add garage node dump-table command for debugging purpose

Merge pull request 'garage_db: reduce frequency of sqlite snapshot progress log (fix #1129 )' (#1146 ) from fix-1129 into main-v1
Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1146
2026-08-13 15:46:53 +00:00 · 2025-08-29 16:35:47 +02:00 · 2025-08-27 22:26:32 +00:00 · 2025-08-28 00:07:35 +02:00 · 2025-08-27 21:21:09 +00:00 · 2025-08-27 21:10:59 +00:00
73 changed files with 1994 additions and 372 deletions
@@ -28,6 +28,11 @@ steps:
    commands:
      - nix-build -j4 --attr flakePackages.tests-sqlite

+  - name: unit + func tests (fjall)
+    image: nixpkgs/nix:nixos-22.05
+    commands:
+      - nix-build -j4 --attr flakePackages.tests-fjall
+
  - name: integration tests
    image: nixpkgs/nix:nixos-22.05
    commands:
@@ -687,6 +687,12 @@ version = "1.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2d2c12f985c78475a6b8d629afd0c360260ef34cfef52efccdcfd31972f81c2e"

+[[package]]
+name = "byteview"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6236364b88b9b6d0bc181ba374cf1ab55ba3ef97a1cb6f8cddad48a273767fb5"
+
 [[package]]
 name = "cc"
 version = "1.2.16"
@@ -798,6 +804,12 @@ version = "1.0.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5b63caa9aa9397e2d9480a9b13673856c78d8ac123288526c37d7839f2a86990"

+[[package]]
+name = "compare"
+version = "0.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ea0095f6103c2a8b44acd6fd15960c801dafebf02e21940360833e0673f48ba7"
+
 [[package]]
 name = "core-foundation"
 version = "0.9.4"
@@ -867,9 +879,18 @@ dependencies = [

 [[package]]
 name = "crossbeam-channel"
-version = "0.5.14"
+version = "0.5.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "06ba6d68e24814cb8de6bb986db8222d3a027d15872cabc0d18817bc3c0e4471"
+checksum = "82b8f8f868b36967f9606790d1903570de9ceaf870a7bf9fbbd3016d636a2cb2"
+dependencies = [
+ "crossbeam-utils",
+]
+
+[[package]]
+name = "crossbeam-epoch"
+version = "0.9.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5b82ac4a3c2ca9c3460964f020e1402edd5753411d7737aa39c3714ad1b5420e"
 dependencies = [
 "crossbeam-utils",
 ]
@@ -883,6 +904,16 @@ dependencies = [
 "crossbeam-utils",
 ]

+[[package]]
+name = "crossbeam-skiplist"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df29de440c58ca2cc6e587ec3d22347551a32435fbde9d2bff64e78a9ffa151b"
+dependencies = [
+ "crossbeam-epoch",
+ "crossbeam-utils",
+]
+
 [[package]]
 name = "crossbeam-utils"
 version = "0.8.21"
@@ -954,6 +985,20 @@ dependencies = [
 "num_cpus",
 ]

+[[package]]
+name = "dashmap"
+version = "6.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5041cc499144891f3790297212f32a74fb938e5136a14943f338ef9e0ae276cf"
+dependencies = [
+ "cfg-if",
+ "crossbeam-utils",
+ "hashbrown 0.14.5",
+ "lock_api",
+ "once_cell",
+ "parking_lot_core 0.9.10",
+]
+
 [[package]]
 name = "deranged"
 version = "0.4.0"
@@ -996,6 +1041,12 @@ dependencies = [
 "syn 2.0.100",
 ]

+[[package]]
+name = "double-ended-peekable"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c0d05e1c0dbad51b52c38bda7adceef61b9efc2baf04acfe8726a8c4630a6f57"
+
 [[package]]
 name = "dyn-clone"
 version = "1.0.19"
@@ -1017,6 +1068,18 @@ dependencies = [
 "cfg-if",
 ]

+[[package]]
+name = "enum_dispatch"
+version = "0.3.13"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aa18ce2bc66555b3218614519ac839ddb759a7d6720732f979ef8d13be147ecd"
+dependencies = [
+ "once_cell",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.100",
+]
+
 [[package]]
 name = "env_logger"
 version = "0.10.2"
@@ -1084,6 +1147,23 @@ version = "0.4.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0ce7134b9999ecaf8bcd65542e436736ef32ddca1b3e06094cb6ec5755203b80"

+[[package]]
+name = "fjall"
+version = "2.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b25ad44cd4360a0448a9b5a0a6f1c7a621101cca4578706d43c9a821418aebc"
+dependencies = [
+ "byteorder",
+ "byteview",
+ "dashmap 6.1.0",
+ "log",
+ "lsm-tree",
+ "path-absolutize",
+ "std-semaphore",
+ "tempfile",
+ "xxhash-rust",
+]
+
 [[package]]
 name = "fnv"
 version = "1.0.7"
@@ -1200,7 +1280,7 @@ dependencies = [

 [[package]]
 name = "garage"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "assert-json-diff",
 "async-trait",
@@ -1241,6 +1321,7 @@ dependencies = [
 "opentelemetry-prometheus",
 "parse_duration",
 "serde",
+ "serde_bytes",
 "serde_json",
 "sha1",
 "sha2",
@@ -1250,12 +1331,13 @@ dependencies = [
 "timeago",
 "tokio",
 "tracing",
+ "tracing-journald",
 "tracing-subscriber",
 ]

 [[package]]
 name = "garage_api_admin"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "argon2",
 "async-trait",
@@ -1281,7 +1363,7 @@ dependencies = [

 [[package]]
 name = "garage_api_common"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "base64 0.21.7",
 "bytes",
@@ -1300,7 +1382,6 @@ dependencies = [
 "http-body-util",
 "hyper 1.6.0",
 "hyper-util",
- "idna 0.5.0",
 "md-5",
 "nom",
 "opentelemetry",
@@ -1316,7 +1397,7 @@ dependencies = [

 [[package]]
 name = "garage_api_k2v"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "base64 0.21.7",
 "err-derive",
@@ -1339,7 +1420,7 @@ dependencies = [

 [[package]]
 name = "garage_api_s3"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "aes-gcm",
 "async-compression",
@@ -1384,7 +1465,7 @@ dependencies = [

 [[package]]
 name = "garage_block"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "arc-swap",
 "async-compression",
@@ -1409,11 +1490,13 @@ dependencies = [

 [[package]]
 name = "garage_db"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "err-derive",
+ "fjall",
 "heed",
 "mktemp",
+ "parking_lot 0.12.3",
 "r2d2",
 "r2d2_sqlite",
 "rusqlite",
@@ -1422,7 +1505,7 @@ dependencies = [

 [[package]]
 name = "garage_model"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "async-trait",
 "base64 0.21.7",
@@ -1449,7 +1532,7 @@ dependencies = [

 [[package]]
 name = "garage_net"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "arc-swap",
 "bytes",
@@ -1474,7 +1557,7 @@ dependencies = [

 [[package]]
 name = "garage_rpc"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "arc-swap",
 "async-trait",
@@ -1506,7 +1589,7 @@ dependencies = [

 [[package]]
 name = "garage_table"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "arc-swap",
 "async-trait",
@@ -1527,7 +1610,7 @@ dependencies = [

 [[package]]
 name = "garage_util"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "arc-swap",
 "async-trait",
@@ -1559,7 +1642,7 @@ dependencies = [

 [[package]]
 name = "garage_web"
-version = "1.1.0"
+version = "1.2.0"
 dependencies = [
 "err-derive",
 "garage_api_common",
@@ -1655,6 +1738,12 @@ dependencies = [
 "syn 2.0.100",
 ]

+[[package]]
+name = "guardian"
+version = "1.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "17e2ac29387b1aa07a1e448f7bb4f35b500787971e965b02842b900afa5c8f6f"
+
 [[package]]
 name = "h2"
 version = "0.3.26"
@@ -2170,16 +2259,6 @@ version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"

-[[package]]
-name = "idna"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "634d9b1461af396cad843f47fdba5597a4f9e6ddd4bfb6ff5d85028c25cb12f6"
-dependencies = [
- "unicode-bidi",
- "unicode-normalization",
-]
-
 [[package]]
 name = "idna"
 version = "1.0.3"
@@ -2239,6 +2318,15 @@ dependencies = [
 "cfg-if",
 ]

+[[package]]
+name = "interval-heap"
+version = "0.0.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "11274e5e8e89b8607cfedc2910b6626e998779b48a019151c7604d0adcb86ac6"
+dependencies = [
+ "compare",
+]
+
 [[package]]
 name = "ipnet"
 version = "2.11.0"
@@ -2595,6 +2683,36 @@ dependencies = [
 "hashbrown 0.15.2",
 ]

+[[package]]
+name = "lsm-tree"
+version = "2.10.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ab73c02eadb3dc12c0024e5b61d6284e6d59064e67e74fbad77856caa56f62c7"
+dependencies = [
+ "byteorder",
+ "crossbeam-skiplist",
+ "double-ended-peekable",
+ "enum_dispatch",
+ "guardian",
+ "interval-heap",
+ "log",
+ "lz4_flex",
+ "path-absolutize",
+ "quick_cache",
+ "rustc-hash",
+ "self_cell",
+ "tempfile",
+ "value-log",
+ "varint-rs",
+ "xxhash-rust",
+]
+
+[[package]]
+name = "lz4_flex"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "75761162ae2b0e580d7e7c390558127e5f01b4194debd6221fd8c207fc80e3f5"
+
 [[package]]
 name = "matchers"
 version = "0.1.0"
@@ -2849,7 +2967,7 @@ checksum = "6105e89802af13fdf48c49d7646d3b533a70e536d818aae7e78ba0433d01acb8"
 dependencies = [
 "async-trait",
 "crossbeam-channel",
- "dashmap",
+ "dashmap 4.0.2",
 "fnv",
 "futures-channel",
 "futures-executor",
@@ -3010,6 +3128,24 @@ version = "1.0.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"

+[[package]]
+name = "path-absolutize"
+version = "3.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e4af381fe79fa195b4909485d99f73a80792331df0625188e707854f0b3383f5"
+dependencies = [
+ "path-dedot",
+]
+
+[[package]]
+name = "path-dedot"
+version = "3.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "07ba0ad7e047712414213ff67533e6dd477af0a4e1d14fb52343e53d30ea9397"
+dependencies = [
+ "once_cell",
+]
+
 [[package]]
 name = "pem"
 version = "3.0.5"
@@ -3305,6 +3441,16 @@ dependencies = [
 "serde",
 ]

+[[package]]
+name = "quick_cache"
+version = "0.6.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9ad6644cb07b7f3488b9f3d2fde3b4c0a7fa367cafefb39dff93a659f76eb786"
+dependencies = [
+ "equivalent",
+ "hashbrown 0.15.2",
+]
+
 [[package]]
 name = "quote"
 version = "1.0.40"
@@ -3542,6 +3688,12 @@ version = "0.1.24"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "719b953e2095829ee67db738b3bfa9fa368c94900df327b3f07fe6e794d2fe1f"

+[[package]]
+name = "rustc-hash"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "357703d41365b4b27c590e3ed91eabb1b663f07c4c084095e60cbed4362dff0d"
+
 [[package]]
 name = "rustc_version"
 version = "0.4.1"
@@ -3784,6 +3936,12 @@ dependencies = [
 "libc",
 ]

+[[package]]
+name = "self_cell"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0f7d95a54511e0c7be3f51e8867aa8cf35148d7b9445d44de2f943e2b206e749"
+
 [[package]]
 name = "semver"
 version = "1.0.26"
@@ -3997,6 +4155,12 @@ dependencies = [
 "syn 1.0.109",
 ]

+[[package]]
+name = "std-semaphore"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "33ae9eec00137a8eed469fb4148acd9fc6ac8c3f9b110f52cd34698c8b5bfa0e"
+
 [[package]]
 name = "strsim"
 version = "0.11.1"
@@ -4252,21 +4416,6 @@ dependencies = [
 "zerovec",
 ]

-[[package]]
-name = "tinyvec"
-version = "1.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "09b3661f17e86524eccd4371ab0429194e0d7c008abb45f7a7495b1719463c71"
-dependencies = [
- "tinyvec_macros",
-]
-
-[[package]]
-name = "tinyvec_macros"
-version = "0.1.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
-
 [[package]]
 name = "tokio"
 version = "1.44.1"
@@ -4540,6 +4689,17 @@ dependencies = [
 "tracing",
 ]

+[[package]]
+name = "tracing-journald"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc0b4143302cf1022dac868d521e36e8b27691f72c84b3311750d5188ebba657"
+dependencies = [
+ "libc",
+ "tracing-core",
+ "tracing-subscriber",
+]
+
 [[package]]
 name = "tracing-log"
 version = "0.2.0"
@@ -4587,27 +4747,12 @@ version = "0.1.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2896d95c02a80c6d6a5d6e953d479f5ddf2dfdb6a244441010e373ac0fb88971"

-[[package]]
-name = "unicode-bidi"
-version = "0.3.18"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5c1cb5db39152898a79168971543b1cb5020dff7fe43c8dc468b0885f5e29df5"
-
 [[package]]
 name = "unicode-ident"
 version = "1.0.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5a5f39404a5da50712a4c1eecf25e90dd62b613502b7e925fd4e4d19b5c96512"

-[[package]]
-name = "unicode-normalization"
-version = "0.1.24"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5033c97c4262335cded6d6fc3e5c18ab755e1a3dc96376350f3d8e9f009ad956"
-dependencies = [
- "tinyvec",
-]
-
 [[package]]
 name = "unicode-segmentation"
 version = "1.12.0"
@@ -4655,7 +4800,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32f8b686cadd1473f4bd0117a5d28d36b1ade384ea9b5069a1c40aefed7fda60"
 dependencies = [
 "form_urlencoded",
- "idna 1.0.3",
+ "idna",
 "percent-encoding",
 ]

@@ -4693,6 +4838,29 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ba73ea9cf16a25df0c8caa16c51acb937d5712a8429db78a3ee29d5dcacd3a65"

+[[package]]
+name = "value-log"
+version = "1.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "62fc7c4ce161f049607ecea654dca3f2d727da5371ae85e2e4f14ce2b98ed67c"
+dependencies = [
+ "byteorder",
+ "byteview",
+ "interval-heap",
+ "log",
+ "path-absolutize",
+ "rustc-hash",
+ "tempfile",
+ "varint-rs",
+ "xxhash-rust",
+]
+
+[[package]]
+name = "varint-rs"
+version = "2.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f54a172d0620933a27a4360d3db3e2ae0dd6cceae9730751a036bbf182c4b23"
+
 [[package]]
 name = "vcpkg"
 version = "0.2.15"
@@ -24,18 +24,18 @@ default-members = ["src/garage"]

 # Internal Garage crates
 format_table = { version = "0.1.1", path = "src/format-table" }
-garage_api_common = { version = "1.1.0", path = "src/api/common" }
-garage_api_admin = { version = "1.1.0", path = "src/api/admin" }
-garage_api_s3 = { version = "1.1.0", path = "src/api/s3" }
-garage_api_k2v = { version = "1.1.0", path = "src/api/k2v" }
-garage_block = { version = "1.1.0", path = "src/block" }
-garage_db = { version = "1.1.0", path = "src/db", default-features = false }
-garage_model = { version = "1.1.0", path = "src/model", default-features = false }
-garage_net = { version = "1.1.0", path = "src/net" }
-garage_rpc = { version = "1.1.0", path = "src/rpc" }
-garage_table = { version = "1.1.0", path = "src/table" }
-garage_util = { version = "1.1.0", path = "src/util" }
-garage_web = { version = "1.1.0", path = "src/web" }
+garage_api_common = { version = "1.2.0", path = "src/api/common" }
+garage_api_admin = { version = "1.2.0", path = "src/api/admin" }
+garage_api_s3 = { version = "1.2.0", path = "src/api/s3" }
+garage_api_k2v = { version = "1.2.0", path = "src/api/k2v" }
+garage_block = { version = "1.2.0", path = "src/block" }
+garage_db = { version = "1.2.0", path = "src/db", default-features = false }
+garage_model = { version = "1.2.0", path = "src/model", default-features = false }
+garage_net = { version = "1.2.0", path = "src/net" }
+garage_rpc = { version = "1.2.0", path = "src/rpc" }
+garage_table = { version = "1.2.0", path = "src/table" }
+garage_util = { version = "1.2.0", path = "src/util" }
+garage_web = { version = "1.2.0", path = "src/web" }
 k2v-client = { version = "0.0.4", path = "src/k2v-client" }

 # External crates from crates.io
@@ -58,7 +58,6 @@ git-version = "0.3.4"
 hex = "0.4"
 hexdump = "0.1"
 hmac = "0.12"
-idna = "0.5"
 itertools = "0.12"
 ipnet = "2.9.0"
 lazy_static = "1.4"
@@ -66,6 +65,7 @@ md-5 = "0.10"
 mktemp = "0.5"
 nix = { version = "0.29", default-features = false, features = ["fs"] }
 nom = "7.1"
+parking_lot = "0.12"
 parse_duration = "2.1"
 pin-project = "1.0.12"
 pnet_datalink = "0.34"
@@ -84,12 +84,14 @@ pretty_env_logger = "0.5"
 structopt = { version = "0.3", default-features = false }
 syslog-tracing = "0.3"
 tracing = "0.1"
+tracing-journald = "0.3.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }

 heed = { version = "0.11", default-features = false, features = ["lmdb"] }
 rusqlite = "0.31.0"
 r2d2 = "0.8"
 r2d2_sqlite = "0.24"
+fjall = "2.4"

 async-compression = { version = "0.4", features = ["tokio", "zstd"] }
 zstd = { version = "0.13", default-features = false }
@@ -12,7 +12,7 @@ In this section, we cover the following web applications:
 | [Mastodon](#mastodon)     | ✅       | Natively supported    |
 | [Matrix](#matrix)     | ✅       |  Tested with `synapse-s3-storage-provider`    |
 | [ejabberd](#ejabberd)     | ✅       |  `mod_s3_upload`    |
-| [Pixelfed](#pixelfed)     | ❓       |  Not yet tested    |
+| [Pixelfed](#pixelfed)     | ✅       |  Natively supported    |
 | [Pleroma](#pleroma)     | ❓       |  Not yet tested    |
 | [Lemmy](#lemmy)     | ✅        |  Supported with pict-rs    |
 | [Funkwhale](#funkwhale)     | ❓       | Not yet tested     |
@@ -191,10 +191,10 @@ garage key create peertube-key

 Keep the Key ID and the Secret key in a pad, they will be needed later.  

-We need two buckets, one for normal videos (named peertube-video) and one for webtorrent videos (named peertube-playlist).
+We need two buckets, one for normal videos (named peertube-videos) and one for webtorrent videos (named peertube-playlists).
 ```bash
 garage bucket create peertube-videos
-garage bucket create peertube-playlist
+garage bucket create peertube-playlists
 ```

 Now we allow our key to read and write on these buckets:
@@ -253,7 +253,7 @@ object_storage:
    proxify_private_files: false

  streaming_playlists:
-    bucket_name: 'peertube-playlist'
+    bucket_name: 'peertube-playlists'

    # Keep it empty for our example
    prefix: ''
@@ -8,18 +8,18 @@ have published Ansible roles.  We list them and compare them below.

 ## Comparison of Ansible roles

-| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) |
-|------------------------------------|---------------------------------------------|---------------------------------------------------------------|
-| **Runtime**                        | Systemd                                     | Docker                                                        |
-| **Target OS**                      | Any Linux                                   | Any Linux                                                     |
-| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  |
-| **Additional software**            | None                                        | Traefik                                                       |
-| **Automatic node connection**      | ❌                                          | ✅                                                            |
-| **Layout management**              | ❌                                          | ✅                                                            |
-| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    |
-| **Allow custom Garage config**     | ✅                                          | ❌                                                            |
-| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            |
-| **Multiple instances on one host** | ✅                                          | ✅                                                            |
+| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) | [eddster ansible-role-garage](#eddster-ansible-role-garage) |
+|------------------------------------|---------------------------------------------|---------------------------------------------------------------|---------------------------------|
+| **Runtime**                        | Systemd                                     | Docker                                                        | Systemd                         |
+| **Target OS**                      | Any Linux                                   | Any Linux                                                     | Any Linux                       |
+| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  | arm64, arm, 386, amd64                            |
+| **Additional software**            | None                                        | Traefik                                                       | Ngnix and Keepalived (optional) |
+| **Automatic node connection**      | ❌                                          | ✅                                                            | ✅                              |
+| **Layout management**              | ❌                                          | ✅                                                            | ✅                              |
+| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    | ✅                              |
+| **Allow custom Garage config**     | ✅                                          | ❌                                                            | ❌                              |
+| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            | ✅                              |
+| **Multiple instances on one host** | ✅                                          | ✅                                                            | ❌                              |


 ## zorun/ansible-role-garage
@@ -49,3 +49,15 @@ structured DNS names, etc).

 As a result, this role makes it easier to start with Garage on Ansible,
 but is less flexible.
+
+## eddster2309/ansible-role-garage
+
+[Source code](https://github.com/eddster2309/ansible-role-garage), [Ansible galaxy](https://galaxy.ansible.com/ui/standalone/roles/eddster2309/garage/)
+
+This role is a opinionated but customisable role using the official Garage
+static binaries and only requires Systemd. As such it should work on any
+Linux based host. It includes all the nesscary configuration to
+automatically setup a clustered Garage deployment. Most Garage
+configuration options are exposed through Ansible variables so while you
+can't provide a custom config you can get very close. It can optionally
+installed a HA nginx deployment with Keepalived.
@@ -26,6 +26,13 @@ Or deploy with custom values:
 helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
 ```

+If you want to manage the CustomRessourceDefinition used by garage for its `kubernetes_discovery` outside of the helm chart, add `garage.kubernetesSkipCrd: true` to your custom values and use the kustomization before deploying the helm chart:
+
+```bash
+kubectl apply -k ../k8s/crd
+helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
+```
+
 After deploying, cluster layout must be configured manually as described in [Creating a cluster layout](@/documentation/quick-start/_index.md#creating-a-cluster-layout). Use the following command to access garage CLI:

 ```bash
@@ -96,14 +96,14 @@ to store 2 TB of data in total.
 ## Get a Docker image

 Our docker image is currently named `dxflrs/garage` and is stored on the [Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).
-We encourage you to use a fixed tag (eg. `v1.1.0`) and not the `latest` tag.
-For this example, we will use the latest published version at the time of the writing which is `v1.1.0` but it's up to you
+We encourage you to use a fixed tag (eg. `v1.2.0`) and not the `latest` tag.
+For this example, we will use the latest published version at the time of the writing which is `v1.2.0` but it's up to you
 to check [the most recent versions on the Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).

 For example:

 ```
-sudo docker pull dxflrs/garage:v1.1.0
+sudo docker pull dxflrs/garage:v1.2.0
 ```

 ## Deploying and configuring Garage
@@ -171,7 +171,7 @@ docker run \
  -v /etc/garage.toml:/etc/garage.toml \
  -v /var/lib/garage/meta:/var/lib/garage/meta \
  -v /var/lib/garage/data:/var/lib/garage/data \
-  dxflrs/garage:v1.1.0
+  dxflrs/garage:v1.2.0
 ```

 With this command line, Garage should be started automatically at each boot.
@@ -185,7 +185,7 @@ If you want to use `docker-compose`, you may use the following `docker-compose.y
 version: "3"
 services:
  garage:
-    image: dxflrs/garage:v1.1.0
+    image: dxflrs/garage:v1.2.0
    network_mode: "host"
    restart: unless-stopped
    volumes:
@@ -28,6 +28,7 @@ StateDirectory=garage
 DynamicUser=true
 ProtectHome=true
 NoNewPrivileges=true
+LimitNOFILE=42000

 [Install]
 WantedBy=multi-user.target
@@ -132,7 +132,7 @@ docker run \
  -v /path/to/garage.toml:/etc/garage.toml \
  -v /path/to/garage/meta:/var/lib/garage/meta \
  -v /path/to/garage/data:/var/lib/garage/data \
-  dxflrs/garage:v1.1.0
+  dxflrs/garage:v1.2.0
 ```

 Under Linux, you can substitute `--network host` for `-p 3900:3900 -p 3901:3901 -p 3902:3902 -p 3903:3903`
@@ -182,11 +182,12 @@ ID                 Hostname  Address         Tag                   Zone  Capacit
 ## Creating a cluster layout

 Creating a cluster layout for a Garage deployment means informing Garage
-of the disk space available on each node of the cluster
-as well as the zone (e.g. datacenter) each machine is located in.
+of the disk space available on each node of the cluster, `-c`,
+as well as the name of the zone (e.g. datacenter), `-z`, each machine is located in.

-For our test deployment, we are using only one node. The way in which we configure
-it does not matter, you can simply write:
+For our test deployment, we are have only one node with zone named `dc1` and a
+capacity of `1G`, though the capacity is ignored for a single node deployment
+and can be changed later when adding new nodes.

 ```bash
 garage layout assign -z dc1 -c 1G <node_id>
@@ -46,6 +46,7 @@ bootstrap_peers = [
    "212fd62eeaca72c122b45a7f4fa0f55e012aa5e24ac384a72a3016413fa724ff@[fc00:F::1]:3901",
 ]

+allow_punycode = false

 [consul_discovery]
 api = "catalog"
@@ -92,29 +93,30 @@ The following gives details about each available configuration option.

 [Environment variables](#env_variables).

-Top-level configuration options:
+Top-level configuration options, in alphabetical order:
+[`allow_punycode`](#allow_punycode),
 [`allow_world_readable_secrets`](#allow_world_readable_secrets),
 [`block_ram_buffer_max`](#block_ram_buffer_max),
 [`block_size`](#block_size),
 [`bootstrap_peers`](#bootstrap_peers),
 [`compression_level`](#compression_level),
+[`consistency_mode`](#consistency_mode),
 [`data_dir`](#data_dir),
 [`data_fsync`](#data_fsync),
 [`db_engine`](#db_engine),
 [`disable_scrub`](#disable_scrub),
-[`use_local_tz`](#use_local_tz),
 [`lmdb_map_size`](#lmdb_map_size),
 [`metadata_auto_snapshot_interval`](#metadata_auto_snapshot_interval),
 [`metadata_dir`](#metadata_dir),
 [`metadata_fsync`](#metadata_fsync),
 [`metadata_snapshots_dir`](#metadata_snapshots_dir),
 [`replication_factor`](#replication_factor),
-[`consistency_mode`](#consistency_mode),
 [`rpc_bind_addr`](#rpc_bind_addr),
 [`rpc_bind_outgoing`](#rpc_bind_outgoing),
 [`rpc_public_addr`](#rpc_public_addr),
 [`rpc_public_addr_subnet`](#rpc_public_addr_subnet)
-[`rpc_secret`/`rpc_secret_file`](#rpc_secret).
+[`rpc_secret`/`rpc_secret_file`](#rpc_secret),
+[`use_local_tz`](#use_local_tz).

 The `[consul_discovery]` section:
 [`api`](#consul_api),
@@ -151,13 +153,17 @@ The `[admin]` section:

 ### Environment variables {#env_variables}

-The following configuration parameter must be specified as an environment
-variable, it does not exist in the configuration file:
+The following configuration parameters must be specified as environment variables,
+they do not exist in the configuration file:

 - `GARAGE_LOG_TO_SYSLOG` (since `v0.9.4`): set this to `1` or `true` to make the
  Garage daemon send its logs to `syslog` (using the libc `syslog` function)
  instead of printing to stderr.

+- `GARAGE_LOG_TO_JOURNALD` (since `v1.2.0`): set this to `1` or `true` to make the
+  Garage daemon send its logs to `journald` (using the native protocol of `systemd-journald`)
+  instead of printing to stderr.
+
 The following environment variables can be used to override the corresponding
 values in the configuration file:

@@ -169,7 +175,7 @@ values in the configuration file:

 ### Top-level configuration options

-#### `replication_factor` {#replication_factor}
+#### `replication_factor` (since `v1.0.0`) {#replication_factor}

 The replication factor can be any positive integer smaller or equal the node count in your cluster.
 The chosen replication factor has a big impact on the cluster's failure tolerancy and performance characteristics.
@@ -217,7 +223,7 @@ is in progress.  In theory, no data should be lost as rebalancing is a
 routine operation for Garage, although we cannot guarantee you that everything
 will go right in such an extreme scenario.

-#### `consistency_mode` {#consistency_mode}
+#### `consistency_mode` (since `v1.0.0`) {#consistency_mode}

 The consistency mode setting determines the read and write behaviour of your cluster.

@@ -327,6 +333,7 @@ Since `v0.8.0`, Garage can use alternative storage backends as follows:
 | --------- | ----------------- | ------------- |
 | [LMDB](https://www.symas.com/lmdb) (since `v0.8.0`, default since `v0.9.0`) | `"lmdb"` | `<metadata_dir>/db.lmdb/` |
 | [Sqlite](https://sqlite.org) (since `v0.8.0`) | `"sqlite"` | `<metadata_dir>/db.sqlite` |
+| [Fjall](https://github.com/fjall-rs/fjall) (**experimental support** since `v1.3.0`) | `"fjall"` | `<metadata_dir>/db.fjall/` |
 | [Sled](https://sled.rs) (old default, removed since `v1.0`) | `"sled"` | `<metadata_dir>/db/` |

 Sled was supported until Garage v0.9.x, and was removed in Garage v1.0.
@@ -363,6 +370,14 @@ LMDB works very well, but is known to have the following limitations:
  so it is not the best choice for high-performance storage clusters,
  but it should work fine in many cases.

+- Fjall: a storage engine based on LSM trees, which theoretically allow for
+  higher write throughput than other storage engines that are based on B-trees.
+  Using Fjall could potentially improve Garage's performance significantly in
+  write-heavy workloads. **Support for Fjall is experimental at this point**,
+  we have added it to Garage for evaluation purposes only. **Do not use it for
+  production-critical workloads.**
+
+
 It is possible to convert Garage's metadata directory from one format to another
 using the `garage convert-db` command, which should be used as follows:

@@ -400,6 +415,7 @@ Here is how this option impacts the different database engines:
 |----------|------------------------------------|-------------------------------|
 | Sqlite   | `PRAGMA synchronous = OFF`         | `PRAGMA synchronous = NORMAL` |
 | LMDB     | `MDB_NOMETASYNC` + `MDB_NOSYNC`    | `MDB_NOMETASYNC`              |
+| Fjall    | default options                    | not supported                 |

 Note that the Sqlite database is always ran in `WAL` mode (`PRAGMA journal_mode = WAL`).

@@ -604,7 +620,7 @@ be obtained by running `garage node id` and then included directly in the
 key will be returned by `garage node id` and you will have to add the IP
 yourself.

-### `allow_world_readable_secrets` or `GARAGE_ALLOW_WORLD_READABLE_SECRETS` (env) {#allow_world_readable_secrets}
+#### `allow_world_readable_secrets` or `GARAGE_ALLOW_WORLD_READABLE_SECRETS` (env) {#allow_world_readable_secrets}

 Garage checks the permissions of your secret files to make sure they're not
 world-readable. In some cases, the check might fail and consider your files as
@@ -616,6 +632,13 @@ permission verification.
 Alternatively, you can set the `GARAGE_ALLOW_WORLD_READABLE_SECRETS`
 environment variable to `true` to bypass the permissions check.

+#### `allow_punycode` {#allow_punycode}
+
+Allow creating buckets with names containing punycode. When used for buckets served
+as websites, this allows using almost any unicode character in the domain name.
+
+Default to `false`.
+
 ### The `[consul_discovery]` section

 Garage supports discovering other nodes of the cluster using Consul.  For this
@@ -23,7 +23,6 @@ Feel free to open a PR to suggest fixes this table. Minio is missing because the
 - 2022-05-25 - Many Ceph S3 endpoints are not documented but implemented. Following a notification from the Ceph community, we added them.


-
 ## High-level features

 | Feature                      | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
@@ -34,6 +33,7 @@ Feel free to open a PR to suggest fixes this table. Minio is missing because the
 | [URL vhost-style](https://docs.aws.amazon.com/AmazonS3/latest/userguide/VirtualHosting.html#virtual-hosted-style-access) URL (eg. `bucket.host.tld/key`) |  ✅ Implemented | ❌| ✅| ✅ | ✅ |
 | [Presigned URLs](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ShareObjectPreSignedURL.html) |  ✅ Implemented | ❌|  ✅ | ✅ |  ✅(❓) |
 | [SSE-C encryption](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ServerSideEncryptionCustomerKeys.html) |  ✅ Implemented | ❓ |  ✅ | ❌ |  ✅ |
+| [Bucket versioning](https://docs.aws.amazon.com/AmazonS3/latest/userguide/Versioning.html) | ❌ Missing | ✅ |  ✅ | ❌ | ✅ |

 *Note:* OpenIO does not says if it supports presigned URLs. Because it is part
 of signature v4 and they claim they support it without additional precisions,
@@ -70,7 +70,7 @@ Example response body:
 ```json
 {
  "node": "b10c110e4e854e5aa3f4637681befac755154b20059ec163254ddbfae86b09df",
-  "garageVersion": "v1.1.0",
+  "garageVersion": "v1.2.0",
  "garageFeatures": [
    "k2v",
    "lmdb",
@@ -53,6 +53,9 @@
          tests-sqlite = testWith {
            GARAGE_TEST_INTEGRATION_DB_ENGINE = "sqlite";
          };
+          tests-fjall = testWith {
+            GARAGE_TEST_INTEGRATION_DB_ENGINE = "fjall";
+          };
        };

        # ---- developpment shell, for making native builds only ----
@@ -68,12 +68,13 @@ let
  rootFeatures = if features != null then
    features
  else
-    ([ "bundled-libs" "lmdb" "sqlite" "k2v" ] ++ (lib.optionals release [
+    ([ "bundled-libs" "lmdb" "sqlite" "fjall" "k2v" ] ++ (lib.optionals release [
      "consul-discovery"
      "kubernetes-discovery"
      "metrics"
      "telemetry-otlp"
      "syslog"
+      "journald"
    ]));

  featuresStr = lib.concatStringsSep "," rootFeatures;
@@ -2,8 +2,8 @@ apiVersion: v2
 name: garage
 description: S3-compatible object store for small self-hosted geo-distributed deployments
 type: application
-version: 0.7.0
-appVersion: "v1.1.0"
+version: 0.7.1
+appVersion: "v1.2.0"
 home: https://garagehq.deuxfleurs.fr/
 icon: https://garagehq.deuxfleurs.fr/images/garage-logo.svg

@@ -15,4 +15,4 @@ keywords:
 sources:
 - https://git.deuxfleurs.fr/Deuxfleurs/garage.git

-maintainers: []
+maintainers: []
@@ -1,6 +1,6 @@
 # garage

-![Version: 0.7.0](https://img.shields.io/badge/Version-0.7.0-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: v1.1.0](https://img.shields.io/badge/AppVersion-v1.1.0-informational?style=flat-square)
+![Version: 0.7.1](https://img.shields.io/badge/Version-0.7.1-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: v1.2.0](https://img.shields.io/badge/AppVersion-v1.2.0-informational?style=flat-square)

 S3-compatible object store for small self-hosted geo-distributed deployments

@@ -0,0 +1,43 @@
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  name: garagenodes.deuxfleurs.fr
+spec:
+  conversion:
+    strategy: None
+  group: deuxfleurs.fr
+  names:
+    kind: GarageNode
+    listKind: GarageNodeList
+    plural: garagenodes
+    singular: garagenode
+  scope: Namespaced
+  versions:
+  - name: v1
+    schema:
+      openAPIV3Schema:
+        description: Auto-generated derived type for Node via `CustomResource`
+        properties:
+          spec:
+            properties:
+              address:
+                format: ip
+                type: string
+              hostname:
+                type: string
+              port:
+                format: uint16
+                minimum: 0
+                type: integer
+            required:
+            - address
+            - hostname
+            - port
+            type: object
+        required:
+        - spec
+        title: GarageNode
+        type: object
+    served: true
+    storage: true
+    subresources: {}
@@ -0,0 +1,5 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+- garagenodes.deuxfleurs.fr.yaml
@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_admin"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -277,7 +277,7 @@ pub async fn handle_create_bucket(
 	let helper = garage.locked_helper().await;

 	if let Some(ga) = &req.global_alias {
-		if !is_valid_bucket_name(ga) {
+		if !is_valid_bucket_name(ga, garage.config.allow_punycode) {
 			return Err(Error::bad_request(format!(
 				"{}: {}",
 				ga, INVALID_BUCKET_NAME_MESSAGE
@@ -292,7 +292,7 @@ pub async fn handle_create_bucket(
 	}

 	if let Some(la) = &req.local_alias {
-		if !is_valid_bucket_name(&la.alias) {
+		if !is_valid_bucket_name(&la.alias, garage.config.allow_punycode) {
 			return Err(Error::bad_request(format!(
 				"{}: {}",
 				la.alias, INVALID_BUCKET_NAME_MESSAGE
@@ -382,7 +382,7 @@ pub async fn handle_delete_bucket(
 	for ((key_id, alias), _, active) in state.local_aliases.items().iter() {
 		if *active {
 			helper
-				.unset_local_bucket_alias(bucket.id, key_id, alias)
+				.purge_local_bucket_alias(bucket.id, key_id, alias)
 				.await?;
 		}
 	}
@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_common"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -28,7 +28,6 @@ err-derive.workspace = true
 hex.workspace = true
 hmac.workspace = true
 md-5.workspace = true
-idna.workspace = true
 tracing.workspace = true
 nom.workspace = true
 pin-project.workspace = true
@@ -33,6 +33,7 @@ use garage_util::metrics::{gen_trace_id, RecordDuration};
 use garage_util::socket_address::UnixOrTCPSocketAddress;

 use crate::helpers::{BoxBody, ErrorBody};
+use crate::signature::payload::Authorization;

 pub trait ApiEndpoint: Send + Sync + 'static {
 	fn name(&self) -> &'static str;
@@ -58,6 +59,12 @@ pub trait ApiHandler: Send + Sync + 'static {
 		req: Request<IncomingBody>,
 		endpoint: Self::Endpoint,
 	) -> impl Future<Output = Result<Response<BoxBody<Self::Error>>, Self::Error>> + Send;
+
+	/// Returns the key id used to authenticate this request. The ID returned must be safe to
+	/// log.
+	fn key_id_from_request(&self, req: &Request<IncomingBody>) -> Option<String> {
+		None
+	}
 }

 pub struct ApiServer<A: ApiHandler> {
@@ -142,19 +149,20 @@ impl<A: ApiHandler> ApiServer<A> {
 	) -> Result<Response<BoxBody<A::Error>>, http::Error> {
 		let uri = req.uri().clone();

-		if let Ok(forwarded_for_ip_addr) =
+		let source = if let Ok(forwarded_for_ip_addr) =
 			forwarded_headers::handle_forwarded_for_headers(req.headers())
 		{
-			info!(
-				"{} (via {}) {} {}",
-				forwarded_for_ip_addr,
-				addr,
-				req.method(),
-				uri
-			);
+			format!("{forwarded_for_ip_addr} (via {addr})")
 		} else {
-			info!("{} {} {}", addr, req.method(), uri);
-		}
+			format!("{addr}")
+		};
+		// we only do this to log the access key, so we can discard any error
+		let key = self
+			.api_handler
+			.key_id_from_request(&req)
+			.map(|k| format!("(key {k}) "))
+			.unwrap_or_default();
+		info!("{source} {key}{} {uri}", req.method());
 		debug!("{:?}", req);

 		let tracer = opentelemetry::global::tracer("garage");
@@ -343,7 +351,11 @@ where

 	while !*must_exit.borrow() {
 		let (stream, client_addr) = tokio::select! {
-			acc = listener.accept() => acc?,
+			acc = listener.accept() => match acc {
+							Ok(r) => r,
+							Err(e) if e.kind() == std::io::ErrorKind::ConnectionAborted => continue,
+							Err(e) => return Err(e.into()),
+						},
 			_ = must_exit.changed() => continue,
 		};

@@ -8,7 +8,6 @@ use hyper::{
 	body::{Body, Bytes},
 	Request, Response,
 };
-use idna::domain_to_unicode;
 use serde::{Deserialize, Serialize};

 use garage_model::bucket_table::BucketParams;
@@ -97,7 +96,7 @@ pub fn authority_to_host(authority: &str) -> Result<String, Error> {
 			authority
 		))),
 	};
-	authority.map(|h| domain_to_unicode(h).0)
+	authority.map(|h| h.to_ascii_lowercase())
 }

 /// Extract the bucket name and the key name from an HTTP path and possibly a bucket provided in
@@ -417,7 +417,7 @@ pub async fn verify_v4(
 // ============ Authorization header, or X-Amz-* query params =========

 pub struct Authorization {
-	key_id: String,
+	pub key_id: String,
 	scope: String,
 	signed_headers: String,
 	signature: String,
@@ -426,7 +426,7 @@ pub struct Authorization {
 }

 impl Authorization {
-	fn parse_header(headers: &HeaderMap) -> Result<Self, Error> {
+	pub fn parse_header(headers: &HeaderMap) -> Result<Self, Error> {
 		let authorization = headers
 			.get(AUTHORIZATION)
 			.ok_or_bad_request("Missing authorization header")?
@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_k2v"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -176,6 +176,12 @@ impl ApiHandler for K2VApiServer {

 		Ok(resp_ok)
 	}
+
+	fn key_id_from_request(&self, req: &Request<IncomingBody>) -> Option<String> {
+		garage_api_common::signature::payload::Authorization::parse_header(req.headers())
+			.map(|auth| auth.key_id)
+			.ok()
+	}
 }

 impl ApiEndpoint for K2VApiEndpoint {
@@ -1,6 +1,6 @@
 [package]
 name = "garage_api_s3"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -226,6 +226,7 @@ impl ApiHandler for S3ApiServer {
 			Endpoint::DeleteBucket {} => handle_delete_bucket(ctx).await,
 			Endpoint::GetBucketLocation {} => handle_get_bucket_location(ctx),
 			Endpoint::GetBucketVersioning {} => handle_get_bucket_versioning(),
+			Endpoint::GetBucketAcl {} => handle_get_bucket_acl(ctx),
 			Endpoint::ListObjects {
 				delimiter,
 				encoding_type,
@@ -342,6 +343,12 @@ impl ApiHandler for S3ApiServer {

 		Ok(resp_ok)
 	}
+
+	fn key_id_from_request(&self, req: &Request<IncomingBody>) -> Option<String> {
+		garage_api_common::signature::payload::Authorization::parse_header(req.headers())
+			.map(|auth| auth.key_id)
+			.ok()
+	}
 }

 impl ApiEndpoint for S3ApiEndpoint {
@@ -5,7 +5,7 @@ use hyper::{Request, Response, StatusCode};
 use garage_model::bucket_alias_table::*;
 use garage_model::bucket_table::Bucket;
 use garage_model::garage::Garage;
-use garage_model::key_table::Key;
+use garage_model::key_table::{Key, KeyParams};
 use garage_model::permission::BucketKeyPerm;
 use garage_table::util::*;
 use garage_util::crdt::*;
@@ -44,6 +44,55 @@ pub fn handle_get_bucket_versioning() -> Result<Response<ResBody>, Error> {
 		.body(string_body(xml))?)
 }

+pub fn handle_get_bucket_acl(ctx: ReqCtx) -> Result<Response<ResBody>, Error> {
+	let ReqCtx {
+		bucket_id, api_key, ..
+	} = ctx;
+	let key_p = api_key.params().ok_or_internal_error(
+		"Key should not be in deleted state at this point (in handle_get_bucket_acl)",
+	)?;
+
+	let mut grants: Vec<s3_xml::Grant> = vec![];
+	let kp = api_key.bucket_permissions(&bucket_id);
+
+	if kp.allow_owner {
+		grants.push(s3_xml::Grant {
+			grantee: create_grantee(&key_p, &api_key),
+			permission: s3_xml::Value("FULL_CONTROL".to_string()),
+		});
+	} else {
+		if kp.allow_read {
+			grants.push(s3_xml::Grant {
+				grantee: create_grantee(&key_p, &api_key),
+				permission: s3_xml::Value("READ".to_string()),
+			});
+			grants.push(s3_xml::Grant {
+				grantee: create_grantee(&key_p, &api_key),
+				permission: s3_xml::Value("READ_ACP".to_string()),
+			});
+		}
+		if kp.allow_write {
+			grants.push(s3_xml::Grant {
+				grantee: create_grantee(&key_p, &api_key),
+				permission: s3_xml::Value("WRITE".to_string()),
+			});
+		}
+	}
+
+	let access_control_policy = s3_xml::AccessControlPolicy {
+		xmlns: (),
+		owner: None,
+		acl: s3_xml::AccessControlList { entries: grants },
+	};
+
+	let xml = s3_xml::to_xml_with_header(&access_control_policy)?;
+	trace!("xml: {}", xml);
+
+	Ok(Response::builder()
+		.header("Content-Type", "application/xml")
+		.body(string_body(xml))?)
+}
+
 pub async fn handle_list_buckets(
 	garage: &Garage,
 	api_key: &Key,
@@ -172,7 +221,7 @@ pub async fn handle_create_bucket(
 		}

 		// Create the bucket!
-		if !is_valid_bucket_name(&bucket_name) {
+		if !is_valid_bucket_name(&bucket_name, garage.config.allow_punycode) {
 			return Err(Error::bad_request(format!(
 				"{}: {}",
 				bucket_name, INVALID_BUCKET_NAME_MESSAGE
@@ -241,11 +290,11 @@ pub async fn handle_delete_bucket(ctx: ReqCtx) -> Result<Response<ResBody>, Erro
 		// 1. delete bucket alias
 		if is_local_alias {
 			helper
-				.unset_local_bucket_alias(*bucket_id, &api_key.key_id, bucket_name)
+				.purge_local_bucket_alias(*bucket_id, &api_key.key_id, bucket_name)
 				.await?;
 		} else {
 			helper
-				.unset_global_bucket_alias(*bucket_id, bucket_name)
+				.purge_global_bucket_alias(*bucket_id, bucket_name)
 				.await?;
 		}

@@ -311,6 +360,15 @@ fn parse_create_bucket_xml(xml_bytes: &[u8]) -> Option<Option<String>> {
 	Some(ret)
 }

+fn create_grantee(key_params: &KeyParams, api_key: &Key) -> s3_xml::Grantee {
+	s3_xml::Grantee {
+		xmlns_xsi: (),
+		typ: "CanonicalUser".to_string(),
+		display_name: Some(s3_xml::Value(key_params.name.get().to_string())),
+		id: Some(s3_xml::Value(api_key.key_id.to_string())),
+	}
+}
+
 #[cfg(test)]
 mod tests {
 	use super::*;
@@ -26,9 +26,10 @@ use garage_api_common::signature::checksum::*;
 use crate::api_server::{ReqBody, ResBody};
 use crate::encryption::EncryptionParams;
 use crate::error::*;
-use crate::get::{full_object_byte_stream, PreconditionHeaders};
+use crate::get::{check_version_not_deleted, full_object_byte_stream, PreconditionHeaders};
 use crate::multipart;
 use crate::put::{extract_metadata_headers, save_stream, ChecksumMode, SaveStreamResult};
+use crate::website::X_AMZ_WEBSITE_REDIRECT_LOCATION;
 use crate::xml::{self as s3_xml, xmlns_tag};

 pub const X_AMZ_COPY_SOURCE_IF_MATCH: HeaderName =
@@ -84,7 +85,18 @@ pub async fn handle_copy(
 			Some(v) if v == hyper::header::HeaderValue::from_static("REPLACE") => {
 				extract_metadata_headers(req.headers())?
 			}
-			_ => source_object_meta_inner.into_owned().headers,
+			_ => {
+				// The x-amz-website-redirect-location header is not copied, instead
+				// it is replaced by the value from the request (or removed if no
+				// value was specified)
+				let is_redirect =
+					|(key, _): &(String, String)| key == X_AMZ_WEBSITE_REDIRECT_LOCATION.as_str();
+				let mut headers: Vec<_> = source_object_meta_inner.headers.clone();
+				headers.retain(|h| !is_redirect(h));
+				let new_headers = extract_metadata_headers(req.headers())?;
+				headers.extend(new_headers.into_iter().filter(is_redirect));
+				headers
+			}
 		},
 		checksum: source_checksum,
 	};
@@ -225,6 +237,7 @@ async fn handle_copy_metaonly(
 				.get(&source_version.uuid, &EmptyKey)
 				.await?;
 			let source_version = source_version.ok_or(Error::NoSuchKey)?;
+			check_version_not_deleted(&source_version)?;

 			// Write an "uploading" marker in Object table
 			// This holds a reference to the object in the Version table
@@ -416,6 +429,7 @@ pub async fn handle_upload_part_copy(
 		.get(&source_object_version.uuid, &EmptyKey)
 		.await?
 		.ok_or(Error::NoSuchKey)?;
+	check_version_not_deleted(&source_version)?;

 	// We want to reuse blocks from the source version as much as possible.
 	// However, we still need to get the data from these blocks
@@ -547,6 +561,7 @@ pub async fn handle_upload_part_copy(

 	let mut current_offset = 0;
 	let mut next_block = defragmenter.next().await?;
+	let mut blocks_to_dup = dest_version.clone();

 	// TODO this could be optimized similarly to read_and_put_blocks
 	// low priority because uploadpartcopy is rarely used
@@ -576,8 +591,7 @@ pub async fn handle_upload_part_copy(
 			.unwrap()?;
 		checksummer = checksummer_updated;

-		dest_version.blocks.clear();
-		dest_version.blocks.put(
+		let (version_block_key, version_block) = (
 			VersionBlockKey {
 				part_number,
 				offset: current_offset,
@@ -589,37 +603,56 @@ pub async fn handle_upload_part_copy(
 		);
 		current_offset += data_len;

-		let block_ref = BlockRef {
-			block: final_hash,
-			version: dest_version_id,
-			deleted: false.into(),
+		let next = if let Some(final_data) = data_to_upload {
+			dest_version.blocks.clear();
+			dest_version.blocks.put(version_block_key, version_block);
+			let block_ref = BlockRef {
+				block: final_hash,
+				version: dest_version_id,
+				deleted: false.into(),
+			};
+			let (_, _, _, next) = futures::try_join!(
+				// Thing 1: if the block is not exactly a block that existed before,
+				// we need to insert that data as a new block.
+				garage.block_manager.rpc_put_block(
+					final_hash,
+					final_data,
+					dest_encryption.is_encrypted(),
+					None
+				),
+				// Thing 2: we need to insert the block in the version
+				garage.version_table.insert(&dest_version),
+				// Thing 3: we need to add a block reference
+				garage.block_ref_table.insert(&block_ref),
+				// Thing 4: we need to read the next block
+				defragmenter.next(),
+			)?;
+			next
+		} else {
+			blocks_to_dup.blocks.put(version_block_key, version_block);
+			defragmenter.next().await?
 		};
-
-		let (_, _, _, next) = futures::try_join!(
-			// Thing 1: if the block is not exactly a block that existed before,
-			// we need to insert that data as a new block.
-			async {
-				if let Some(final_data) = data_to_upload {
-					garage
-						.block_manager
-						.rpc_put_block(final_hash, final_data, dest_encryption.is_encrypted(), None)
-						.await
-				} else {
-					Ok(())
-				}
-			},
-			// Thing 2: we need to insert the block in the version
-			garage.version_table.insert(&dest_version),
-			// Thing 3: we need to add a block reference
-			garage.block_ref_table.insert(&block_ref),
-			// Thing 4: we need to read the next block
-			defragmenter.next(),
-		)?;
 		next_block = next;
 	}

 	assert_eq!(current_offset, source_range.length);

+	// Put the duplicated blocks into the version & block_refs tables
+	let block_refs_to_put = blocks_to_dup
+		.blocks
+		.items()
+		.iter()
+		.map(|b| BlockRef {
+			block: b.1.hash,
+			version: dest_version_id,
+			deleted: false.into(),
+		})
+		.collect::<Vec<_>>();
+	futures::try_join!(
+		garage.version_table.insert(&blocks_to_dup),
+		garage.block_ref_table.insert_many(&block_refs_to_put[..]),
+	)?;
+
 	let checksums = checksummer.finalize();
 	let etag = dest_encryption.etag_from_md5(&checksums.md5);
 	let checksum = checksums.extract(dest_object_checksum_algorithm);
@@ -19,12 +19,13 @@ use garage_net::stream::ByteStream;
 use garage_rpc::rpc_helper::OrderTag;
 use garage_table::EmptyKey;
 use garage_util::data::*;
-use garage_util::error::OkOrMessage;
+use garage_util::error::{Error as UtilError, OkOrMessage};

 use garage_model::garage::Garage;
 use garage_model::s3::object_table::*;
 use garage_model::s3::version_table::*;

+use garage_api_common::common_error::CommonError;
 use garage_api_common::helpers::*;
 use garage_api_common::signature::checksum::{add_checksum_response_headers, X_AMZ_CHECKSUM_MODE};

@@ -215,6 +216,7 @@ pub async fn handle_head_without_ctx(
 					.get(&object_version.uuid, &EmptyKey)
 					.await?
 					.ok_or(Error::NoSuchKey)?;
+				check_version_not_deleted(&version)?;

 				let (part_offset, part_end) =
 					calculate_part_bounds(&version, pn).ok_or(Error::InvalidPart)?;
@@ -365,6 +367,21 @@ pub async fn handle_get_without_ctx(
 	}
 }

+pub(crate) fn check_version_not_deleted(version: &Version) -> Result<(), Error> {
+	if version.deleted.get() {
+		// the version was deleted between when the object_table was consulted
+		// and now, this could mean the object was deleted, or overriden.
+		// Rather than say the key doesn't exist, return a transient error
+		// to signal the client to try again.
+		return Err(CommonError::InternalError(UtilError::Message(
+			"conflict/inconsistency between object and version state, version is deleted"
+				.to_string(),
+		))
+		.into());
+	}
+	Ok(())
+}
+
 async fn handle_get_full(
 	garage: Arc<Garage>,
 	version: &ObjectVersion,
@@ -431,6 +448,7 @@ pub fn full_object_byte_stream(
 						.ok_or_message("channel closed")?;

 					let version = version_fut.await.unwrap()?.ok_or(Error::NoSuchKey)?;
+					check_version_not_deleted(&version)?;
 					for (i, (_, vb)) in version.blocks.items().iter().enumerate().skip(1) {
 						let stream_block_i = encryption
 							.get_block(&garage, &vb.hash, Some(order_stream.order(i as u64)))
@@ -446,6 +464,14 @@ pub fn full_object_byte_stream(
 				{
 					Ok(()) => (),
 					Err(e) => {
+						// TODO i think this is a bad idea, we should log
+						// an error and stop there. If the error happens to
+						// be exactly the size of what hasn't been streamed
+						// yet, the client will see the request as a
+						// success
+						// instead truncating the output notify the client
+						// something happened with their download, so that
+						// they can retry it
 						let _ = tx.send(error_stream_item(e)).await;
 					}
 				}
@@ -497,7 +523,7 @@ async fn handle_get_range(
 				.get(&version.uuid, &EmptyKey)
 				.await?
 				.ok_or(Error::NoSuchKey)?;
-
+			check_version_not_deleted(&version)?;
 			let body =
 				body_from_blocks_range(garage, encryption, version.blocks.items(), begin, end);
 			Ok(resp_builder.body(body)?)
@@ -548,6 +574,8 @@ async fn handle_get_part(
 				.await?
 				.ok_or(Error::NoSuchKey)?;

+			check_version_not_deleted(&version)?;
+
 			let (begin, end) =
 				calculate_part_bounds(&version, part_number).ok_or(Error::InvalidPart)?;

@@ -27,7 +27,7 @@ pub async fn handle_get_lifecycle(ctx: ReqCtx) -> Result<Response<ResBody>, Erro
 			.body(string_body(xml))?)
 	} else {
 		Ok(Response::builder()
-			.status(StatusCode::NO_CONTENT)
+			.status(StatusCode::NOT_FOUND)
 			.body(empty_body())?)
 	}
 }
@@ -13,6 +13,10 @@ pub fn xmlns_tag<S: Serializer>(_v: &(), s: S) -> Result<S::Ok, S::Error> {
 	s.serialize_str("http://s3.amazonaws.com/doc/2006-03-01/")
 }

+pub fn xmlns_xsi_tag<S: Serializer>(_v: &(), s: S) -> Result<S::Ok, S::Error> {
+	s.serialize_str("http://www.w3.org/2001/XMLSchema-instance")
+}
+
 #[derive(Debug, Serialize, Deserialize, PartialEq, Eq, PartialOrd, Ord)]
 pub struct Value(#[serde(rename = "$value")] pub String);

@@ -319,6 +323,42 @@ pub struct PostObject {
 	pub etag: Value,
 }

+#[derive(Debug, Serialize, PartialEq, Eq)]
+pub struct Grantee {
+	#[serde(rename = "xmlns:xsi", serialize_with = "xmlns_xsi_tag")]
+	pub xmlns_xsi: (),
+	#[serde(rename = "xsi:type")]
+	pub typ: String,
+	#[serde(rename = "DisplayName")]
+	pub display_name: Option<Value>,
+	#[serde(rename = "ID")]
+	pub id: Option<Value>,
+}
+
+#[derive(Debug, Serialize, PartialEq, Eq)]
+pub struct Grant {
+	#[serde(rename = "Grantee")]
+	pub grantee: Grantee,
+	#[serde(rename = "Permission")]
+	pub permission: Value,
+}
+
+#[derive(Debug, Serialize, PartialEq, Eq)]
+pub struct AccessControlList {
+	#[serde(rename = "Grant")]
+	pub entries: Vec<Grant>,
+}
+
+#[derive(Debug, Serialize, PartialEq, Eq)]
+pub struct AccessControlPolicy {
+	#[serde(serialize_with = "xmlns_tag")]
+	pub xmlns: (),
+	#[serde(rename = "Owner")]
+	pub owner: Option<Owner>,
+	#[serde(rename = "AccessControlList")]
+	pub acl: AccessControlList,
+}
+
 #[cfg(test)]
 mod tests {
 	use super::*;
@@ -427,6 +467,43 @@ mod tests {
 		Ok(())
 	}

+	#[test]
+	fn get_bucket_acl_result() -> Result<(), ApiError> {
+		let grant = Grant {
+			grantee: Grantee {
+				xmlns_xsi: (),
+				typ: "CanonicalUser".to_string(),
+				display_name: Some(Value("owner_name".to_string())),
+				id: Some(Value("qsdfjklm".to_string())),
+			},
+			permission: Value("FULL_CONTROL".to_string()),
+		};
+
+		let get_bucket_acl = AccessControlPolicy {
+			xmlns: (),
+			owner: None,
+			acl: AccessControlList {
+				entries: vec![grant],
+			},
+		};
+		assert_eq!(
+			to_xml_with_header(&get_bucket_acl)?,
+			"<?xml version=\"1.0\" encoding=\"UTF-8\"?>\
+<AccessControlPolicy xmlns=\"http://s3.amazonaws.com/doc/2006-03-01/\">\
+	<AccessControlList>\
+		<Grant>\
+			<Grantee xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\" xsi:type=\"CanonicalUser\">\
+				<DisplayName>owner_name</DisplayName>\
+				<ID>qsdfjklm</ID>\
+			</Grantee>\
+			<Permission>FULL_CONTROL</Permission>\
+		</Grant>\
+	</AccessControlList>\
+</AccessControlPolicy>"
+		);
+		Ok(())
+	}
+
 	#[test]
 	fn delete_result() -> Result<(), ApiError> {
 		let delete_result = DeleteResult {
@@ -1,6 +1,6 @@
 [package]
 name = "garage_block"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -408,8 +408,8 @@ impl BlockManager {
 	}

 	/// Get number of items in the refcount table
-	pub fn rc_len(&self) -> Result<usize, Error> {
-		Ok(self.rc.rc_table.len()?)
+	pub fn rc_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.rc.rc_table.approximate_len()?)
 	}

 	/// Send command to start/stop/manager scrub worker
@@ -427,7 +427,7 @@ impl BlockManager {

 	/// List all resync errors
 	pub fn list_resync_errors(&self) -> Result<Vec<BlockResyncErrorInfo>, Error> {
-		let mut blocks = Vec::with_capacity(self.resync.errors.len()?);
+		let mut blocks = Vec::with_capacity(self.resync.errors.approximate_len()?);
 		for ent in self.resync.errors.iter()? {
 			let (hash, cnt) = ent?;
 			let cnt = ErrorCounter::decode(&cnt);
@@ -50,7 +50,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_rc_size: meter
 				.u64_value_observer("block.rc_size", move |observer| {
-					if let Ok(value) = rc_tree.len() {
+					if let Ok(value) = rc_tree.approximate_len() {
 						observer.observe(value as u64, &[])
 					}
 				})
@@ -58,7 +58,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_resync_queue_len: meter
 				.u64_value_observer("block.resync_queue_length", move |observer| {
-					if let Ok(value) = resync_queue.len() {
+					if let Ok(value) = resync_queue.approximate_len() {
 						observer.observe(value as u64, &[]);
 					}
 				})
@@ -68,7 +68,7 @@ impl BlockManagerMetrics {
 				.init(),
 			_resync_errored_blocks: meter
 				.u64_value_observer("block.resync_errored_blocks", move |observer| {
-					if let Ok(value) = resync_errors.len() {
+					if let Ok(value) = resync_errors.approximate_len() {
 						observer.observe(value as u64, &[]);
 					}
 				})
@@ -106,13 +106,13 @@ impl BlockResyncManager {
 	}

 	/// Get length of resync queue
-	pub fn queue_len(&self) -> Result<usize, Error> {
-		Ok(self.queue.len()?)
+	pub fn queue_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.queue.approximate_len()?)
 	}

 	/// Get number of blocks that have an error
-	pub fn errors_len(&self) -> Result<usize, Error> {
-		Ok(self.errors.len()?)
+	pub fn errors_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.errors.approximate_len()?)
 	}

 	/// Clear the error counter for a block and put it in queue immediately
@@ -548,9 +548,11 @@ impl Worker for ResyncWorker {
 		}

 		WorkerStatus {
-			queue_length: Some(self.manager.resync.queue_len().unwrap_or(0) as u64),
+			queue_length: Some(self.manager.resync.queue_approximate_len().unwrap_or(0) as u64),
 			tranquility: Some(tranquility),
-			persistent_errors: Some(self.manager.resync.errors_len().unwrap_or(0) as u64),
+			persistent_errors: Some(
+				self.manager.resync.errors_approximate_len().unwrap_or(0) as u64
+			),
 			..Default::default()
 		}
 	}
@@ -1,6 +1,6 @@
 [package]
 name = "garage_db"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -16,10 +16,14 @@ err-derive.workspace = true
 tracing.workspace = true

 heed = { workspace = true, optional = true }
+
 rusqlite = { workspace = true, optional = true, features = ["backup"] }
 r2d2 = { workspace = true, optional = true }
 r2d2_sqlite = { workspace = true, optional = true }

+fjall = { workspace = true, optional = true }
+parking_lot = { workspace = true, optional = true }
+
 [dev-dependencies]
 mktemp.workspace = true

@@ -27,4 +31,5 @@ mktemp.workspace = true
 default = [ "lmdb", "sqlite" ]
 bundled-libs = [ "rusqlite?/bundled" ]
 lmdb = [ "heed" ]
+fjall = [ "dep:fjall", "dep:parking_lot" ]
 sqlite = [ "rusqlite", "r2d2", "r2d2_sqlite" ]
@@ -0,0 +1,453 @@
+use core::ops::Bound;
+
+use std::path::PathBuf;
+use std::sync::Arc;
+
+use parking_lot::{MappedRwLockReadGuard, RwLock, RwLockReadGuard};
+
+use fjall::{
+	PartitionCreateOptions, PersistMode, TransactionalKeyspace, TransactionalPartitionHandle,
+	WriteTransaction,
+};
+
+use crate::{
+	open::{Engine, OpenOpt},
+	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
+	TxResult, TxValueIter, Value, ValueIter,
+};
+
+pub use fjall;
+
+// --
+
+pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
+	info!("Opening Fjall database at: {}", path.display());
+	if opt.fsync {
+		return Err(Error(
+			"metadata_fsync is not supported with the Fjall database engine".into(),
+		));
+	}
+	let mut config = fjall::Config::new(path);
+	if let Some(block_cache_size) = opt.fjall_block_cache_size {
+		config = config.cache_size(block_cache_size as u64);
+	}
+	let keyspace = config.open_transactional()?;
+	Ok(FjallDb::init(keyspace))
+}
+
+// -- err
+
+impl From<fjall::Error> for Error {
+	fn from(e: fjall::Error) -> Error {
+		Error(format!("fjall: {}", e).into())
+	}
+}
+
+impl From<fjall::LsmError> for Error {
+	fn from(e: fjall::LsmError) -> Error {
+		Error(format!("fjall lsm_tree: {}", e).into())
+	}
+}
+
+impl From<fjall::Error> for TxOpError {
+	fn from(e: fjall::Error) -> TxOpError {
+		TxOpError(e.into())
+	}
+}
+
+// -- db
+
+pub struct FjallDb {
+	keyspace: TransactionalKeyspace,
+	trees: RwLock<Vec<(String, TransactionalPartitionHandle)>>,
+}
+
+type ByteRefRangeBound<'r> = (Bound<&'r [u8]>, Bound<&'r [u8]>);
+
+impl FjallDb {
+	pub fn init(keyspace: TransactionalKeyspace) -> Db {
+		let s = Self {
+			keyspace,
+			trees: RwLock::new(Vec::new()),
+		};
+		Db(Arc::new(s))
+	}
+
+	fn get_tree(
+		&self,
+		i: usize,
+	) -> Result<MappedRwLockReadGuard<'_, TransactionalPartitionHandle>> {
+		RwLockReadGuard::try_map(self.trees.read(), |trees: &Vec<_>| {
+			trees.get(i).map(|tup| &tup.1)
+		})
+		.map_err(|_| Error("invalid tree id".into()))
+	}
+}
+
+impl IDb for FjallDb {
+	fn engine(&self) -> String {
+		"Fjall (EXPERIMENTAL!)".into()
+	}
+
+	fn open_tree(&self, name: &str) -> Result<usize> {
+		let mut trees = self.trees.write();
+		let safe_name = encode_name(name)?;
+		if let Some(i) = trees.iter().position(|(name, _)| *name == safe_name) {
+			Ok(i)
+		} else {
+			let tree = self
+				.keyspace
+				.open_partition(&safe_name, PartitionCreateOptions::default())?;
+			let i = trees.len();
+			trees.push((safe_name, tree));
+			Ok(i)
+		}
+	}
+
+	fn list_trees(&self) -> Result<Vec<String>> {
+		Ok(self
+			.keyspace
+			.list_partitions()
+			.iter()
+			.map(|n| decode_name(&n))
+			.collect::<Result<Vec<_>>>()?)
+	}
+
+	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
+		std::fs::create_dir_all(base_path)?;
+		let path = Engine::Fjall.db_path(base_path);
+
+		let source_state = self.keyspace.read_tx();
+		let copy_keyspace = fjall::Config::new(path).open()?;
+
+		for partition_name in self.keyspace.list_partitions() {
+			let source_partition = self
+				.keyspace
+				.open_partition(&partition_name, PartitionCreateOptions::default())?;
+			let copy_partition =
+				copy_keyspace.open_partition(&partition_name, PartitionCreateOptions::default())?;
+
+			for entry in source_state.iter(&source_partition) {
+				let (key, value) = entry?;
+				copy_partition.insert(key, value)?;
+			}
+		}
+
+		copy_keyspace.persist(PersistMode::SyncAll)?;
+		Ok(())
+	}
+
+	// ----
+
+	fn get(&self, tree_idx: usize, key: &[u8]) -> Result<Option<Value>> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		let val = tx.get(&tree, key)?;
+		match val {
+			None => Ok(None),
+			Some(v) => Ok(Some(v.to_vec())),
+		}
+	}
+
+	fn approximate_len(&self, tree_idx: usize) -> Result<usize> {
+		let tree = self.get_tree(tree_idx)?;
+		Ok(tree.approximate_len())
+	}
+	fn is_empty(&self, tree_idx: usize) -> Result<bool> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		Ok(tx.is_empty(&tree)?)
+	}
+
+	fn insert(&self, tree_idx: usize, key: &[u8], value: &[u8]) -> Result<()> {
+		let tree = self.get_tree(tree_idx)?;
+		let mut tx = self.keyspace.write_tx();
+		tx.insert(&tree, key, value);
+		tx.commit()?;
+		Ok(())
+	}
+
+	fn remove(&self, tree_idx: usize, key: &[u8]) -> Result<()> {
+		let tree = self.get_tree(tree_idx)?;
+		let mut tx = self.keyspace.write_tx();
+		tx.remove(&tree, key);
+		tx.commit()?;
+		Ok(())
+	}
+
+	fn clear(&self, tree_idx: usize) -> Result<()> {
+		let mut trees = self.trees.write();
+
+		if tree_idx >= trees.len() {
+			return Err(Error("invalid tree id".into()));
+		}
+		let (name, tree) = trees.remove(tree_idx);
+
+		self.keyspace.delete_partition(tree)?;
+		let tree = self
+			.keyspace
+			.open_partition(&name, PartitionCreateOptions::default())?;
+		trees.insert(tree_idx, (name, tree));
+
+		Ok(())
+	}
+
+	fn iter(&self, tree_idx: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		Ok(Box::new(tx.iter(&tree).map(iterator_remap)))
+	}
+
+	fn iter_rev(&self, tree_idx: usize) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		Ok(Box::new(tx.iter(&tree).rev().map(iterator_remap)))
+	}
+
+	fn range<'r>(
+		&self,
+		tree_idx: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		Ok(Box::new(
+			tx.range::<&'r [u8], ByteRefRangeBound>(&tree, (low, high))
+				.map(iterator_remap),
+		))
+	}
+	fn range_rev<'r>(
+		&self,
+		tree_idx: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> Result<ValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let tx = self.keyspace.read_tx();
+		Ok(Box::new(
+			tx.range::<&'r [u8], ByteRefRangeBound>(&tree, (low, high))
+				.rev()
+				.map(iterator_remap),
+		))
+	}
+
+	// ----
+
+	fn transaction(&self, f: &dyn ITxFn) -> TxResult<OnCommit, ()> {
+		let trees = self.trees.read();
+		let mut tx = FjallTx {
+			trees: &trees[..],
+			tx: self.keyspace.write_tx(),
+		};
+
+		let res = f.try_on(&mut tx);
+		match res {
+			TxFnResult::Ok(on_commit) => {
+				tx.tx.commit().map_err(Error::from).map_err(TxError::Db)?;
+				Ok(on_commit)
+			}
+			TxFnResult::Abort => {
+				tx.tx.rollback();
+				Err(TxError::Abort(()))
+			}
+			TxFnResult::DbErr => {
+				tx.tx.rollback();
+				Err(TxError::Db(Error(
+					"(this message will be discarded)".into(),
+				)))
+			}
+		}
+	}
+}
+
+// ----
+
+struct FjallTx<'a> {
+	trees: &'a [(String, TransactionalPartitionHandle)],
+	tx: WriteTransaction<'a>,
+}
+
+impl<'a> FjallTx<'a> {
+	fn get_tree(&self, i: usize) -> TxOpResult<&TransactionalPartitionHandle> {
+		self.trees.get(i).map(|tup| &tup.1).ok_or_else(|| {
+			TxOpError(Error(
+				"invalid tree id (it might have been openned after the transaction started)".into(),
+			))
+		})
+	}
+}
+
+impl<'a> ITx for FjallTx<'a> {
+	fn get(&self, tree_idx: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
+		let tree = self.get_tree(tree_idx)?;
+		match self.tx.get(tree, key)? {
+			Some(v) => Ok(Some(v.to_vec())),
+			None => Ok(None),
+		}
+	}
+	fn len(&self, tree_idx: usize) -> TxOpResult<usize> {
+		let tree = self.get_tree(tree_idx)?;
+		Ok(self.tx.len(tree)? as usize)
+	}
+
+	fn insert(&mut self, tree_idx: usize, key: &[u8], value: &[u8]) -> TxOpResult<()> {
+		let tree = self.get_tree(tree_idx)?.clone();
+		self.tx.insert(&tree, key, value);
+		Ok(())
+	}
+	fn remove(&mut self, tree_idx: usize, key: &[u8]) -> TxOpResult<()> {
+		let tree = self.get_tree(tree_idx)?.clone();
+		self.tx.remove(&tree, key);
+		Ok(())
+	}
+	fn clear(&mut self, _tree_idx: usize) -> TxOpResult<()> {
+		unimplemented!("LSM tree clearing in cross-partition transaction is not supported")
+	}
+
+	fn iter(&self, tree_idx: usize) -> TxOpResult<TxValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?.clone();
+		Ok(Box::new(self.tx.iter(&tree).map(iterator_remap_tx)))
+	}
+	fn iter_rev(&self, tree_idx: usize) -> TxOpResult<TxValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?.clone();
+		Ok(Box::new(self.tx.iter(&tree).rev().map(iterator_remap_tx)))
+	}
+
+	fn range<'r>(
+		&self,
+		tree_idx: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let low = clone_bound(low);
+		let high = clone_bound(high);
+		Ok(Box::new(
+			self.tx
+				.range::<Vec<u8>, ByteVecRangeBounds>(&tree, (low, high))
+				.map(iterator_remap_tx),
+		))
+	}
+	fn range_rev<'r>(
+		&self,
+		tree_idx: usize,
+		low: Bound<&'r [u8]>,
+		high: Bound<&'r [u8]>,
+	) -> TxOpResult<TxValueIter<'_>> {
+		let tree = self.get_tree(tree_idx)?;
+		let low = clone_bound(low);
+		let high = clone_bound(high);
+		Ok(Box::new(
+			self.tx
+				.range::<Vec<u8>, ByteVecRangeBounds>(&tree, (low, high))
+				.rev()
+				.map(iterator_remap_tx),
+		))
+	}
+}
+
+// -- maps fjall's (k, v) to ours
+
+fn iterator_remap(r: fjall::Result<(fjall::Slice, fjall::Slice)>) -> Result<(Value, Value)> {
+	r.map(|(k, v)| (k.to_vec(), v.to_vec()))
+		.map_err(|e| e.into())
+}
+
+fn iterator_remap_tx(r: fjall::Result<(fjall::Slice, fjall::Slice)>) -> TxOpResult<(Value, Value)> {
+	r.map(|(k, v)| (k.to_vec(), v.to_vec()))
+		.map_err(|e| e.into())
+}
+
+// -- utils to deal with Garage's tightness on Bound lifetimes
+
+type ByteVecBound = Bound<Vec<u8>>;
+type ByteVecRangeBounds = (ByteVecBound, ByteVecBound);
+
+fn clone_bound(bound: Bound<&[u8]>) -> ByteVecBound {
+	let value = match bound {
+		Bound::Excluded(v) | Bound::Included(v) => v.to_vec(),
+		Bound::Unbounded => vec![],
+	};
+
+	match bound {
+		Bound::Included(_) => Bound::Included(value),
+		Bound::Excluded(_) => Bound::Excluded(value),
+		Bound::Unbounded => Bound::Unbounded,
+	}
+}
+
+// -- utils to encode table names --
+
+fn encode_name(s: &str) -> Result<String> {
+	let base = 'A' as u32;
+
+	let mut ret = String::with_capacity(s.len() + 10);
+	for c in s.chars() {
+		if c.is_alphanumeric() || c == '_' || c == '-' || c == '#' {
+			ret.push(c);
+		} else if c <= u8::MAX as char {
+			ret.push('$');
+			let c_hi = c as u32 / 16;
+			let c_lo = c as u32 % 16;
+			ret.push(char::from_u32(base + c_hi).unwrap());
+			ret.push(char::from_u32(base + c_lo).unwrap());
+		} else {
+			return Err(Error(
+				format!("table name {} could not be safely encoded", s).into(),
+			));
+		}
+	}
+	Ok(ret)
+}
+
+fn decode_name(s: &str) -> Result<String> {
+	use std::convert::TryFrom;
+
+	let errfn = || Error(format!("encoded table name {} is invalid", s).into());
+	let c_map = |c: char| {
+		let c = c as u32;
+		let base = 'A' as u32;
+		if (base..base + 16).contains(&c) {
+			Some(c - base)
+		} else {
+			None
+		}
+	};
+
+	let mut ret = String::with_capacity(s.len());
+	let mut it = s.chars();
+	while let Some(c) = it.next() {
+		if c == '$' {
+			let c_hi = it.next().and_then(c_map).ok_or_else(errfn)?;
+			let c_lo = it.next().and_then(c_map).ok_or_else(errfn)?;
+			let c_dec = char::try_from(c_hi * 16 + c_lo).map_err(|_| errfn())?;
+			ret.push(c_dec);
+		} else {
+			ret.push(c);
+		}
+	}
+	Ok(ret)
+}
+
+#[cfg(test)]
+mod tests {
+	use super::*;
+
+	#[test]
+	fn test_encdec_name() {
+		for name in [
+			"testname",
+			"test_name",
+			"test name",
+			"test$name",
+			"test:name@help.me$get/this**right",
+		] {
+			let encname = encode_name(name).unwrap();
+			assert!(!encname.contains(' '));
+			assert!(!encname.contains('.'));
+			assert!(!encname.contains('*'));
+			assert_eq!(*name, decode_name(&encname).unwrap());
+		}
+	}
+}
@@ -1,6 +1,8 @@
 #[macro_use]
 extern crate tracing;

+#[cfg(feature = "fjall")]
+pub mod fjall_adapter;
 #[cfg(feature = "lmdb")]
 pub mod lmdb_adapter;
 #[cfg(feature = "sqlite")]
@@ -152,7 +154,7 @@ impl Db {
 		let tree_names = other.list_trees()?;
 		for name in tree_names {
 			let tree = self.open_tree(&name)?;
-			if tree.len()? > 0 {
+			if !tree.is_empty()? {
 				return Err(Error(format!("tree {} already contains data", name).into()));
 			}

@@ -194,8 +196,12 @@ impl Tree {
 		self.0.get(self.1, key.as_ref())
 	}
 	#[inline]
-	pub fn len(&self) -> Result<usize> {
-		self.0.len(self.1)
+	pub fn approximate_len(&self) -> Result<usize> {
+		self.0.approximate_len(self.1)
+	}
+	#[inline]
+	pub fn is_empty(&self) -> Result<bool> {
+		self.0.is_empty(self.1)
 	}

 	#[inline]
@@ -333,7 +339,8 @@ pub(crate) trait IDb: Send + Sync {
 	fn snapshot(&self, path: &PathBuf) -> Result<()>;

 	fn get(&self, tree: usize, key: &[u8]) -> Result<Option<Value>>;
-	fn len(&self, tree: usize) -> Result<usize>;
+	fn approximate_len(&self, tree: usize) -> Result<usize>;
+	fn is_empty(&self, tree: usize) -> Result<bool>;

 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()>;
 	fn remove(&self, tree: usize, key: &[u8]) -> Result<()>;
@@ -1,8 +1,8 @@
 use core::ops::Bound;
-use core::ptr::NonNull;

 use std::collections::HashMap;
 use std::convert::TryInto;
+use std::marker::PhantomPinned;
 use std::path::PathBuf;
 use std::pin::Pin;
 use std::sync::{Arc, RwLock};
@@ -11,12 +11,55 @@ use heed::types::ByteSlice;
 use heed::{BytesDecode, Env, RoTxn, RwTxn, UntypedDatabase as Database};

 use crate::{
+	open::{Engine, OpenOpt},
 	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
 	TxResult, TxValueIter, Value, ValueIter,
 };

 pub use heed;

+// ---- top-level open function
+
+pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
+	info!("Opening LMDB database at: {}", path.display());
+	if let Err(e) = std::fs::create_dir_all(&path) {
+		return Err(Error(
+			format!("Unable to create LMDB data directory: {}", e).into(),
+		));
+	}
+
+	let map_size = match opt.lmdb_map_size {
+		None => recommended_map_size(),
+		Some(v) => v - (v % 4096),
+	};
+
+	let mut env_builder = heed::EnvOpenOptions::new();
+	env_builder.max_dbs(100);
+	env_builder.map_size(map_size);
+	env_builder.max_readers(2048);
+	unsafe {
+		env_builder.flag(heed::flags::Flags::MdbNoRdAhead);
+		env_builder.flag(heed::flags::Flags::MdbNoMetaSync);
+		if !opt.fsync {
+			env_builder.flag(heed::flags::Flags::MdbNoSync);
+		}
+	}
+	match env_builder.open(&path) {
+		Err(heed::Error::Io(e)) if e.kind() == std::io::ErrorKind::OutOfMemory => {
+			return Err(Error(
+				"OutOfMemory error while trying to open LMDB database. This can happen \
+                if your operating system is not allowing you to use sufficient virtual \
+                memory address space. Please check that no limit is set (ulimit -v). \
+                You may also try to set a smaller `lmdb_map_size` configuration parameter. \
+                On 32-bit machines, you should probably switch to another database engine."
+					.into(),
+			))
+		}
+		Err(e) => Err(Error(format!("Cannot open LMDB database: {}", e).into())),
+		Ok(db) => Ok(LmdbDb::init(db)),
+	}
+}
+
 // -- err

 impl From<heed::Error> for Error {
@@ -104,10 +147,9 @@ impl IDb for LmdbDb {
 		Ok(ret2)
 	}

-	fn snapshot(&self, to: &PathBuf) -> Result<()> {
-		std::fs::create_dir_all(to)?;
-		let mut path = to.clone();
-		path.push("data.mdb");
+	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
+		std::fs::create_dir_all(base_path)?;
+		let path = Engine::Lmdb.db_path(base_path);
 		self.db
 			.copy_to_path(path, heed::CompactionOption::Enabled)?;
 		Ok(())
@@ -126,11 +168,16 @@ impl IDb for LmdbDb {
 		}
 	}

-	fn len(&self, tree: usize) -> Result<usize> {
+	fn approximate_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 is_empty(&self, tree: usize) -> Result<bool> {
+		let tree = self.get_tree(tree)?;
+		let tx = self.db.read_txn()?;
+		Ok(tree.is_empty(&tx)?)
+	}

 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
@@ -159,13 +206,15 @@ impl IDb for LmdbDb {
 	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)?))
+		// Safety: the cloture does not store its argument anywhere,
+		unsafe { 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)?))
+		// Safety: the cloture does not store its argument anywhere,
+		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.rev_iter(tx)?)) }
 	}

 	fn range<'r>(
@@ -176,7 +225,8 @@ impl IDb for LmdbDb {
 	) -> Result<ValueIter<'_>> {
 		let tree = self.get_tree(tree)?;
 		let tx = self.db.read_txn()?;
-		TxAndIterator::make(tx, |tx| Ok(tree.range(tx, &(low, high))?))
+		// Safety: the cloture does not store its argument anywhere,
+		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.range(tx, &(low, high))?)) }
 	}
 	fn range_rev<'r>(
 		&self,
@@ -186,7 +236,8 @@ impl IDb for LmdbDb {
 	) -> 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))?))
+		// Safety: the cloture does not store its argument anywhere,
+		unsafe { TxAndIterator::make(tx, |tx| Ok(tree.rev_range(tx, &(low, high))?)) }
 	}

 	// ----
@@ -316,28 +367,41 @@ where
 {
 	tx: RoTxn<'a>,
 	iter: Option<I>,
+	_pin: PhantomPinned,
 }

 impl<'a, I> TxAndIterator<'a, I>
 where
 	I: Iterator<Item = IteratorItem<'a>> + 'a,
 {
-	fn make<F>(tx: RoTxn<'a>, iterfun: F) -> Result<ValueIter<'a>>
+	fn iter(self: Pin<&mut Self>) -> &mut Option<I> {
+		// Safety: iter is not structural
+		unsafe { &mut self.get_unchecked_mut().iter }
+	}
+
+	/// Safety: iterfun must not store its argument anywhere but in its result.
+	unsafe fn make<F>(tx: RoTxn<'a>, iterfun: F) -> Result<ValueIter<'a>>
 	where
 		F: FnOnce(&'a RoTxn<'a>) -> Result<I>,
 	{
-		let res = TxAndIterator { tx, iter: None };
+		let res = TxAndIterator {
+			tx,
+			iter: None,
+			_pin: PhantomPinned,
+		};
 		let mut boxed = Box::pin(res);

-		// This unsafe allows us to bypass lifetime checks
-		let tx = unsafe { NonNull::from(&boxed.tx).as_ref() };
-		let iter = iterfun(tx)?;
+		let tx_lifetime_overextended: &'a RoTxn<'a> = {
+			let tx = &boxed.tx;
+			// Safety: Artificially extending the lifetime because
+			// this reference will only be stored and accessed from the
+			// returned ValueIter which guarantees that it is destroyed
+			// before the tx it is pointing  to.
+			unsafe { &*&raw const *tx }
+		};
+		let iter = iterfun(&tx_lifetime_overextended)?;

-		let mut_ref = Pin::as_mut(&mut boxed);
-		// This unsafe allows us to write in a field of the pinned struct
-		unsafe {
-			Pin::get_unchecked_mut(mut_ref).iter = Some(iter);
-		}
+		*boxed.as_mut().iter() = Some(iter);

 		Ok(Box::new(TxAndIteratorPin(boxed)))
 	}
@@ -348,8 +412,10 @@ where
 	I: Iterator<Item = IteratorItem<'a>> + 'a,
 {
 	fn drop(&mut self) {
-		// ensure the iterator is dropped before the RoTxn it references
-		drop(self.iter.take());
+		// Safety: `new_unchecked` is okay because we know this value is never
+		// used again after being dropped.
+		let this = unsafe { Pin::new_unchecked(self) };
+		drop(this.iter().take());
 	}
 }

@@ -365,13 +431,12 @@ where

 	fn next(&mut self) -> Option<Self::Item> {
 		let mut_ref = Pin::as_mut(&mut self.0);
-		// This unsafe allows us to mutably access the iterator field
-		let next = unsafe { Pin::get_unchecked_mut(mut_ref).iter.as_mut()?.next() };
-		match next {
-			None => None,
-			Some(Err(e)) => Some(Err(e.into())),
-			Some(Ok((k, v))) => Some(Ok((k.to_vec(), v.to_vec()))),
-		}
+		let next = mut_ref.iter().as_mut()?.next()?;
+		let res = match next {
+			Err(e) => Err(e.into()),
+			Ok((k, v)) => Ok((k.to_vec(), v.to_vec())),
+		};
+		Some(res)
 	}
 }

@@ -11,6 +11,7 @@ use crate::{Db, Error, Result};
 pub enum Engine {
 	Lmdb,
 	Sqlite,
+	Fjall,
 }

 impl Engine {
@@ -19,8 +20,26 @@ impl Engine {
 		match self {
 			Self::Lmdb => "lmdb",
 			Self::Sqlite => "sqlite",
+			Self::Fjall => "fjall",
 		}
 	}
+
+	/// Return engine-specific DB path from base path
+	pub fn db_path(&self, base_path: &PathBuf) -> PathBuf {
+		let mut ret = base_path.clone();
+		match self {
+			Self::Lmdb => {
+				ret.push("db.lmdb");
+			}
+			Self::Sqlite => {
+				ret.push("db.sqlite");
+			}
+			Self::Fjall => {
+				ret.push("db.fjall");
+			}
+		}
+		ret
+	}
 }

 impl std::fmt::Display for Engine {
@@ -36,10 +55,11 @@ impl std::str::FromStr for Engine {
 		match text {
 			"lmdb" | "heed" => Ok(Self::Lmdb),
 			"sqlite" | "sqlite3" | "rusqlite" => Ok(Self::Sqlite),
+            "fjall" => Ok(Self::Fjall),
 			"sled" => Err(Error("Sled is no longer supported as a database engine. Converting your old metadata db can be done using an older Garage binary (e.g. v0.9.4).".into())),
 			kind => Err(Error(
 				format!(
-					"Invalid DB engine: {} (options are: lmdb, sqlite)",
+					"Invalid DB engine: {} (options are: lmdb, sqlite, fjall)",
 					kind
 				)
 				.into(),
@@ -51,6 +71,7 @@ impl std::str::FromStr for Engine {
 pub struct OpenOpt {
 	pub fsync: bool,
 	pub lmdb_map_size: Option<usize>,
+	pub fjall_block_cache_size: Option<usize>,
 }

 impl Default for OpenOpt {
@@ -58,6 +79,7 @@ impl Default for OpenOpt {
 		Self {
 			fsync: false,
 			lmdb_map_size: None,
+			fjall_block_cache_size: None,
 		}
 	}
 }
@@ -66,53 +88,15 @@ pub fn open_db(path: &PathBuf, engine: Engine, opt: &OpenOpt) -> Result<Db> {
 	match engine {
 		// ---- Sqlite DB ----
 		#[cfg(feature = "sqlite")]
-		Engine::Sqlite => {
-			info!("Opening Sqlite database at: {}", path.display());
-			let manager = r2d2_sqlite::SqliteConnectionManager::file(path);
-			Ok(crate::sqlite_adapter::SqliteDb::new(manager, opt.fsync)?)
-		}
+		Engine::Sqlite => crate::sqlite_adapter::open_db(path, opt),

 		// ---- LMDB DB ----
 		#[cfg(feature = "lmdb")]
-		Engine::Lmdb => {
-			info!("Opening LMDB database at: {}", path.display());
-			if let Err(e) = std::fs::create_dir_all(&path) {
-				return Err(Error(
-					format!("Unable to create LMDB data directory: {}", e).into(),
-				));
-			}
+		Engine::Lmdb => crate::lmdb_adapter::open_db(path, opt),

-			let map_size = match opt.lmdb_map_size {
-				None => crate::lmdb_adapter::recommended_map_size(),
-				Some(v) => v - (v % 4096),
-			};
-
-			let mut env_builder = heed::EnvOpenOptions::new();
-			env_builder.max_dbs(100);
-			env_builder.map_size(map_size);
-			env_builder.max_readers(2048);
-			unsafe {
-				env_builder.flag(crate::lmdb_adapter::heed::flags::Flags::MdbNoRdAhead);
-				env_builder.flag(crate::lmdb_adapter::heed::flags::Flags::MdbNoMetaSync);
-				if !opt.fsync {
-					env_builder.flag(heed::flags::Flags::MdbNoSync);
-				}
-			}
-			match env_builder.open(&path) {
-				Err(heed::Error::Io(e)) if e.kind() == std::io::ErrorKind::OutOfMemory => {
-					return Err(Error(
-						"OutOfMemory error while trying to open LMDB database. This can happen \
-                        if your operating system is not allowing you to use sufficient virtual \
-                        memory address space. Please check that no limit is set (ulimit -v). \
-                        You may also try to set a smaller `lmdb_map_size` configuration parameter. \
-                        On 32-bit machines, you should probably switch to another database engine."
-							.into(),
-					))
-				}
-				Err(e) => Err(Error(format!("Cannot open LMDB database: {}", e).into())),
-				Ok(db) => Ok(crate::lmdb_adapter::LmdbDb::init(db)),
-			}
-		}
+		// ---- Fjall DB ----
+		#[cfg(feature = "fjall")]
+		Engine::Fjall => crate::fjall_adapter::open_db(path, opt),

 		// Pattern is unreachable when all supported DB engines are compiled into binary. The allow
 		// attribute is added so that we won't have to change this match in case stop building
@@ -11,12 +11,23 @@ use r2d2_sqlite::SqliteConnectionManager;
 use rusqlite::{params, Rows, Statement, Transaction};

 use crate::{
+	open::{Engine, OpenOpt},
 	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
 	TxResult, TxValueIter, Value, ValueIter,
 };

 pub use rusqlite;

+// ---- top-level open function
+
+pub(crate) fn open_db(path: &PathBuf, opt: &OpenOpt) -> Result<Db> {
+	info!("Opening Sqlite database at: {}", path.display());
+	let manager = r2d2_sqlite::SqliteConnectionManager::file(path);
+	Ok(SqliteDb::new(manager, opt.fsync)?)
+}
+
+// ----
+
 type Connection = r2d2::PooledConnection<SqliteConnectionManager>;

 // --- err
@@ -139,17 +150,32 @@ impl IDb for SqliteDb {
 		Ok(trees)
 	}

-	fn snapshot(&self, to: &PathBuf) -> Result<()> {
+	fn snapshot(&self, base_path: &PathBuf) -> Result<()> {
 		fn progress(p: rusqlite::backup::Progress) {
-			let percent = (p.pagecount - p.remaining) * 100 / p.pagecount;
-			info!("Sqlite snapshot progress: {}%", percent);
+			use std::sync::atomic::{AtomicU64, Ordering};
+			use std::time::{SystemTime, UNIX_EPOCH};
+
+			static LAST_LOG_TIME: AtomicU64 = AtomicU64::new(0);
+
+			let now = SystemTime::now()
+				.duration_since(UNIX_EPOCH)
+				.expect("Fix your clock :o")
+				.as_millis() as u64;
+			if now >= LAST_LOG_TIME.load(Ordering::Relaxed) + 10 * 1000 {
+				let percent = (p.pagecount - p.remaining) * 100 / p.pagecount;
+				info!("Sqlite snapshot progress: {}%", percent);
+
+				LAST_LOG_TIME.fetch_max(now, Ordering::Relaxed);
+			}
 		}
-		std::fs::create_dir_all(to)?;
-		let mut path = to.clone();
-		path.push("db.sqlite");
+
+		std::fs::create_dir_all(base_path)?;
+		let path = Engine::Sqlite.db_path(&base_path);
+
 		self.db
 			.get()?
 			.backup(rusqlite::DatabaseName::Main, path, Some(progress))?;
+
 		Ok(())
 	}

@@ -160,7 +186,7 @@ impl IDb for SqliteDb {
 		self.internal_get(&self.db.get()?, &tree, key)
 	}

-	fn len(&self, tree: usize) -> Result<usize> {
+	fn approximate_len(&self, tree: usize) -> Result<usize> {
 		let tree = self.get_tree(tree)?;
 		let db = self.db.get()?;

@@ -172,6 +198,10 @@ impl IDb for SqliteDb {
 		}
 	}

+	fn is_empty(&self, tree: usize) -> Result<bool> {
+		Ok(self.approximate_len(tree)? == 0)
+	}
+
 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
 		let db = self.db.get()?;
@@ -1,7 +1,7 @@
 use crate::*;

 fn test_suite(db: Db) {
-	let tree = db.open_tree("tree").unwrap();
+	let tree = db.open_tree("tree:this_is_a_tree").unwrap();

 	let ka: &[u8] = &b"test"[..];
 	let kb: &[u8] = &b"zwello"[..];
@@ -14,7 +14,7 @@ fn test_suite(db: Db) {

 	assert!(tree.insert(ka, va).is_ok());
 	assert_eq!(tree.get(ka).unwrap().unwrap(), va);
-	assert_eq!(tree.len().unwrap(), 1);
+	assert_eq!(tree.iter().unwrap().count(), 1);

 	// ---- test transaction logic ----

@@ -148,3 +148,15 @@ fn test_sqlite_db() {
 	let db = SqliteDb::new(manager, false).unwrap();
 	test_suite(db);
 }
+
+#[test]
+#[cfg(feature = "fjall")]
+fn test_fjall_db() {
+	use crate::fjall_adapter::{fjall, FjallDb};
+
+	let path = mktemp::Temp::new_dir().unwrap();
+	let config = fjall::Config::new(path).temporary(true);
+	let keyspace = config.open_transactional().unwrap();
+	let db = FjallDb::init(keyspace);
+	test_suite(db);
+}
@@ -1,6 +1,6 @@
 [package]
 name = "garage"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -49,6 +49,8 @@ structopt.workspace = true
 git-version.workspace = true

 serde.workspace = true
+serde_json.workspace = true
+serde_bytes.workspace = true

 futures.workspace = true
 tokio.workspace = true
@@ -57,6 +59,7 @@ opentelemetry.workspace = true
 opentelemetry-prometheus = { workspace = true, optional = true }
 opentelemetry-otlp = { workspace = true, optional = true }
 syslog-tracing = { workspace = true, optional = true }
+tracing-journald = { workspace = true, optional = true }

 [dev-dependencies]
 garage_api_common.workspace = true
@@ -90,6 +93,7 @@ k2v = [ "garage_util/k2v", "garage_api_k2v" ]
 # Database engines
 lmdb = [ "garage_model/lmdb" ]
 sqlite = [ "garage_model/sqlite" ]
+fjall = [ "garage_model/fjall" ]

 # Automatic registration and discovery via Consul API
 consul-discovery = [ "garage_rpc/consul-discovery" ]
@@ -101,6 +105,8 @@ metrics = [ "garage_api_admin/metrics", "opentelemetry-prometheus" ]
 telemetry-otlp = [ "opentelemetry-otlp" ]
 # Logging to syslog
 syslog = [ "syslog-tracing" ]
+# Logging to journald
+journald = [ "tracing-journald" ]

 # NOTE: bundled-libs and system-libs should be treat as mutually exclusive;
 # exactly one of them should be enabled.
@@ -101,6 +101,7 @@ impl AdminRpcHandler {
 		let mut obj_dels = 0;
 		let mut mpu_dels = 0;
 		let mut ver_dels = 0;
+		let mut br_dels = 0;

 		for hash in blocks {
 			let hash = hex::decode(hash).ok_or_bad_request("invalid hash")?;
@@ -131,12 +132,19 @@ impl AdminRpcHandler {
 						ver_dels += 1;
 					}
 				}
+				if !br.deleted.get() {
+					let mut br = br;
+					br.deleted.set();
+					self.garage.block_ref_table.insert(&br).await?;
+					br_dels += 1;
+				}
 			}
 		}

 		Ok(AdminRpc::Ok(format!(
-			"Purged {} blocks, {} versions, {} objects, {} multipart uploads",
+			"Purged {} blocks: marked {} block refs, {} versions, {} objects and {} multipart uploads as deleted",
 			blocks.len(),
+			br_dels,
 			ver_dels,
 			obj_dels,
 			mpu_dels,
@@ -126,7 +126,7 @@ impl AdminRpcHandler {

 	#[allow(clippy::ptr_arg)]
 	async fn handle_create_bucket(&self, name: &String) -> Result<AdminRpc, Error> {
-		if !is_valid_bucket_name(name) {
+		if !is_valid_bucket_name(name, self.garage.config.allow_punycode) {
 			return Err(Error::BadRequest(format!(
 				"{}: {}",
 				name, INVALID_BUCKET_NAME_MESSAGE
@@ -219,7 +219,7 @@ impl AdminRpcHandler {

 		// Gather block manager statistics
 		writeln!(&mut ret, "\nBlock manager stats:").unwrap();
-		let rc_len = self.garage.block_manager.rc_len()?.to_string();
+		let rc_len = self.garage.block_manager.rc_approximate_len()?.to_string();

 		writeln!(
 			&mut ret,
@@ -230,13 +230,13 @@ impl AdminRpcHandler {
 		writeln!(
 			&mut ret,
 			"  resync queue length: {}",
-			self.garage.block_manager.resync.queue_len()?
+			self.garage.block_manager.resync.queue_approximate_len()?
 		)
 		.unwrap();
 		writeln!(
 			&mut ret,
 			"  blocks with resync errors: {}",
-			self.garage.block_manager.resync.errors_len()?
+			self.garage.block_manager.resync.errors_approximate_len()?
 		)
 		.unwrap();

@@ -346,16 +346,21 @@ impl AdminRpcHandler {
 		F: TableSchema + 'static,
 		R: TableReplication + 'static,
 	{
-		let data_len = t.data.store.len().map_err(GarageError::from)?.to_string();
-		let mkl_len = t.merkle_updater.merkle_tree_len()?.to_string();
+		let data_len = t
+			.data
+			.store
+			.approximate_len()
+			.map_err(GarageError::from)?
+			.to_string();
+		let mkl_len = t.merkle_updater.merkle_tree_approximate_len()?.to_string();

 		Ok(format!(
 			"  {}\t{}\t{}\t{}\t{}",
 			F::TABLE_NAME,
 			data_len,
 			mkl_len,
-			t.merkle_updater.todo_len()?,
-			t.data.gc_todo_len()?
+			t.merkle_updater.todo_approximate_len()?,
+			t.data.gc_todo_approximate_len()?
 		))
 	}

@@ -71,6 +71,10 @@ pub enum NodeOperation {
 	/// Connect to Garage node that is currently isolated from the system
 	#[structopt(name = "connect", version = garage_version())]
 	Connect(ConnectNodeOpt),
+
+	/// Dump the content of a metadata table as JSON lines
+	#[structopt(name = "dump", version = garage_version())]
+	Dump(DumpNodeOpt),
 }

 #[derive(StructOpt, Debug)]
@@ -88,6 +92,12 @@ pub struct ConnectNodeOpt {
 	pub(crate) node: String,
 }

+#[derive(StructOpt, Debug)]
+pub struct DumpNodeOpt {
+	/// Name of the data table to dump
+	pub(crate) what: String,
+}
+
 #[derive(StructOpt, Debug)]
 pub enum LayoutOperation {
 	/// Assign role to Garage node
@@ -478,6 +488,9 @@ pub enum RepairWhat {
 	/// Recalculate block reference counters
 	#[structopt(name = "block-rc", version = garage_version())]
 	BlockRc,
+	/// Fix inconsistency in bucket aliases (WARNING: EXPERIMENTAL)
+	#[structopt(name = "aliases", version = garage_version())]
+	Aliases,
 	/// Verify integrity of all blocks on disc
 	#[structopt(name = "scrub", version = garage_version())]
 	Scrub {
@@ -145,11 +145,14 @@ async fn main() {
 	let res = match opt.cmd {
 		Command::Server => server::run_server(opt.config_file, opt.secrets).await,
 		Command::OfflineRepair(repair_opt) => {
-			repair::offline::offline_repair(opt.config_file, opt.secrets, repair_opt).await
+			repair::offline::offline_repair(opt.config_file, opt.secrets, repair_opt)
 		}
 		Command::ConvertDb(conv_opt) => {
 			cli::convert_db::do_conversion(conv_opt).map_err(From::from)
 		}
+		Command::Node(NodeOperation::Dump(dump_opt)) => {
+			repair::offline::dump(opt.config_file, opt.secrets, dump_opt)
+		}
 		Command::Node(NodeOperation::NodeId(node_id_opt)) => {
 			node_id_command(opt.config_file, node_id_opt.quiet)
 		}
@@ -208,6 +211,43 @@ fn init_logging(opt: &Opt) {
 		}
 	}

+	if std::env::var("GARAGE_LOG_TO_JOURNALD")
+		.map(|x| x == "1" || x == "true")
+		.unwrap_or(false)
+	{
+		#[cfg(feature = "journald")]
+		{
+			use tracing_journald::{Priority, PriorityMappings};
+			use tracing_subscriber::layer::SubscriberExt;
+			use tracing_subscriber::util::SubscriberInitExt;
+
+			let registry = tracing_subscriber::registry()
+				.with(tracing_subscriber::fmt::layer().with_writer(std::io::sink))
+				.with(env_filter);
+			match tracing_journald::layer() {
+				Ok(layer) => {
+					registry
+						.with(layer.with_priority_mappings(PriorityMappings {
+							info: Priority::Informational,
+							debug: Priority::Debug,
+							..PriorityMappings::new()
+						}))
+						.init();
+				}
+				Err(e) => {
+					eprintln!("Couldn't connect to journald: {}.", e);
+					std::process::exit(1);
+				}
+			}
+			return;
+		}
+		#[cfg(not(feature = "journald"))]
+		{
+			eprintln!("Journald support is not enabled in this build.");
+			std::process::exit(1);
+		}
+	}
+
 	tracing_subscriber::fmt()
 		.with_writer(std::io::stderr)
 		.with_env_filter(env_filter)
@@ -1,14 +1,18 @@
+use std::io::Write;
 use std::path::PathBuf;

+use serde::Serialize;
+
 use garage_util::config::*;
 use garage_util::error::*;

 use garage_model::garage::Garage;
+use garage_table::{replication::TableReplication, *};

 use crate::cli::structs::*;
 use crate::secrets::{fill_secrets, Secrets};

-pub async fn offline_repair(
+pub fn offline_repair(
 	config_file: PathBuf,
 	secrets: Secrets,
 	opt: OfflineRepairOpt,
@@ -45,3 +49,100 @@ pub async fn offline_repair(

 	Ok(())
 }
+
+pub fn dump(config_file: PathBuf, secrets: Secrets, opt: DumpNodeOpt) -> Result<(), Error> {
+	let what = opt.what.as_str();
+
+	info!("Loading configuration...");
+	let config = fill_secrets(read_config(config_file)?, secrets)?;
+
+	info!("Initializing Garage main data store...");
+	let garage = Garage::new(config)?;
+
+	match what {
+		"bucket" | "buckets" => dump_table_inner(&garage.bucket_table),
+		"bucket_alias" | "bucket_aliases" => dump_table_inner(&garage.bucket_alias_table),
+		"key" | "keys" => dump_table_inner(&garage.key_table),
+		"object" | "objects" => dump_table_inner(&garage.object_table),
+		"object_counter" | "object_counters" => Err(Error::Message(
+			"object_counters cannot be JSON-serialized".into(),
+		)),
+		"mpu" => dump_table_inner(&garage.mpu_table),
+		"mpu_counter" | "mpu_counters" => Err(Error::Message(
+			"mpu_counters cannot be JSON-serialized".into(),
+		)),
+		"version" | "versions" => dump_table_inner(&garage.version_table),
+		"block_ref" | "block_refs" => dump_table_inner(&garage.block_ref_table),
+		#[cfg(feature = "k2v")]
+		"k2v_item" | "k2v_items" => dump_table_inner(&garage.k2v.item_table),
+		//#[cfg(feature = "k2v")]
+		"k2v_counter" | "k2v_counters" => Err(Error::Message(
+			"k2v_counters cannot be JSON-serialized".into(),
+		)),
+		other => {
+			let mut stdout = std::io::stdout().lock();
+			match other {
+				"cluster_layout" => Err(Error::Message(
+					"cluster_layout cannot be JSON-serialized".into(),
+				)),
+				_ => Err(Error::Message(format!("invalid thing to dump: {}", what))),
+			}
+		}
+	}
+}
+
+#[derive(Serialize)]
+struct DumpEntry<'a, T: Serialize> {
+	#[serde(with = "serde_bytes")]
+	partition_key: &'a [u8],
+	#[serde(with = "serde_bytes")]
+	sort_key: &'a [u8],
+	entry: &'a T,
+}
+
+fn dump_table_inner<F, R>(table: &Table<F, R>) -> Result<(), Error>
+where
+	F: TableSchema,
+	R: TableReplication,
+{
+	eprintln!("Dumping table {}...", F::TABLE_NAME);
+
+	let mut stdout = std::io::stdout().lock();
+
+	for line in table.data.store.iter()? {
+		let (_k, v) = line?;
+		let v_dec = table.data.decode_entry(&v)?;
+		let pkh = v_dec.partition_key().hash();
+		let dump_entry = DumpEntry {
+			partition_key: pkh.as_slice(),
+			sort_key: v_dec.sort_key().sort_key(),
+			entry: &v_dec,
+		};
+		dump_line(&mut stdout, dump_entry)?;
+	}
+	stdout.flush()?;
+
+	Ok(())
+}
+
+fn dump_line<T: Serialize>(
+	mut stdout: &mut std::io::StdoutLock<'static>,
+	dump_entry: T,
+) -> Result<(), Error> {
+	let mut ser = serde_json::ser::Serializer::with_formatter(&mut stdout, DumpFormatter);
+	dump_entry.serialize(&mut ser)?;
+	stdout.write_all(b"\n")?;
+	Ok(())
+}
+
+struct DumpFormatter;
+impl serde_json::ser::Formatter for DumpFormatter {
+	fn write_byte_array<W>(&mut self, writer: &mut W, value: &[u8]) -> std::io::Result<()>
+	where
+		W: ?Sized + std::io::Write,
+	{
+		writer.write_all(b"\"")?;
+		writer.write_all(hex::encode(&value).as_bytes())?;
+		writer.write_all(b"\"")
+	}
+}
@@ -88,6 +88,10 @@ pub async fn launch_online_repair(
 				garage.block_manager.clone(),
 			));
 		}
+		RepairWhat::Aliases => {
+			info!("Repairing bucket aliases (foreground)");
+			garage.locked_helper().await.repair_aliases().await?;
+		}
 	}
 	Ok(())
 }
@@ -183,10 +183,21 @@ fn watch_shutdown_signal() -> watch::Receiver<bool> {
 		let mut sigterm =
 			signal(SignalKind::terminate()).expect("Failed to install SIGTERM handler");
 		let mut sighup = signal(SignalKind::hangup()).expect("Failed to install SIGHUP handler");
-		tokio::select! {
-			_ = sigint.recv() => info!("Received SIGINT, shutting down."),
-			_ = sigterm.recv() => info!("Received SIGTERM, shutting down."),
-			_ = sighup.recv() => info!("Received SIGHUP, shutting down."),
+		loop {
+			tokio::select! {
+					_ = sigint.recv() => {
+						info!("Received SIGINT, shutting down.");
+						break
+					}
+					_ = sigterm.recv() => {
+						info!("Received SIGTERM, shutting down.");
+						break
+					}
+					_ = sighup.recv() => {
+						info!("Received SIGHUP, reload not supported.");
+						continue
+					}
+			}
 		}
 		send_cancel.send(true).unwrap();
 	});
@@ -63,6 +63,8 @@ rpc_bind_addr = "127.0.0.1:{rpc_port}"
 rpc_public_addr = "127.0.0.1:{rpc_port}"
 rpc_secret = "{secret}"

+allow_punycode = true
+
 [s3_api]
 s3_region = "{region}"
 api_bind_addr = "127.0.0.1:{s3_port}"
@@ -533,3 +533,118 @@ async fn test_website_check_domain() {
 		})
 	);
 }
+
+#[tokio::test]
+async fn test_website_puny() {
+	const BCKT_NAME: &str = "xn--pda.eu";
+	let ctx = common::context();
+	let bucket = ctx.create_bucket(BCKT_NAME);
+
+	let data = ByteStream::from_static(BODY);
+
+	ctx.client
+		.put_object()
+		.bucket(&bucket)
+		.key("index.html")
+		.body(data)
+		.send()
+		.await
+		.unwrap();
+
+	let client = Client::builder(TokioExecutor::new()).build_http();
+
+	let req = |suffix| {
+		Request::builder()
+			.method("GET")
+			.uri(format!("http://127.0.0.1:{}/", ctx.garage.web_port))
+			.header("Host", format!("{}{}", BCKT_NAME, suffix))
+			.body(Body::new(Bytes::new()))
+			.unwrap()
+	};
+
+	ctx.garage
+		.command()
+		.args(["bucket", "website", "--allow", BCKT_NAME])
+		.quiet()
+		.expect_success_status("Could not allow website on bucket");
+
+	let mut resp = client.request(req("")).await.unwrap();
+	assert_eq!(resp.status(), StatusCode::OK);
+	assert_eq!(
+		resp.into_body().collect().await.unwrap().to_bytes(),
+		BODY.as_ref()
+	);
+
+	resp = client.request(req(".web.garage")).await.unwrap();
+	assert_eq!(resp.status(), StatusCode::OK);
+	assert_eq!(
+		resp.into_body().collect().await.unwrap().to_bytes(),
+		BODY.as_ref()
+	);
+
+	for bname in [
+		BCKT_NAME.to_string(),
+		format!("{BCKT_NAME}.web.garage"),
+		format!("{BCKT_NAME}.s3.garage"),
+	] {
+		let admin_req = || {
+			Request::builder()
+				.method("GET")
+				.uri(format!(
+					"http://127.0.0.1:{0}/check?domain={1}",
+					ctx.garage.admin_port, bname
+				))
+				.body(Body::new(Bytes::new()))
+				.unwrap()
+		};
+
+		let admin_resp = client.request(admin_req()).await.unwrap();
+		assert_eq!(admin_resp.status(), StatusCode::OK);
+		assert_eq!(
+			admin_resp.into_body().collect().await.unwrap().to_bytes(),
+			format!("Domain '{bname}' is managed by Garage").as_bytes()
+		);
+	}
+}
+
+#[tokio::test]
+async fn test_website_object_not_found() {
+	const BCKT_NAME: &str = "not-found";
+	let ctx = common::context();
+	let _bucket = ctx.create_bucket(BCKT_NAME);
+
+	let client = Client::builder(TokioExecutor::new()).build_http();
+
+	let req = |suffix| {
+		Request::builder()
+			.method("GET")
+			.uri(format!("http://127.0.0.1:{}/", ctx.garage.web_port))
+			.header("Host", format!("{}{}", BCKT_NAME, suffix))
+			.body(Body::new(Bytes::new()))
+			.unwrap()
+	};
+
+	ctx.garage
+		.command()
+		.args(["bucket", "website", "--allow", BCKT_NAME])
+		.quiet()
+		.expect_success_status("Could not allow website on bucket");
+
+	let resp = client.request(req("")).await.unwrap();
+	assert_eq!(resp.status(), StatusCode::NOT_FOUND);
+	// the error we return by default are *not* xml
+	assert_eq!(
+		resp.headers().get(http::header::CONTENT_TYPE).unwrap(),
+		"text/html; charset=utf-8"
+	);
+	let result = String::from_utf8(
+		resp.into_body()
+			.collect()
+			.await
+			.unwrap()
+			.to_bytes()
+			.to_vec(),
+	)
+	.unwrap();
+	assert!(result.contains("not found"));
+}
@@ -72,6 +72,16 @@ impl K2vClient {
 			.enable_http2()
 			.build();
 		let client = HttpClient::builder(TokioExecutor::new()).build(connector);
+		Self::new_with_client(config, client)
+	}
+
+	/// Create a new K2V client with an external client.
+	/// Useful for example if you plan on creating many clients but you want to mutualize the
+	/// underlying thread pools & co.
+	pub fn new_with_client(
+		config: K2vClientConfig,
+		client: HttpClient<HttpsConnector<HttpConnector>, Body>,
+	) -> Result<Self, Error> {
 		let user_agent: std::borrow::Cow<str> = match &config.user_agent {
 			Some(ua) => ua.into(),
 			None => format!("k2v/{}", env!("CARGO_PKG_VERSION")).into(),
@@ -1,6 +1,6 @@
 [package]
 name = "garage_model"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -44,3 +44,4 @@ default = [ "lmdb", "sqlite" ]
 k2v = [ "garage_util/k2v" ]
 lmdb = [ "garage_db/lmdb" ]
 sqlite = [ "garage_db/sqlite" ]
+fjall = [ "garage_db/fjall" ]
@@ -22,14 +22,10 @@ mod v08 {
 pub use v08::*;

 impl BucketAlias {
-	pub fn new(name: String, ts: u64, bucket_id: Option<Uuid>) -> Option<Self> {
-		if !is_valid_bucket_name(&name) {
-			None
-		} else {
-			Some(BucketAlias {
-				name,
-				state: crdt::Lww::raw(ts, bucket_id),
-			})
+	pub fn new(name: String, ts: u64, bucket_id: Option<Uuid>) -> Self {
+		BucketAlias {
+			name,
+			state: crdt::Lww::raw(ts, bucket_id),
 		}
 	}

@@ -80,7 +76,7 @@ impl TableSchema for BucketAliasTable {
 /// In the case of Garage, bucket names must not be hex-encoded
 /// 32 byte string, which is excluded thanks to the
 /// maximum length of 63 bytes given in the spec.
-pub fn is_valid_bucket_name(n: &str) -> bool {
+pub fn is_valid_bucket_name(n: &str, puny: bool) -> bool {
 	// Bucket names must be between 3 and 63 characters
 	n.len() >= 3 && n.len() <= 63
 	// Bucket names must be composed of lowercase letters, numbers,
@@ -92,7 +88,9 @@ pub fn is_valid_bucket_name(n: &str) -> bool {
 	// Bucket names must not be formatted as an IP address
 	&& n.parse::<std::net::IpAddr>().is_err()
 	// Bucket names must not start with "xn--"
-	&& !n.starts_with("xn--")
+	&& (!n.starts_with("xn--") || puny)
+        // We are a bit stricter, to properly restrict punycode in all labels
+	&& (!n.contains(".xn--") || puny)
 	// Bucket names must not end with "-s3alias"
 	&& !n.ends_with("-s3alias")
 }
@@ -116,21 +116,17 @@ impl Garage {
 		info!("Opening database...");
 		let db_engine = db::Engine::from_str(&config.db_engine)
 			.ok_or_message("Invalid `db_engine` value in configuration file")?;
-		let mut db_path = config.metadata_dir.clone();
-		match db_engine {
-			db::Engine::Sqlite => {
-				db_path.push("db.sqlite");
-			}
-			db::Engine::Lmdb => {
-				db_path.push("db.lmdb");
-			}
-		}
+		let db_path = db_engine.db_path(&config.metadata_dir);
 		let db_opt = db::OpenOpt {
 			fsync: config.metadata_fsync,
 			lmdb_map_size: match config.lmdb_map_size {
 				v if v == usize::default() => None,
 				v => Some(v),
 			},
+			fjall_block_cache_size: match config.fjall_block_cache_size {
+				v if v == usize::default() => None,
+				v => Some(v),
+			},
 		};
 		let db = db::open_db(&db_path, db_engine, &db_opt)
 			.ok_or_message("Unable to open metadata db")?;
@@ -1,3 +1,7 @@
+use std::collections::{HashMap, HashSet};
+
+use garage_db as db;
+
 use garage_util::crdt::*;
 use garage_util::data::*;
 use garage_util::error::{Error as GarageError, OkOrMessage};
@@ -47,6 +51,10 @@ impl<'a> LockedHelper<'a> {
 		KeyHelper(self.0)
 	}

+	// ================================================
+	//      global bucket aliases
+	// ================================================
+
 	/// Sets a new alias for a bucket in global namespace.
 	/// This function fails if:
 	/// - alias name is not valid according to S3 spec
@@ -57,7 +65,7 @@ impl<'a> LockedHelper<'a> {
 		bucket_id: Uuid,
 		alias_name: &String,
 	) -> Result<(), Error> {
-		if !is_valid_bucket_name(alias_name) {
+		if !is_valid_bucket_name(alias_name, self.0.config.allow_punycode) {
 			return Err(Error::InvalidBucketName(alias_name.to_string()));
 		}

@@ -88,8 +96,7 @@ impl<'a> LockedHelper<'a> {
 		// writes are now done and all writes use timestamp alias_ts

 		let alias = match alias {
-			None => BucketAlias::new(alias_name.clone(), alias_ts, Some(bucket_id))
-				.ok_or_else(|| Error::InvalidBucketName(alias_name.clone()))?,
+			None => BucketAlias::new(alias_name.clone(), alias_ts, Some(bucket_id)),
 			Some(mut a) => {
 				a.state = Lww::raw(alias_ts, Some(bucket_id));
 				a
@@ -180,13 +187,14 @@ impl<'a> LockedHelper<'a> {
 			.ok_or_else(|| Error::NoSuchBucket(alias_name.to_string()))?;

 		// Checks ok, remove alias
-		let alias_ts = match bucket.state.as_option() {
-			Some(bucket_state) => increment_logical_clock_2(
-				alias.state.timestamp(),
-				bucket_state.aliases.get_timestamp(alias_name),
-			),
-			None => increment_logical_clock(alias.state.timestamp()),
-		};
+		let alias_ts = increment_logical_clock_2(
+			alias.state.timestamp(),
+			bucket
+				.state
+				.as_option()
+				.map(|p| p.aliases.get_timestamp(alias_name))
+				.unwrap_or(0),
+		);

 		// ---- timestamp-ensured causality barrier ----
 		// writes are now done and all writes use timestamp alias_ts
@@ -204,6 +212,10 @@ impl<'a> LockedHelper<'a> {
 		Ok(())
 	}

+	// ================================================
+	//      local bucket aliases
+	// ================================================
+
 	/// Sets a new alias for a bucket in the local namespace of a key.
 	/// This function fails if:
 	/// - alias name is not valid according to S3 spec
@@ -216,14 +228,12 @@ impl<'a> LockedHelper<'a> {
 		key_id: &String,
 		alias_name: &String,
 	) -> Result<(), Error> {
-		let key_helper = KeyHelper(self.0);
-
-		if !is_valid_bucket_name(alias_name) {
+		if !is_valid_bucket_name(alias_name, self.0.config.allow_punycode) {
 			return Err(Error::InvalidBucketName(alias_name.to_string()));
 		}

 		let mut bucket = self.bucket().get_existing_bucket(bucket_id).await?;
-		let mut key = key_helper.get_existing_key(key_id).await?;
+		let mut key = self.key().get_existing_key(key_id).await?;

 		let key_param = key.state.as_option_mut().unwrap();

@@ -272,23 +282,13 @@ impl<'a> LockedHelper<'a> {
 		key_id: &String,
 		alias_name: &String,
 	) -> Result<(), Error> {
-		let key_helper = KeyHelper(self.0);
-
 		let mut bucket = self.bucket().get_existing_bucket(bucket_id).await?;
-		let mut key = key_helper.get_existing_key(key_id).await?;
+		let mut key = self.key().get_existing_key(key_id).await?;

+		let key_p = key.state.as_option().unwrap();
 		let bucket_p = bucket.state.as_option_mut().unwrap();

-		if key
-			.state
-			.as_option()
-			.unwrap()
-			.local_aliases
-			.get(alias_name)
-			.cloned()
-			.flatten()
-			!= Some(bucket_id)
-		{
+		if key_p.local_aliases.get(alias_name).cloned().flatten() != Some(bucket_id) {
 			return Err(GarageError::Message(format!(
 				"Bucket {:?} does not have alias {} in namespace of key {}",
 				bucket_id, alias_name, key_id
@@ -305,17 +305,17 @@ impl<'a> LockedHelper<'a> {
 			.local_aliases
 			.items()
 			.iter()
-			.any(|((k, n), _, active)| *k == key.key_id && n == alias_name && *active);
+			.any(|((k, n), _, active)| (*k != key.key_id || n != alias_name) && *active);
+
 		if !has_other_global_aliases && !has_other_local_aliases {
 			return Err(Error::BadRequest(format!("Bucket {} doesn't have other aliases, please delete it instead of just unaliasing.", alias_name)));
 		}

 		// Checks ok, remove alias
-		let key_param = key.state.as_option_mut().unwrap();
 		let bucket_p_local_alias_key = (key.key_id.clone(), alias_name.clone());

 		let alias_ts = increment_logical_clock_2(
-			key_param.local_aliases.get_timestamp(alias_name),
+			key_p.local_aliases.get_timestamp(alias_name),
 			bucket_p
 				.local_aliases
 				.get_timestamp(&bucket_p_local_alias_key),
@@ -324,7 +324,8 @@ impl<'a> LockedHelper<'a> {
 		// ---- timestamp-ensured causality barrier ----
 		// writes are now done and all writes use timestamp alias_ts

-		key_param.local_aliases = LwwMap::raw_item(alias_name.clone(), alias_ts, None);
+		key.state.as_option_mut().unwrap().local_aliases =
+			LwwMap::raw_item(alias_name.clone(), alias_ts, None);
 		self.0.key_table.insert(&key).await?;

 		bucket_p.local_aliases = LwwMap::raw_item(bucket_p_local_alias_key, alias_ts, false);
@@ -333,21 +334,68 @@ impl<'a> LockedHelper<'a> {
 		Ok(())
 	}

+	/// Ensures a bucket does not have a certain local alias.
+	/// Contrarily to unset_local_bucket_alias, this does not
+	/// fail on any condition other than:
+	/// - bucket cannot be found (its fine if it is in deleted state)
+	/// - key cannot be found (its fine if alias in key points to nothing
+	///   or to another bucket)
+	pub async fn purge_local_bucket_alias(
+		&self,
+		bucket_id: Uuid,
+		key_id: &String,
+		alias_name: &String,
+	) -> Result<(), Error> {
+		let mut bucket = self.bucket().get_internal_bucket(bucket_id).await?;
+		let mut key = self.key().get_internal_key(key_id).await?;
+
+		let bucket_p_local_alias_key = (key.key_id.clone(), alias_name.clone());
+
+		let alias_ts = increment_logical_clock_2(
+			key.state
+				.as_option()
+				.map(|p| p.local_aliases.get_timestamp(alias_name))
+				.unwrap_or(0),
+			bucket
+				.state
+				.as_option()
+				.map(|p| p.local_aliases.get_timestamp(&bucket_p_local_alias_key))
+				.unwrap_or(0),
+		);
+
+		// ---- timestamp-ensured causality barrier ----
+		// writes are now done and all writes use timestamp alias_ts
+
+		if let Some(kp) = key.state.as_option_mut() {
+			kp.local_aliases = LwwMap::raw_item(alias_name.clone(), alias_ts, None);
+			self.0.key_table.insert(&key).await?;
+		}
+
+		if let Some(bp) = bucket.state.as_option_mut() {
+			bp.local_aliases = LwwMap::raw_item(bucket_p_local_alias_key, alias_ts, false);
+			self.0.bucket_table.insert(&bucket).await?;
+		}
+
+		Ok(())
+	}
+
+	// ================================================
+	//      permissions
+	// ================================================
+
 	/// Sets permissions for a key on a bucket.
 	/// This function fails if:
 	/// - bucket or key cannot be found at all (its ok if they are in deleted state)
-	/// - bucket or key is in deleted state and we are trying to set permissions other than "deny
-	/// all"
+	/// - bucket or key is in deleted state and we are trying to set
+	///   permissions other than "deny all"
 	pub async fn set_bucket_key_permissions(
 		&self,
 		bucket_id: Uuid,
 		key_id: &String,
 		mut perm: BucketKeyPerm,
 	) -> Result<(), Error> {
-		let key_helper = KeyHelper(self.0);
-
 		let mut bucket = self.bucket().get_internal_bucket(bucket_id).await?;
-		let mut key = key_helper.get_internal_key(key_id).await?;
+		let mut key = self.key().get_internal_key(key_id).await?;

 		if let Some(bstate) = bucket.state.as_option() {
 			if let Some(kp) = bstate.authorized_keys.get(key_id) {
@@ -384,21 +432,20 @@ impl<'a> LockedHelper<'a> {
 		Ok(())
 	}

-	// ----
+	// ================================================
+	//      keys
+	// ================================================

 	/// Deletes an API access key
 	pub async fn delete_key(&self, key: &mut Key) -> Result<(), Error> {
 		let state = key.state.as_option_mut().unwrap();

 		// --- done checking, now commit ---
-		// (the step at unset_local_bucket_alias will fail if a bucket
-		// does not have another alias, the deletion will be
-		// interrupted in the middle if that happens)

 		// 1. Delete local aliases
 		for (alias, _, to) in state.local_aliases.items().iter() {
 			if let Some(bucket_id) = to {
-				self.unset_local_bucket_alias(*bucket_id, &key.key_id, alias)
+				self.purge_local_bucket_alias(*bucket_id, &key.key_id, alias)
 					.await?;
 			}
 		}
@@ -415,4 +462,193 @@ impl<'a> LockedHelper<'a> {

 		Ok(())
 	}
+
+	// ================================================
+	//      repair procedure
+	// ================================================
+
+	pub async fn repair_aliases(&self) -> Result<(), GarageError> {
+		self.0.db.transaction(|tx| {
+			info!("--- begin repair_aliases transaction ----");
+
+			// 1. List all non-deleted buckets, so that we can fix bad aliases
+			let mut all_buckets: HashSet<Uuid> = HashSet::new();
+
+			for item in tx.range::<&[u8], _>(&self.0.bucket_table.data.store, ..)? {
+				let bucket = self
+					.0
+					.bucket_table
+					.data
+					.decode_entry(&(item?.1))
+					.map_err(db::TxError::Abort)?;
+				if !bucket.is_deleted() {
+					all_buckets.insert(bucket.id);
+				}
+			}
+
+			info!("number of buckets: {}", all_buckets.len());
+
+			// 2. List all aliases declared in bucket_alias_table and key_table
+			//    Take note of aliases that point to non-existing buckets
+			let mut global_aliases: HashMap<String, Uuid> = HashMap::new();
+
+			{
+				let mut delete_global = vec![];
+				for item in tx.range::<&[u8], _>(&self.0.bucket_alias_table.data.store, ..)? {
+					let mut alias = self
+						.0
+						.bucket_alias_table
+						.data
+						.decode_entry(&(item?.1))
+						.map_err(db::TxError::Abort)?;
+					if let Some(id) = alias.state.get() {
+						if all_buckets.contains(id) {
+							// keep aliases
+							global_aliases.insert(alias.name().to_string(), *id);
+						} else {
+							// delete alias
+							warn!(
+								"global alias: remove {} -> {:?} (bucket is deleted)",
+								alias.name(),
+								id
+							);
+							alias.state.update(None);
+							delete_global.push(alias);
+						}
+					}
+				}
+
+				info!("number of global aliases: {}", global_aliases.len());
+
+				info!("global alias table: {} entries fixed", delete_global.len());
+				for ga in delete_global {
+					debug!("Enqueue update to global alias table: {:?}", ga);
+					self.0.bucket_alias_table.queue_insert(tx, &ga)?;
+				}
+			}
+
+			let mut local_aliases: HashMap<(String, String), Uuid> = HashMap::new();
+
+			{
+				let mut delete_local = vec![];
+
+				for item in tx.range::<&[u8], _>(&self.0.key_table.data.store, ..)? {
+					let mut key = self
+						.0
+						.key_table
+						.data
+						.decode_entry(&(item?.1))
+						.map_err(db::TxError::Abort)?;
+					let Some(p) = key.state.as_option_mut() else {
+						continue;
+					};
+					let mut has_changes = false;
+					for (name, _, to) in p.local_aliases.items().to_vec() {
+						if let Some(id) = to {
+							if all_buckets.contains(&id) {
+								local_aliases.insert((key.key_id.clone(), name), id);
+							} else {
+								warn!(
+									"local alias: remove ({}, {}) -> {:?} (bucket is deleted)",
+									key.key_id, name, id
+								);
+								p.local_aliases.update_in_place(name, None);
+								has_changes = true;
+							}
+						}
+					}
+					if has_changes {
+						delete_local.push(key);
+					}
+				}
+
+				info!("number of local aliases: {}", local_aliases.len());
+
+				info!("key table: {} entries fixed", delete_local.len());
+				for la in delete_local {
+					debug!("Enqueue update to key table: {:?}", la);
+					self.0.key_table.queue_insert(tx, &la)?;
+				}
+			}
+
+			// 4. Reverse the alias maps to determine the aliases per-bucket
+			let mut bucket_global: HashMap<Uuid, Vec<String>> = HashMap::new();
+			let mut bucket_local: HashMap<Uuid, Vec<(String, String)>> = HashMap::new();
+
+			for (name, bucket) in global_aliases {
+				bucket_global.entry(bucket).or_default().push(name);
+			}
+			for ((key, name), bucket) in local_aliases {
+				bucket_local.entry(bucket).or_default().push((key, name));
+			}
+
+			// 5. Fix the bucket table to ensure consistency
+			let mut bucket_updates = vec![];
+
+			for item in tx.range::<&[u8], _>(&self.0.bucket_table.data.store, ..)? {
+				let bucket = self
+					.0
+					.bucket_table
+					.data
+					.decode_entry(&(item?.1))
+					.map_err(db::TxError::Abort)?;
+				let mut bucket2 = bucket.clone();
+				let Some(param) = bucket2.state.as_option_mut() else {
+					continue;
+				};
+
+				// fix global aliases
+				{
+					let ga = bucket_global.remove(&bucket.id).unwrap_or_default();
+					for (name, _, active) in param.aliases.items().to_vec() {
+						if active && !ga.contains(&name) {
+							warn!("bucket {:?}: remove global alias {}", bucket.id, name);
+							param.aliases.update_in_place(name, false);
+						}
+					}
+					for name in ga {
+						if param.aliases.get(&name).copied() != Some(true) {
+							warn!("bucket {:?}: add global alias {}", bucket.id, name);
+							param.aliases.update_in_place(name, true);
+						}
+					}
+				}
+
+				// fix local aliases
+				{
+					let la = bucket_local.remove(&bucket.id).unwrap_or_default();
+					for (pair, _, active) in param.local_aliases.items().to_vec() {
+						if active && !la.contains(&pair) {
+							warn!("bucket {:?}: remove local alias {:?}", bucket.id, pair);
+							param.local_aliases.update_in_place(pair, false);
+						}
+					}
+					for pair in la {
+						if param.local_aliases.get(&pair).copied() != Some(true) {
+							warn!("bucket {:?}: add local alias {:?}", bucket.id, pair);
+							param.local_aliases.update_in_place(pair, true);
+						}
+					}
+				}
+
+				if bucket2 != bucket {
+					bucket_updates.push(bucket2);
+				}
+			}
+
+			info!("bucket table: {} entries fixed", bucket_updates.len());
+			for b in bucket_updates {
+				debug!("Enqueue update to bucket table: {:?}", b);
+				self.0.bucket_table.queue_insert(tx, &b)?;
+			}
+
+			info!("--- end repair_aliases transaction ----");
+
+			Ok(())
+		})?;
+
+		info!("repair_aliases is done");
+
+		Ok(())
+	}
 }
@@ -121,13 +121,13 @@ impl Worker for LifecycleWorker {
 				mpu_aborted,
 				..
 			} => {
-				let n_objects = self.garage.object_table.data.store.len().ok();
+				let n_objects = self.garage.object_table.data.store.approximate_len().ok();
 				let progress = match n_objects {
-					None => "...".to_string(),
-					Some(total) => format!(
+					Some(total) if total > 0 => format!(
 						"~{:.2}%",
 						100. * std::cmp::min(*counter, total) as f32 / total as f32
 					),
+					_ => "...".to_string(),
 				};
 				WorkerStatus {
 					progress: Some(progress),
@@ -1,6 +1,6 @@
 [package]
 name = "garage_net"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -1,6 +1,6 @@
 [package]
 name = "garage_rpc"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -1,6 +1,6 @@
 [package]
 name = "garage_table"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -367,7 +367,7 @@ impl<F: TableSchema, R: TableReplication> TableData<F, R> {
 		}
 	}

-	pub fn gc_todo_len(&self) -> Result<usize, Error> {
-		Ok(self.gc_todo.len()?)
+	pub fn gc_todo_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.gc_todo.approximate_len()?)
 	}
 }
@@ -313,7 +313,7 @@ impl<F: TableSchema, R: TableReplication> Worker for GcWorker<F, R> {

 	fn status(&self) -> WorkerStatus {
 		WorkerStatus {
-			queue_length: Some(self.gc.data.gc_todo_len().unwrap_or(0) as u64),
+			queue_length: Some(self.gc.data.gc_todo_approximate_len().unwrap_or(0) as u64),
 			..Default::default()
 		}
 	}
@@ -287,12 +287,12 @@ impl<F: TableSchema, R: TableReplication> MerkleUpdater<F, R> {
 		MerkleNode::decode_opt(&ent)
 	}

-	pub fn merkle_tree_len(&self) -> Result<usize, Error> {
-		Ok(self.data.merkle_tree.len()?)
+	pub fn merkle_tree_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.data.merkle_tree.approximate_len()?)
 	}

-	pub fn todo_len(&self) -> Result<usize, Error> {
-		Ok(self.data.merkle_todo.len()?)
+	pub fn todo_approximate_len(&self) -> Result<usize, Error> {
+		Ok(self.data.merkle_todo.approximate_len()?)
 	}
 }

@@ -306,7 +306,7 @@ impl<F: TableSchema, R: TableReplication> Worker for MerkleWorker<F, R> {

 	fn status(&self) -> WorkerStatus {
 		WorkerStatus {
-			queue_length: Some(self.0.todo_len().unwrap_or(0) as u64),
+			queue_length: Some(self.0.todo_approximate_len().unwrap_or(0) as u64),
 			..Default::default()
 		}
 	}
@@ -34,7 +34,7 @@ impl TableMetrics {
 				.u64_value_observer(
 					"table.size",
 					move |observer| {
-						if let Ok(value) = store.len() {
+						if let Ok(value) = store.approximate_len() {
 							observer.observe(
 								value as u64,
 								&[KeyValue::new("table_name", table_name)],
@@ -48,7 +48,7 @@ impl TableMetrics {
 				.u64_value_observer(
 					"table.merkle_tree_size",
 					move |observer| {
-						if let Ok(value) = merkle_tree.len() {
+						if let Ok(value) = merkle_tree.approximate_len() {
 							observer.observe(
 								value as u64,
 								&[KeyValue::new("table_name", table_name)],
@@ -62,7 +62,7 @@ impl TableMetrics {
 				.u64_value_observer(
 					"table.merkle_updater_todo_queue_length",
 					move |observer| {
-						if let Ok(v) = merkle_todo.len() {
+						if let Ok(v) = merkle_todo.approximate_len() {
 							observer.observe(
 								v as u64,
 								&[KeyValue::new("table_name", table_name)],
@@ -76,7 +76,7 @@ impl TableMetrics {
 				.u64_value_observer(
 					"table.gc_todo_queue_length",
 					move |observer| {
-                        if let Ok(value) = gc_todo.len() {
+                        if let Ok(value) = gc_todo.approximate_len() {
                            observer.observe(
                                value as u64,
                                &[KeyValue::new("table_name", table_name)],
@@ -27,7 +27,7 @@ impl<F: TableSchema, R: TableReplication> Worker for InsertQueueWorker<F, R> {

 	fn status(&self) -> WorkerStatus {
 		WorkerStatus {
-			queue_length: Some(self.0.data.insert_queue.len().unwrap_or(0) as u64),
+			queue_length: Some(self.0.data.insert_queue.approximate_len().unwrap_or(0) as u64),
 			..Default::default()
 		}
 	}
@@ -1,6 +1,6 @@
 [package]
 name = "garage_util"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -122,6 +122,10 @@ pub struct Config {
 	#[serde(deserialize_with = "deserialize_capacity", default)]
 	pub lmdb_map_size: usize,

+	/// Fjall block cache size
+	#[serde(deserialize_with = "deserialize_capacity", default)]
+	pub fjall_block_cache_size: usize,
+
 	// -- APIs
 	/// Configuration for S3 api
 	pub s3_api: S3ApiConfig,
@@ -135,6 +139,10 @@ pub struct Config {
 	/// Configuration for the admin API endpoint
 	#[serde(default = "Default::default")]
 	pub admin: AdminConfig,
+
+	/// Allow punycode in bucket names
+	#[serde(default)]
+	pub allow_punycode: bool,
 }

 /// Value for data_dir: either a single directory or a list of dirs with attributes
@@ -1,6 +1,6 @@
 [package]
 name = "garage_web"
-version = "1.1.0"
+version = "1.2.0"
 authors = ["Alex Auvolat <alex@adnab.me>", "Quentin Dufour <quentin@dufour.io>"]
 edition = "2018"
 license = "AGPL-3.0"
@@ -397,10 +397,30 @@ fn error_to_res(e: Error) -> Response<BoxBody<Error>> {
 	//   was a HEAD request or we couldn't get the error document)
 	// We do NOT enter this code path when returning the bucket's
 	// error document (this is handled in serve_file)
-	let body = string_body(format!("{}\n", e));
-	let mut http_error = Response::new(body);
+	let mut body_str = format!(
+		r"<title>{http_code} {code_text}</title>
+<h1>{http_code} {code_text}</h1>",
+		http_code = e.http_status_code().as_u16(),
+		code_text = e.http_status_code().canonical_reason().unwrap_or("Unknown"),
+	);
+	if let Error::ApiError(ref err) = e {
+		body_str.push_str(&format!(
+			r"
+<ul>
+<li>Code: {s3_code}</li>
+<li>Message: {s3_message}.</li>
+</ul>",
+			s3_code = err.aws_code(),
+			s3_message = err,
+		));
+	}
+	let mut http_error = Response::new(string_body(body_str));
 	*http_error.status_mut() = e.http_status_code();
 	e.add_headers(http_error.headers_mut());
+	http_error.headers_mut().insert(
+		http::header::CONTENT_TYPE,
+		"text/html; charset=utf-8".parse().unwrap(),
+	);
 	http_error
 }
Author	SHA1	Message	Date
Alex Auvolat	8bccddb504	add `garage node dump-table` command for debugging purpose	2025-08-29 16:35:47 +02:00
Alex	0b901bf291	Merge pull request 'garage_db: reduce frequency of sqlite snapshot progress log (fix #1129 )' (#1146 ) from fix-1129 into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1146	2025-08-27 22:26:32 +00:00
Alex Auvolat	c8c20d6f47	garage_db: reduce frequency of sqlite snapshot progress log (fix #1129 )	2025-08-28 00:07:35 +02:00
Alex	e5db610e4c	Merge pull request 'K2V client: allow custom HTTP client' (#731 ) from k2v/shared_http_client into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/731 Reviewed-by: maximilien <me@mricher.fr>	2025-08-27 21:21:09 +00:00
Alex	65c6f8adea	Merge pull request 'garage_db: refactor open function' (#1142 ) from factor-db-open into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1142	2025-08-27 21:10:59 +00:00
Alex Auvolat	54b9bf02a3	garage_db: refactor open function	2025-08-27 23:03:09 +02:00
Alex	469153233f	Merge pull request 'garage_db: rename len to approximate_len as it is used for stats only' (#1141 ) from db-approximate-len into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1141	2025-08-27 20:44:50 +00:00
Alex Auvolat	90bba5889a	garage_db: rename len to approximate_len as it is used for stats only	2025-08-27 21:23:45 +02:00
Alex	a64b567d43	Merge pull request 'Add experimental support for Fjall DB engine' (#906 ) from withings/garage:feat/fjall-db-engine into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/906	2025-08-27 19:09:40 +00:00
Alex Auvolat	6ea86db8cd	document fjall db engine, remove flakey metadata_fsync implementation	2025-08-27 20:22:41 +02:00
Alex Auvolat	aa69c06f2b	fix potential race condition and naming bug in fjall adapter	2025-08-27 20:22:38 +02:00
Alex Auvolat	a6c6c44310	nix: build and test fjall feature	2025-08-27 18:54:42 +02:00
Julien Kritter	96d7713915	Add support for an LSM-tree-based backend with Fjall	2025-08-27 18:54:34 +02:00
Alex	d64498c3d3	Merge pull request 'log access keys' (#1122 ) from 1686a/log-access-key into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1122	2025-08-27 16:18:16 +00:00
trinity-1686a	b340599e68	log access keys	2025-08-03 15:30:56 +02:00
Alex	5448012b27	Merge pull request 'Pixelfed_support' (#1118 ) from Lapineige/garage:Pixelfed_support into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1118	2025-08-02 15:03:57 +00:00
Alex	ce34d11a65	Merge pull request 'don't die on SIGHUP' (#1121 ) from 1686a/handle-sighup into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1121	2025-08-02 14:53:58 +00:00
Alex	8cb7623ebd	Merge pull request 'handle ECONNABORTED' (#1120 ) from 1686a/handle-econnaborted into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1120	2025-08-02 14:53:45 +00:00
trinity-1686a	5469c95877	handle ECONNABORTED	2025-08-02 13:14:01 +02:00
trinity-1686a	f930c6f643	don't die on SIGHUP	2025-08-02 13:09:33 +02:00
Alex	afcb22bf16	Merge pull request 'Fix typo in peertube buckets names' (#1117 ) from Lapineige/garage:main into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1117	2025-08-02 08:27:01 +00:00
Lapineige	cc29a40d51	Actualiser doc/book/connect/apps/index.md	2025-08-01 21:35:15 +00:00
Lapineige	0f3f180c3e	Merge branch 'main-v1' into main	2025-08-01 21:33:58 +00:00
Lapineige	70cf6004ae	Fix typo in peertube buckets names	2025-08-01 21:32:59 +00:00
Alex	c7571ff89b	Merge pull request 'Fix some unsoundness in lmdb adapter unsafe' (#1099 ) from krtab/garage:fix_some_ub into main-v1 Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1099	2025-07-31 19:38:23 +00:00
Arthur Carcano	1b42919bf7	Fix some unsoundness in lmdb adapter unsafe	2025-07-25 23:33:51 +02:00
Alex	3f4ab3a4a3	Merge pull request 'Garage v1.2.0' (#1068 ) from rel-1.2.0 into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1068	2025-06-13 16:12:29 +00:00
Alex Auvolat	3a4afc04a9	cargo: update crossbeam-channel to avoid yanked version	2025-06-13 17:22:47 +02:00
Alex Auvolat	fbf03e9378	bump version to v1.2.0	2025-06-13 14:21:28 +02:00
Alex	9eb07d4c7b	Merge pull request 'cli: mark block refs as deleted in garage block purge (fix #1055 )' (#1067 ) from fix-1055 into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1067	2025-06-13 11:53:41 +00:00
Alex Auvolat	85ee4f5d8c	cli: mark block refs as deleted in garage block purge	2025-06-13 13:52:02 +02:00
Alex	328072d122	Merge pull request 'put web error in a basic webpage' (#1064 ) from trinity-1686a/garage:1686a/non-xml-web-error into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1064	2025-06-12 06:06:38 +00:00
trinity-1686a	26bc807905	put web error in a basic webpage before, it was a plain string, with an xml content type this caused browsers to show very ugly and meaningless pages	2025-06-10 22:23:06 +02:00
Alex	a9f5f242b2	Merge pull request 'feat: add log to journald feature' (#1056 ) from ragazenta/garage:feat/tracing-journald into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1056	2025-06-10 18:38:23 +00:00
maximilien	ae98abca5c	Merge pull request 'Add eddster2309/ansible-role-garage as deployment option' (#1057 ) from eddster2309/garage:main into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1057 Reviewed-by: maximilien <me@mricher.fr>	2025-06-08 11:56:31 +00:00
eddster2309	adfa44ad70	Add architecture support	2025-06-03 09:22:43 +00:00
eddster2309	47143b88ad	Add eddster2309/ansible-role-garage as deployment option	2025-06-03 09:15:57 +00:00
Renjaya Raga Zenta	8843aa92fa	feat: add log to journald feature The systemd-journald is used in most major Linux distros that use systemd. This enables logging using the systemd-journald native protocol, instead of just writing to stderr.	2025-06-02 11:55:27 +07:00
Alex	b601b3e46d	Merge pull request 'documentation: Minor doc change to clarify why the capacity does not matter and how the zone name is used' (#1051 ) from ddxv/garage:docs-quick-start into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1051	2025-05-30 16:26:19 +00:00
Alex	a19d2f16e2	Merge pull request 'api: s3: implement get bucket acl' (#1045 ) from ragazenta/garage:feat/dummy-acl into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1045	2025-05-30 16:25:04 +00:00
trinity-1686a	fc8fc60f6d	emit internal error when we detect race condition (#1053 ) (fix #1050 ) i went with a `500`/`InternalError`/`Please try again.` because that is something i've seen AWS S3 report while developing other software, and i'm not convinced all clients would understand a 409 conflict properly (GET don't usually conflict) Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1053 Co-authored-by: trinity-1686a <trinity@deuxfleurs.fr> Co-committed-by: trinity-1686a <trinity@deuxfleurs.fr>	2025-05-30 16:24:12 +00:00
Alex	77079a1498	Merge pull request '[1.1.x] speed up UploadPartCopy' (#1047 ) from yuka/garage:uploadpartcopy-v1 into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1047	2025-05-30 16:22:35 +00:00
James O'Claire	2a4f729b57	Minor doc change to clarify why the capacity does not matter and how the zone name is used	2025-05-28 09:49:50 +08:00
Renjaya Raga Zenta	1b042e379e	api: s3: implement get bucket acl	2025-05-26 09:43:15 +07:00
Yureka	ffbce0f689	speed up UploadPartCopy (cherry picked from commit `db54bf96c7`)	2025-05-23 20:36:32 +02:00
Alex	37e5621dde	Merge pull request 'documentation updates' (#1046 ) from doc-updates into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1046	2025-05-23 15:05:19 +00:00
Alex Auvolat	6529ff379a	documentation updates	2025-05-23 17:02:23 +02:00
Alex	a8d73682a4	Merge pull request 'more resilience to inconsistent alias states' (#989 ) from fix-bucket-aliases into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/989	2025-05-22 17:41:42 +00:00
Alex Auvolat	8654eb19bf	implement repair procedure to fix inconsistent bucket aliases	2025-05-22 19:34:38 +02:00
maximilien	54ea412188	Merge pull request 'Add kubernetes CRD' (#994 ) from babykart/garage:k8s-crd into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/994	2025-05-22 17:15:56 +00:00
Alex Auvolat	2ade8c86f6	more resilience to inconsistent alias states	2025-05-22 19:12:05 +02:00
babykart	b15e2cbb6c	Update Kubernetes cookbook Signed-off-by: babykart <babykart@gmail.com>	2025-05-22 17:11:14 +00:00
babykart	0fd1b7342b	Add Kubernetes CRD and the related kustomization Signed-off-by: babykart <babykart@gmail.com>	2025-05-22 17:11:14 +00:00
Alex	be16bc7a05	Merge pull request 'Fix behavior of CopyObject wrt x-amz-website-redirect-location' (#1037 ) from Armael/garage:copy-website-redirect into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1037	2025-05-22 13:57:28 +00:00
Alex	bfaa1ca6b7	Merge pull request 'api: lifecycle: 404 if missing lifecycle config' (#1043 ) from ragazenta/garage:no-lifecycle-response into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1043	2025-05-22 13:56:52 +00:00
Alex	de8eeab4ad	Merge pull request 'optionally support puny code (fix #273 )' (#1042 ) from trinity-1686a/garage:1686a/punnycode into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1042	2025-05-22 12:49:10 +00:00
Renjaya Raga Zenta	ae3f7ee76c	api: lifecycle: 404 if missing lifecycle config	2025-05-22 19:33:54 +07:00
trinity-1686a	2dc3a6dbbe	document allow_punycode configuration option	2025-05-22 14:08:06 +02:00
Armaël Guéneau	c6bc3f229b	Fix behavior of CopyObject wrt x-amz-website-redirect-location	2025-05-22 14:03:11 +02:00
trinity-1686a	bba9202f31	add test for punycode	2025-05-19 20:36:03 +02:00
trinity-1686a	a605a80806	support punnycode in api/web endpoint	2025-05-19 18:11:55 +02:00
trinity-1686a	539af12d21	allow punnycode in bucket name	2025-05-19 18:07:04 +02:00
Alex	a2a9e3cec4	Merge pull request 'doc: Add systemd example to increase file descriptors limit' (#1023 ) from baptiste/garage:systemd_openfiles into main Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1023	2025-05-09 10:34:07 +00:00
Baptiste Jonglez	14274bc13c	doc: Add systemd example to increase file descriptors limit	2025-05-08 10:27:53 +02:00
Quentin Dufour	8b35a946d9	Allow external HTTP client	2024-02-23 17:09:47 +01:00