-garage_table = { version = "0.7.0", path = "../table" }
-garage_util = { version = "0.7.0", path = "../util" }
-garage_web = { version = "0.7.0", path = "../web" }
+garage_db = { version = "0.8.0", path = "../db" }
+garage_api = { version = "0.8.0", path = "../api" }
+garage_block = { version = "0.8.0", path = "../block" }
+garage_model = { version = "0.8.0", path = "../model" }
+garage_rpc = { version = "0.8.0", path = "../rpc" }
+garage_table = { version = "0.8.0", path = "../table" }
+garage_util = { version = "0.8.0", path = "../util" }
+garage_web = { version = "0.8.0", path = "../web" }
 
 bytes = "1.0"
+bytesize = "1.1"
+timeago = "0.3"
 hex = "0.4"
 tracing = { version = "0.1.30", features = ["log-always"] }
-pretty_env_logger = "0.4"
+tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 rand = "0.8"
 async-trait = "0.1.7"
 sodiumoxide = { version = "0.2.5-0", package = "kuska-sodiumoxide" }
 
-sled = "0.34"
-
 rmp-serde = "0.15"
 serde = { version = "1.0", default-features = false, features = ["derive", "rc"] }
 serde_bytes = "0.11"
@@ -48,22 +50,20 @@ futures = "0.3"
 futures-util = "0.3"
 tokio = { version = "1.0", default-features = false, features = ["rt", "rt-multi-thread", "io-util", "net", "time", "macros", "sync", "signal", "fs"] }
 
-#netapp = { version = "0.3.2", git = "https://git.deuxfleurs.fr/lx/netapp" }
-#netapp = { version = "0.4", path = "../../../netapp" }
-netapp = "0.4"
+netapp = "0.5"
 
 opentelemetry = { version = "0.17", features = [ "rt-tokio" ] }
-opentelemetry-prometheus = "0.10"
-opentelemetry-otlp = "0.10"
-prometheus = "0.13"
+opentelemetry-prometheus = { version = "0.10", optional = true }
+opentelemetry-otlp = { version = "0.10", optional = true }
+prometheus = { version = "0.13", optional = true }
 
 [dev-dependencies]
 aws-sdk-s3 = "0.8"
 chrono = "0.4"
 http = "0.2"
-hmac = "0.10"
+hmac = "0.12"
 hyper = { version = "0.14", features = ["client", "http1", "runtime"] }
-sha2 = "0.9"
+sha2 = "0.10"
 
 static_init = "1.0"
 assert-json-diff = "2.0"
@@ -72,5 +72,26 @@ base64 = "0.13"
 
 
 [features]
-kubernetes-discovery = [ "garage_rpc/kubernetes-discovery" ]
+default = [ "bundled-libs", "metrics", "sled" ]
+
 k2v = [ "garage_util/k2v", "garage_api/k2v" ]
+
+# Database engines, Sled is still our default even though we don't like it
+sled = [ "garage_model/sled" ]
+lmdb = [ "garage_model/lmdb" ]
+sqlite = [ "garage_model/sqlite" ]
+
+# Automatic registration and discovery via Kubernetes API
+kubernetes-discovery = [ "garage_rpc/kubernetes-discovery" ]
+# Prometheus exporter (/metrics endpoint).
+metrics = [ "garage_api/metrics", "opentelemetry-prometheus", "prometheus" ]
+# Exporter for the OpenTelemetry Collector.
+telemetry-otlp = [ "opentelemetry-otlp" ]
+
+# NOTE: bundled-libs and system-libs should be treat as mutually exclusive;
+# exactly one of them should be enabled.
+
+# Use bundled libsqlite instead of linking against system-provided.
+bundled-libs = [ "garage_db/bundled-libs" ]
+# Link against system-provided libsodium and libzstd.
+system-libs = [ "garage_block/system-libs", "garage_rpc/system-libs", "sodiumoxide/use-pkg-config" ]

src/garage/admin.rs

@@ -15,6 +15,8 @@ use garage_table::*;
 
 use garage_rpc::*;
 
+use garage_block::repair::ScrubWorkerCommand;
+
 use garage_model::bucket_alias_table::*;
 use garage_model::bucket_table::*;
 use garage_model::garage::Garage;
@@ -24,24 +26,34 @@ use garage_model::migrate::Migrate;
 use garage_model::permission::*;
 
 use crate::cli::*;
-use crate::repair::Repair;
+use crate::repair::online::launch_online_repair;
 
 pub const ADMIN_RPC_PATH: &str = "garage/admin_rpc.rs/Rpc";
 
 #[derive(Debug, Serialize, Deserialize)]
+#[allow(clippy::large_enum_variant)]
 pub enum AdminRpc {
 	BucketOperation(BucketOperation),
 	KeyOperation(KeyOperation),
 	LaunchRepair(RepairOpt),
 	Migrate(MigrateOpt),
 	Stats(StatsOpt),
+	Worker(WorkerOpt),
 
 	// Replies
 	Ok(String),
 	BucketList(Vec<Bucket>),
-	BucketInfo(Bucket, HashMap<String, Key>),
+	BucketInfo {
+		bucket: Bucket,
+		relevant_keys: HashMap<String, Key>,
+		counters: HashMap<String, i64>,
+	},
 	KeyList(Vec<(String, String)>),
 	KeyInfo(Key, HashMap<Uuid, Bucket>),
+	WorkerList(
+		HashMap<usize, garage_util::background::WorkerInfo>,
+		WorkerListOpt,
+	),
 }
 
 impl Rpc for AdminRpc {
@@ -72,6 +84,7 @@ impl AdminRpcHandler {
 			BucketOperation::Allow(query) => self.handle_bucket_allow(query).await,
 			BucketOperation::Deny(query) => self.handle_bucket_deny(query).await,
 			BucketOperation::Website(query) => self.handle_bucket_website(query).await,
+			BucketOperation::SetQuotas(query) => self.handle_bucket_set_quotas(query).await,
 		}
 	}
 
@@ -87,6 +100,7 @@ impl AdminRpcHandler {
 				EnumerationOrder::Forward,
 			)
 			.await?;
+
 		Ok(AdminRpc::BucketList(buckets))
 	}
 
@@ -104,6 +118,15 @@ impl AdminRpcHandler {
 			.get_existing_bucket(bucket_id)
 			.await?;
 
+		let counters = self
+			.garage
+			.object_counter_table
+			.table
+			.get(&bucket_id, &EmptyKey)
+			.await?
+			.map(|x| x.filtered_values(&self.garage.system.ring.borrow()))
+			.unwrap_or_default();
+
 		let mut relevant_keys = HashMap::new();
 		for (k, _) in bucket
 			.state
@@ -139,7 +162,11 @@ impl AdminRpcHandler {
 			}
 		}
 
-		Ok(AdminRpc::BucketInfo(bucket, relevant_keys))
+		Ok(AdminRpc::BucketInfo {
+			bucket,
+			relevant_keys,
+			counters,
+		})
 	}
 
 	#[allow(clippy::ptr_arg)]
@@ -431,6 +458,60 @@ impl AdminRpcHandler {
 		Ok(AdminRpc::Ok(msg))
 	}
 
+	async fn handle_bucket_set_quotas(&self, query: &SetQuotasOpt) -> Result<AdminRpc, Error> {
+		let bucket_id = self
+			.garage
+			.bucket_helper()
+			.resolve_global_bucket_name(&query.bucket)
+			.await?
+			.ok_or_bad_request("Bucket not found")?;
+
+		let mut bucket = self
+			.garage
+			.bucket_helper()
+			.get_existing_bucket(bucket_id)
+			.await?;
+		let bucket_state = bucket.state.as_option_mut().unwrap();
+
+		if query.max_size.is_none() && query.max_objects.is_none() {
+			return Err(Error::BadRequest(
+				"You must specify either --max-size or --max-objects (or both) for this command to do something.".to_string(),
+			));
+		}
+
+		let mut quotas = bucket_state.quotas.get().clone();
+
+		match query.max_size.as_ref().map(String::as_ref) {
+			Some("none") => quotas.max_size = None,
+			Some(v) => {
+				let bs = v
+					.parse::<bytesize::ByteSize>()
+					.ok_or_bad_request(format!("Invalid size specified: {}", v))?;
+				quotas.max_size = Some(bs.as_u64());
+			}
+			_ => (),
+		}
+
+		match query.max_objects.as_ref().map(String::as_ref) {
+			Some("none") => quotas.max_objects = None,
+			Some(v) => {
+				let mo = v
+					.parse::<u64>()
+					.ok_or_bad_request(format!("Invalid number specified: {}", v))?;
+				quotas.max_objects = Some(mo);
+			}
+			_ => (),
+		}
+
+		bucket_state.quotas.update(quotas);
+		self.garage.bucket_table.insert(&bucket).await?;
+
+		Ok(AdminRpc::Ok(format!(
+			"Quotas updated for {}",
+			&query.bucket
+		)))
+	}
+
 	async fn handle_key_cmd(&self, cmd: &KeyOperation) -> Result<AdminRpc, Error> {
 		match cmd {
 			KeyOperation::List => self.handle_list_keys().await,
@@ -602,7 +683,7 @@ impl AdminRpcHandler {
 					.endpoint
 					.call(
 						&node,
-						&AdminRpc::LaunchRepair(opt_to_send.clone()),
+						AdminRpc::LaunchRepair(opt_to_send.clone()),
 						PRIO_NORMAL,
 					)
 					.await;
@@ -619,15 +700,7 @@ impl AdminRpcHandler {
 				)))
 			}
 		} else {
-			let repair = Repair {
-				garage: self.garage.clone(),
-			};
-			self.garage
-				.system
-				.background
-				.spawn_worker("Repair worker".into(), move |must_exit| async move {
-					repair.repair_worker(opt, must_exit).await
-				});
+			launch_online_repair(self.garage.clone(), opt).await;
 			Ok(AdminRpc::Ok(format!(
 				"Repair launched on {:?}",
 				self.garage.system.id
@@ -650,7 +723,7 @@ impl AdminRpcHandler {
 				let node_id = (*node).into();
 				match self
 					.endpoint
-					.call(&node_id, &AdminRpc::Stats(opt), PRIO_NORMAL)
+					.call(&node_id, AdminRpc::Stats(opt), PRIO_NORMAL)
 					.await?
 				{
 					Ok(AdminRpc::Ok(s)) => writeln!(&mut ret, "{}", s).unwrap(),
@@ -660,18 +733,22 @@ impl AdminRpcHandler {
 			}
 			Ok(AdminRpc::Ok(ret))
 		} else {
-			Ok(AdminRpc::Ok(self.gather_stats_local(opt)))
+			Ok(AdminRpc::Ok(self.gather_stats_local(opt)?))
 		}
 	}
 
-	fn gather_stats_local(&self, opt: StatsOpt) -> String {
+	fn gather_stats_local(&self, opt: StatsOpt) -> Result<String, Error> {
 		let mut ret = String::new();
 		writeln!(
 			&mut ret,
-			"\nGarage version: {}",
-			self.garage.system.garage_version(),
+			"\nGarage version: {} [features: {}]",
+			garage_util::version::garage_version(),
+			garage_util::version::garage_features()
+				.map(|list| list.join(", "))
+				.unwrap_or_else(|| "(unknown)".into()),
 		)
 		.unwrap();
+		writeln!(&mut ret, "\nDatabase engine: {}", self.garage.db.engine()).unwrap();
 
 		// Gather ring statistics
 		let ring = self.garage.system.ring.borrow().clone();
@@ -689,59 +766,108 @@ impl AdminRpcHandler {
 			writeln!(&mut ret, "  {:?} {}", n, c).unwrap();
 		}
 
-		self.gather_table_stats(&mut ret, &self.garage.bucket_table, &opt);
-		self.gather_table_stats(&mut ret, &self.garage.key_table, &opt);
-		self.gather_table_stats(&mut ret, &self.garage.object_table, &opt);
-		self.gather_table_stats(&mut ret, &self.garage.version_table, &opt);
-		self.gather_table_stats(&mut ret, &self.garage.block_ref_table, &opt);
+		self.gather_table_stats(&mut ret, &self.garage.bucket_table, &opt)?;
+		self.gather_table_stats(&mut ret, &self.garage.key_table, &opt)?;
+		self.gather_table_stats(&mut ret, &self.garage.object_table, &opt)?;
+		self.gather_table_stats(&mut ret, &self.garage.version_table, &opt)?;
+		self.gather_table_stats(&mut ret, &self.garage.block_ref_table, &opt)?;
 
 		writeln!(&mut ret, "\nBlock manager stats:").unwrap();
 		if opt.detailed {
 			writeln!(
 				&mut ret,
 				"  number of RC entries (~= number of blocks): {}",
-				self.garage.block_manager.rc_len()
+				self.garage.block_manager.rc_len()?
 			)
 			.unwrap();
 		}
 		writeln!(
 			&mut ret,
 			"  resync queue length: {}",
-			self.garage.block_manager.resync_queue_len()
+			self.garage.block_manager.resync.queue_len()?
 		)
 		.unwrap();
 		writeln!(
 			&mut ret,
 			"  blocks with resync errors: {}",
-			self.garage.block_manager.resync_errors_len()
+			self.garage.block_manager.resync.errors_len()?
 		)
 		.unwrap();
 
-		ret
+		Ok(ret)
 	}
 
-	fn gather_table_stats<F, R>(&self, to: &mut String, t: &Arc<Table<F, R>>, opt: &StatsOpt)
+	fn gather_table_stats<F, R>(
+		&self,
+		to: &mut String,
+		t: &Arc<Table<F, R>>,
+		opt: &StatsOpt,
+	) -> Result<(), Error>
 	where
 		F: TableSchema + 'static,
 		R: TableReplication + 'static,
 	{
 		writeln!(to, "\nTable stats for {}", F::TABLE_NAME).unwrap();
 		if opt.detailed {
-			writeln!(to, "  number of items: {}", t.data.store.len()).unwrap();
+			writeln!(
+				to,
+				"  number of items: {}",
+				t.data.store.len().map_err(GarageError::from)?
+			)
+			.unwrap();
 			writeln!(
 				to,
 				"  Merkle tree size: {}",
-				t.merkle_updater.merkle_tree_len()
+				t.merkle_updater.merkle_tree_len()?
 			)
 			.unwrap();
 		}
 		writeln!(
 			to,
 			"  Merkle updater todo queue length: {}",
-			t.merkle_updater.todo_len()
+			t.merkle_updater.todo_len()?
 		)
 		.unwrap();
-		writeln!(to, "  GC todo queue length: {}", t.data.gc_todo_len()).unwrap();
+		writeln!(to, "  GC todo queue length: {}", t.data.gc_todo_len()?).unwrap();
+
+		Ok(())
+	}
+
+	// ----
+
+	async fn handle_worker_cmd(&self, opt: WorkerOpt) -> Result<AdminRpc, Error> {
+		match opt.cmd {
+			WorkerCmd::List { opt } => {
+				let workers = self.garage.background.get_worker_info();
+				Ok(AdminRpc::WorkerList(workers, opt))
+			}
+			WorkerCmd::Set { opt } => match opt {
+				WorkerSetCmd::ScrubTranquility { tranquility } => {
+					let scrub_command = ScrubWorkerCommand::SetTranquility(tranquility);
+					self.garage
+						.block_manager
+						.send_scrub_command(scrub_command)
+						.await;
+					Ok(AdminRpc::Ok("Scrub tranquility updated".into()))
+				}
+				WorkerSetCmd::ResyncNWorkers { n_workers } => {
+					self.garage
+						.block_manager
+						.resync
+						.set_n_workers(n_workers)
+						.await?;
+					Ok(AdminRpc::Ok("Number of resync workers updated".into()))
+				}
+				WorkerSetCmd::ResyncTranquility { tranquility } => {
+					self.garage
+						.block_manager
+						.resync
+						.set_tranquility(tranquility)
+						.await?;
+					Ok(AdminRpc::Ok("Resync tranquility updated".into()))
+				}
+			},
+		}
 	}
 }
 
@@ -758,6 +884,7 @@ impl EndpointHandler<AdminRpc> for AdminRpcHandler {
 			AdminRpc::Migrate(opt) => self.handle_migrate(opt.clone()).await,
 			AdminRpc::LaunchRepair(opt) => self.handle_launch_repair(opt.clone()).await,
 			AdminRpc::Stats(opt) => self.handle_stats(opt.clone()).await,
+			AdminRpc::Worker(opt) => self.handle_worker_cmd(opt.clone()).await,
 			m => Err(GarageError::unexpected_rpc_message(m).into()),
 		}
 	}

src/garage/cli/cmd.rs

@@ -1,4 +1,5 @@
 use std::collections::HashSet;
+use std::time::Duration;
 
 use garage_util::error::*;
 use garage_util::formater::format_table;
@@ -39,13 +40,14 @@ pub async fn cli_command_dispatch(
 			cmd_admin(admin_rpc_endpoint, rpc_host, AdminRpc::LaunchRepair(ro)).await
 		}
 		Command::Stats(so) => cmd_admin(admin_rpc_endpoint, rpc_host, AdminRpc::Stats(so)).await,
+		Command::Worker(wo) => cmd_admin(admin_rpc_endpoint, rpc_host, AdminRpc::Worker(wo)).await,
 		_ => unreachable!(),
 	}
 }
 
 pub async fn cmd_status(rpc_cli: &Endpoint<SystemRpc, ()>, rpc_host: NodeID) -> Result<(), Error> {
 	let status = match rpc_cli
-		.call(&rpc_host, &SystemRpc::GetKnownNodes, PRIO_NORMAL)
+		.call(&rpc_host, SystemRpc::GetKnownNodes, PRIO_NORMAL)
 		.await??
 	{
 		SystemRpc::ReturnKnownNodes(nodes) => nodes,
@@ -100,6 +102,7 @@ pub async fn cmd_status(rpc_cli: &Endpoint<SystemRpc, ()>, rpc_host: NodeID) ->
 			vec!["ID\tHostname\tAddress\tTags\tZone\tCapacity\tLast seen".to_string()];
 		for adv in status.iter().filter(|adv| !adv.is_up) {
 			if let Some(NodeRoleV(Some(cfg))) = layout.roles.get(&adv.id) {
+				let tf = timeago::Formatter::new();
 				failed_nodes.push(format!(
 					"{id:?}\t{host}\t{addr}\t[{tags}]\t{zone}\t{capacity}\t{last_seen}",
 					id = adv.id,
@@ -110,7 +113,7 @@ pub async fn cmd_status(rpc_cli: &Endpoint<SystemRpc, ()>, rpc_host: NodeID) ->
 					capacity = cfg.capacity_string(),
 					last_seen = adv
 						.last_seen_secs_ago
-						.map(|s| format!("{}s ago", s))
+						.map(|s| tf.convert(Duration::from_secs(s)))
 						.unwrap_or_else(|| "never seen".into()),
 				));
 			}
@@ -146,7 +149,7 @@ pub async fn cmd_connect(
 	args: ConnectNodeOpt,
 ) -> Result<(), Error> {
 	match rpc_cli
-		.call(&rpc_host, &SystemRpc::Connect(args.node), PRIO_NORMAL)
+		.call(&rpc_host, SystemRpc::Connect(args.node), PRIO_NORMAL)
 		.await??
 	{
 		SystemRpc::Ok => {
@@ -162,15 +165,19 @@ pub async fn cmd_admin(
 	rpc_host: NodeID,
 	args: AdminRpc,
 ) -> Result<(), HelperError> {
-	match rpc_cli.call(&rpc_host, &args, PRIO_NORMAL).await?? {
+	match rpc_cli.call(&rpc_host, args, PRIO_NORMAL).await?? {
 		AdminRpc::Ok(msg) => {
 			println!("{}", msg);
 		}
 		AdminRpc::BucketList(bl) => {
 			print_bucket_list(bl);
 		}
-		AdminRpc::BucketInfo(bucket, rk) => {
-			print_bucket_info(&bucket, &rk);
+		AdminRpc::BucketInfo {
+			bucket,
+			relevant_keys,
+			counters,
+		} => {
+			print_bucket_info(&bucket, &relevant_keys, &counters);
 		}
 		AdminRpc::KeyList(kl) => {
 			print_key_list(kl);
@@ -178,6 +185,9 @@ pub async fn cmd_admin(
 		AdminRpc::KeyInfo(key, rb) => {
 			print_key_info(&key, &rb);
 		}
+		AdminRpc::WorkerList(wi, wlo) => {
+			print_worker_info(wi, wlo);
+		}
 		r => {
 			error!("Unexpected response: {:?}", r);
 		}

src/garage/cli/layout.rs

@@ -36,7 +36,7 @@ pub async fn cmd_assign_role(
 	args: AssignRoleOpt,
 ) -> Result<(), Error> {
 	let status = match rpc_cli
-		.call(&rpc_host, &SystemRpc::GetKnownNodes, PRIO_NORMAL)
+		.call(&rpc_host, SystemRpc::GetKnownNodes, PRIO_NORMAL)
 		.await??
 	{
 		SystemRpc::ReturnKnownNodes(nodes) => nodes,
@@ -245,7 +245,7 @@ pub async fn fetch_layout(
 	rpc_host: NodeID,
 ) -> Result<ClusterLayout, Error> {
 	match rpc_cli
-		.call(&rpc_host, &SystemRpc::PullClusterLayout, PRIO_NORMAL)
+		.call(&rpc_host, SystemRpc::PullClusterLayout, PRIO_NORMAL)
 		.await??
 	{
 		SystemRpc::AdvertiseClusterLayout(t) => Ok(t),
@@ -261,7 +261,7 @@ pub async fn send_layout(
 	rpc_cli
 		.call(
 			&rpc_host,
-			&SystemRpc::AdvertiseClusterLayout(layout),
+			SystemRpc::AdvertiseClusterLayout(layout),
 			PRIO_NORMAL,
 		)
 		.await??;