feat: replace jemalloc with mimalloc (#4174)

* feat: replace jemalloc with mimalloc

* docs: record allocator rounds5 retest

* fix(replication): satisfy clippy unwrap lints

* docs: keep allocator migration plan local only

* feat(profiling): rely on pyroscope cpu profiling

* refactor(profiling): centralize unsupported pprof responses

* chore(deps): update s3s revision
This commit is contained in:
houseme
2026-07-02 19:03:38 +08:00
committed by GitHub
parent 54496f2796
commit 70e1d79dfd
17 changed files with 217 additions and 1529 deletions
Generated
+72 -480
View File
@@ -1659,12 +1659,6 @@ version = "3.20.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "72f5acc6cb2ba439de613abc23857ec3d78374d8ed5ac84e9d11336e87da8649"
[[package]]
name = "bytemuck"
version = "1.25.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c8efb64bd706a16a1bdde310ae86b351e4d21550d98d056f22f8a7f7a2183fec"
[[package]]
name = "byteorder"
version = "1.5.0"
@@ -2172,15 +2166,6 @@ version = "0.8.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "773648b94d0e5d620f64f280777445740e61fe701025087ec8b57f45c791888b"
[[package]]
name = "cpp_demangle"
version = "0.4.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f2bb79cb74d735044c972aae58ed0aaa9a837e85b01106a54c39e42e97f62253"
dependencies = [
"cfg-if",
]
[[package]]
name = "cpubits"
version = "0.1.1"
@@ -3619,7 +3604,7 @@ dependencies = [
"libc",
"option-ext",
"redox_users",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -3901,26 +3886,6 @@ dependencies = [
"syn 2.0.118",
]
[[package]]
name = "env_filter"
version = "2.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "900d271a03799a1ee8d1ca9b19893b48ca674a9284fefcfb85f05e74ed314217"
dependencies = [
"log",
"regex",
]
[[package]]
name = "env_logger"
version = "0.11.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "de671bd27a75a797dc9ae289ba1e77276e75e2026408aab65185384e2d5cd3f6"
dependencies = [
"env_filter",
"log",
]
[[package]]
name = "equator"
version = "0.4.2"
@@ -3963,7 +3928,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
dependencies = [
"libc",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -4901,15 +4866,6 @@ dependencies = [
"digest 0.11.3",
]
[[package]]
name = "home"
version = "0.5.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cc627f471c528ff0c4a49e1d5e60450c8f6461dd6d10ba9dcd3a61d3dff7728d"
dependencies = [
"windows-sys 0.61.2",
]
[[package]]
name = "hostname"
version = "0.4.2"
@@ -5246,46 +5202,6 @@ dependencies = [
"serde_core",
]
[[package]]
name = "inferno"
version = "0.11.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "232929e1d75fe899576a3d5c7416ad0d88dbfbb3c3d6aa00873a7408a50ddb88"
dependencies = [
"ahash",
"indexmap 2.14.0",
"is-terminal",
"itoa",
"log",
"num-format",
"once_cell",
"quick-xml 0.26.0",
"rgb",
"str_stack",
]
[[package]]
name = "inferno"
version = "0.12.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "90807d610575744524d9bdc69f3885d96f0e6c3354565b0828354a7ff2a262b8"
dependencies = [
"ahash",
"clap",
"crossbeam-channel",
"crossbeam-utils",
"dashmap",
"env_logger",
"indexmap 2.14.0",
"itoa",
"log",
"num-format",
"once_cell",
"quick-xml 0.39.4",
"rgb",
"str_stack",
]
[[package]]
name = "inout"
version = "0.1.4"
@@ -5373,17 +5289,6 @@ dependencies = [
"serde",
]
[[package]]
name = "is-terminal"
version = "0.4.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3640c1c38b8e4e43584d8df18be5fc6b0aa314ce6ebf51b53313d4306cca8e46"
dependencies = [
"hermit-abi",
"libc",
"windows-sys 0.61.2",
]
[[package]]
name = "is_debug"
version = "1.1.0"
@@ -5429,23 +5334,6 @@ version = "1.0.18"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8f42a60cbdf9a97f5d2305f08a87dc4e09308d1276d28c869c684d7777685682"
[[package]]
name = "jemalloc_pprof"
version = "0.8.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8a0d44c349cfe2654897fadcb9de4f0bfbf48288ec344f700b2bd59f152dd209"
dependencies = [
"anyhow",
"libc",
"mappings",
"once_cell",
"pprof_util",
"tempfile",
"tikv-jemalloc-ctl",
"tokio",
"tracing",
]
[[package]]
name = "jiff"
version = "0.2.31"
@@ -6022,19 +5910,6 @@ dependencies = [
"quote",
]
[[package]]
name = "mappings"
version = "0.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8bab1e61a4b76757edb59cd81fcaa7f3ba9018d43b527d9abfad877b4c6c60f2"
dependencies = [
"anyhow",
"libc",
"once_cell",
"pprof_util",
"tracing",
]
[[package]]
name = "matchers"
version = "0.2.0"
@@ -6088,15 +5963,6 @@ version = "2.8.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "88904434abc2901f197fe8cc55f0445e7ded921dba5911dad2e2b39b48e663c4"
[[package]]
name = "memmap2"
version = "0.5.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "83faa42c0a078c393f6b29d5db232d8be22776a891f8f56e5284faee4a20b327"
dependencies = [
"libc",
]
[[package]]
name = "memmap2"
version = "0.9.11"
@@ -6615,7 +6481,7 @@ version = "0.50.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -6627,20 +6493,6 @@ dependencies = [
"rand 0.8.6",
]
[[package]]
name = "num"
version = "0.4.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "35bd024e8b2ff75562e5f34e7f4905839deb4b22955ef5e73d2fea1b9813cb23"
dependencies = [
"num-bigint",
"num-complex",
"num-integer",
"num-iter",
"num-rational",
"num-traits",
]
[[package]]
name = "num-bigint"
version = "0.4.6"
@@ -6683,16 +6535,6 @@ version = "0.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "521739c6d2bac4aa25192232afe6841231376b2b26d4d9fae5ecf8ca5772e441"
[[package]]
name = "num-format"
version = "0.4.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a652d9771a63711fd3c3deb670acfbe5c30a4072e664d7a3bf5a9e1056ac72c3"
dependencies = [
"arrayvec",
"itoa",
]
[[package]]
name = "num-integer"
version = "0.1.46"
@@ -6713,17 +6555,6 @@ dependencies = [
"num-traits",
]
[[package]]
name = "num-rational"
version = "0.4.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
dependencies = [
"num-bigint",
"num-integer",
"num-traits",
]
[[package]]
name = "num-traits"
version = "0.2.19"
@@ -6788,7 +6619,7 @@ version = "5.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "51e219e79014df21a225b1860a479e2dcd7cbd9130f4defd4bd0e191ea31d67d"
dependencies = [
"base64 0.22.1",
"base64 0.21.7",
"chrono",
"getrandom 0.2.17",
"http 1.4.2",
@@ -6867,16 +6698,6 @@ dependencies = [
"objc2-core-foundation",
]
[[package]]
name = "object"
version = "0.29.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "21158b2c33aa6d4561f1c0a6ea283ca92bc54802a93b263e910746d679a7eb53"
dependencies = [
"flate2",
"memchr",
]
[[package]]
name = "object"
version = "0.37.3"
@@ -6886,6 +6707,17 @@ dependencies = [
"memchr",
]
[[package]]
name = "object"
version = "0.39.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2e5a6c098c7a3b6547378093f5cc30bc54fd361ce711e05293a5cc589562739b"
dependencies = [
"flate2",
"memchr",
"ruzstd",
]
[[package]]
name = "object_store"
version = "0.13.2"
@@ -7734,48 +7566,6 @@ version = "1.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "efca4c95a19a79d1c98f791f10aebd5c1363b473244630bb7dbde1dc98455a24"
[[package]]
name = "pprof-pyroscope-fork"
version = "0.1500.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "228f0967ab4c785d2daa6afdecd844a85a08227b19bfe14f1cf3f7819bf42f43"
dependencies = [
"aligned-vec",
"backtrace",
"cfg-if",
"findshlibs",
"framehop",
"inferno 0.11.21",
"libc",
"log",
"memmap2 0.5.10",
"nix 0.26.4",
"object 0.29.0",
"once_cell",
"protobuf",
"protobuf-codegen",
"smallvec",
"spin 0.10.0",
"symbolic-demangle",
"tempfile",
"thiserror 2.0.18",
]
[[package]]
name = "pprof_util"
version = "0.8.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eea0cc524de808a6d98d192a3d99fe95617031ad4a52ec0a0f987ef4432e8fe1"
dependencies = [
"anyhow",
"backtrace",
"flate2",
"inferno 0.12.6",
"num",
"paste",
"prost 0.14.4",
]
[[package]]
name = "ppv-lite86"
version = "0.2.21"
@@ -7959,7 +7749,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "be769465445e8c1474e9c5dac2018218498557af32d9ed057325ec9a41ae81bf"
dependencies = [
"heck",
"itertools 0.14.0",
"itertools 0.10.5",
"log",
"multimap",
"once_cell",
@@ -7979,7 +7769,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "03da047801ff44bb6a4d407d4860c05fd70bb81714e6b2f3812603d5b145b042"
dependencies = [
"heck",
"itertools 0.14.0",
"itertools 0.10.5",
"log",
"multimap",
"petgraph 0.8.3",
@@ -8000,7 +7790,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8a56d757972c98b346a9b766e3f02746cde6dd1cd1d1d563472929fdd74bec4d"
dependencies = [
"anyhow",
"itertools 0.14.0",
"itertools 0.10.5",
"proc-macro2",
"quote",
"syn 2.0.118",
@@ -8013,7 +7803,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b570b25f7617e43d59005d0990ccb79e950a423952cea19671b7a876da390adf"
dependencies = [
"anyhow",
"itertools 0.14.0",
"itertools 0.10.5",
"proc-macro2",
"quote",
"syn 2.0.118",
@@ -8037,57 +7827,6 @@ dependencies = [
"prost 0.14.4",
]
[[package]]
name = "protobuf"
version = "3.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d65a1d4ddae7d8b5de68153b48f6aa3bba8cb002b243dbdbc55a5afbc98f99f4"
dependencies = [
"once_cell",
"protobuf-support",
"thiserror 1.0.69",
]
[[package]]
name = "protobuf-codegen"
version = "3.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5d3976825c0014bbd2f3b34f0001876604fe87e0c86cd8fa54251530f1544ace"
dependencies = [
"anyhow",
"once_cell",
"protobuf",
"protobuf-parse",
"regex",
"tempfile",
"thiserror 1.0.69",
]
[[package]]
name = "protobuf-parse"
version = "3.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b4aeaa1f2460f1d348eeaeed86aea999ce98c1bded6f089ff8514c9d9dbdc973"
dependencies = [
"anyhow",
"indexmap 2.14.0",
"log",
"protobuf",
"protobuf-support",
"tempfile",
"thiserror 1.0.69",
"which",
]
[[package]]
name = "protobuf-support"
version = "3.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3e36c2f31e0a47f9280fb347ef5e461ffcd2c52dd520d8e216b52f93b0b0d7d6"
dependencies = [
"thiserror 1.0.69",
]
[[package]]
name = "proxy-protocol"
version = "0.5.0"
@@ -8161,19 +7900,29 @@ dependencies = [
[[package]]
name = "pyroscope"
version = "2.0.5"
version = "2.0.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "01ebd83db08e07a69b4291e7043f5f2ca964c8869e0bd0b6d09b8b062de775bd"
checksum = "b6869e0f2cae5cef10281c57d5a260f91372d7f190a7f717eb1a54f643d92151"
dependencies = [
"jemalloc_pprof",
"aligned-vec",
"backtrace",
"findshlibs",
"framehop",
"lazy_static",
"libc",
"libflate",
"log",
"pprof-pyroscope-fork",
"memmap2",
"nix 0.26.4",
"object 0.39.1",
"once_cell",
"prost 0.14.4",
"reqwest",
"serde_json",
"smallvec",
"spin 0.12.1",
"symbolic-demangle",
"tempfile",
"thiserror 2.0.18",
"url",
"uuid",
@@ -8206,34 +7955,6 @@ version = "1.2.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a1d01941d82fa2ab50be1e79e6714289dd7cde78eba4c074bc5a4374f650dfe0"
[[package]]
name = "quick-xml"
version = "0.26.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7f50b1c63b38611e7d4d7f68b82d3ad0cc71a2ad2e7f61fc10f1328d917c93cd"
dependencies = [
"memchr",
]
[[package]]
name = "quick-xml"
version = "0.39.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cdcc8dd4e2f670d309a5f0e83fe36dfdc05af317008fea29144da1a2ac858e5e"
dependencies = [
"memchr",
]
[[package]]
name = "quick-xml"
version = "0.40.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2474bd2e5029e7ccb6abb2ba48cf2383a333851dedf495901544281590c7da7f"
dependencies = [
"memchr",
"serde",
]
[[package]]
name = "quick-xml"
version = "0.41.0"
@@ -8299,7 +8020,7 @@ dependencies = [
"once_cell",
"socket2",
"tracing",
"windows-sys 0.60.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -8728,15 +8449,6 @@ dependencies = [
"subtle",
]
[[package]]
name = "rgb"
version = "0.8.53"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "47b34b781b31e5d73e9fbc8689c70551fd1ade9a19e3e28cfec8580a79290cc4"
dependencies = [
"bytemuck",
]
[[package]]
name = "ring"
version = "0.17.14"
@@ -9076,7 +8788,6 @@ dependencies = [
"hyper",
"hyper-util",
"insta",
"jemalloc_pprof",
"jiff",
"libc",
"libmimalloc-sys",
@@ -9090,7 +8801,6 @@ dependencies = [
"opentelemetry_sdk",
"percent-encoding",
"pin-project-lite",
"pprof-pyroscope-fork",
"proptest",
"rand 0.10.1",
"rcgen",
@@ -9153,8 +8863,6 @@ dependencies = [
"temp-env",
"tempfile",
"thiserror 2.0.18",
"tikv-jemalloc-ctl",
"tikv-jemallocator",
"time",
"tokio",
"tokio-rustls",
@@ -9333,7 +9041,7 @@ dependencies = [
"lazy_static",
"libc",
"md-5 0.11.0",
"memmap2 0.9.11",
"memmap2",
"metrics",
"moka",
"num_cpus",
@@ -9343,7 +9051,7 @@ dependencies = [
"path-absolutize",
"pin-project-lite",
"proptest",
"quick-xml 0.41.0",
"quick-xml",
"rand 0.10.1",
"ratelimit",
"rcgen",
@@ -9524,7 +9232,7 @@ name = "rustfs-io-core"
version = "1.0.0-beta.8"
dependencies = [
"bytes",
"memmap2 0.9.11",
"memmap2",
"rustfs-io-metrics",
"thiserror 2.0.18",
"tokio",
@@ -9696,7 +9404,7 @@ dependencies = [
"hashbrown 0.17.1",
"metrics",
"percent-encoding",
"quick-xml 0.41.0",
"quick-xml",
"rayon",
"rustc-hash",
"rustfs-config",
@@ -9745,7 +9453,6 @@ dependencies = [
"dial9-tokio-telemetry",
"flate2",
"glob",
"jemalloc_pprof",
"jiff",
"metrics",
"num_cpus",
@@ -9835,7 +9542,7 @@ dependencies = [
"md5",
"percent-encoding",
"proptest",
"quick-xml 0.41.0",
"quick-xml",
"regex",
"russh",
"russh-sftp",
@@ -10303,7 +10010,7 @@ dependencies = [
"errno",
"libc",
"linux-raw-sys 0.12.1",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -10377,7 +10084,7 @@ dependencies = [
"security-framework",
"security-framework-sys",
"webpki-root-certs",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -10416,6 +10123,15 @@ dependencies = [
"wait-timeout",
]
[[package]]
name = "ruzstd"
version = "0.8.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a7c1c839d570d835527c9a5e4db7cb2198683a988cb9d7293fc8674e6bd58fc8"
dependencies = [
"twox-hash",
]
[[package]]
name = "ryu"
version = "1.0.23"
@@ -10425,7 +10141,7 @@ checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f"
[[package]]
name = "s3s"
version = "0.14.0"
source = "git+https://github.com/rustfs/s3s?rev=300c657fb895b87e01e30b06e41a7996e57191f9#300c657fb895b87e01e30b06e41a7996e57191f9"
source = "git+https://github.com/rustfs/s3s?rev=921e9c5911951ae9efb91dc237e11be9f2f10a14#921e9c5911951ae9efb91dc237e11be9f2f10a14"
dependencies = [
"arc-swap",
"arrayvec",
@@ -10452,7 +10168,7 @@ dependencies = [
"nom 8.0.0",
"numeric_cast",
"pin-project-lite",
"quick-xml 0.40.1",
"quick-xml",
"serde",
"serde_json",
"serde_urlencoded",
@@ -11142,15 +10858,15 @@ name = "spin"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d5fe4ccb98d9c292d56fec89a5e07da7fc4cf0dc11e156b41793132775d3e591"
dependencies = [
"lock_api",
]
[[package]]
name = "spin"
version = "0.12.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bd5231412d905519dca6a5deb0327d407be68d6c941feec004533401d3a0a715"
dependencies = [
"lock_api",
]
[[package]]
name = "spki"
@@ -11316,12 +11032,6 @@ dependencies = [
"regex",
]
[[package]]
name = "str_stack"
version = "0.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7f446288b699d66d0fd2e30d1cfe7869194312524b3b9252594868ed26ef056a"
[[package]]
name = "stringprep"
version = "0.1.5"
@@ -11404,23 +11114,22 @@ dependencies = [
[[package]]
name = "symbolic-common"
version = "12.18.3"
version = "13.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "332615d90111d8eeaf86a84dc9bbe9f65d0d8c5cf11b4caccedc37754eb0dcfd"
checksum = "82bfbe03a011e4d3e2a94f03e054c33b101767f6905ec2a608dd85aa0cd71c97"
dependencies = [
"debugid",
"memmap2 0.9.11",
"memmap2",
"stable_deref_trait",
"uuid",
]
[[package]]
name = "symbolic-demangle"
version = "12.18.3"
version = "13.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "912017718eb4d21930546245af9a3475c9dccf15675a5c215664e76621afc471"
checksum = "f9d2ed98c0fd513dd4788554e1a62d9736f3642f8aaf7923644ec647158c1ded"
dependencies = [
"cpp_demangle",
"rustc-demangle",
"symbolic-common",
]
@@ -11561,7 +11270,7 @@ dependencies = [
"getrandom 0.4.3",
"once_cell",
"rustix 1.1.4",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -11655,37 +11364,6 @@ dependencies = [
"cfg-if",
]
[[package]]
name = "tikv-jemalloc-ctl"
version = "0.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "661f1f6a57b3a36dc9174a2c10f19513b4866816e13425d3e418b11cc37bc24c"
dependencies = [
"libc",
"paste",
"tikv-jemalloc-sys",
]
[[package]]
name = "tikv-jemalloc-sys"
version = "0.6.1+5.3.0-1-ge13ca993e8ccb9ba9847cc330696e02839f328f7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cd8aa5b2ab86a2cefa406d889139c162cbb230092f7d1d7cbc1716405d852a3b"
dependencies = [
"cc",
"libc",
]
[[package]]
name = "tikv-jemallocator"
version = "0.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0359b4327f954e0567e69fb191cf1436617748813819c94b8cd4a431422d053a"
dependencies = [
"libc",
"tikv-jemalloc-sys",
]
[[package]]
name = "time"
version = "0.3.53"
@@ -12625,18 +12303,6 @@ dependencies = [
"rustls-pki-types",
]
[[package]]
name = "which"
version = "4.4.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "87ba24419a2078cd2b0f2ede2691b6c66d8e47836da3b6db8265ebad47afbfc7"
dependencies = [
"either",
"home",
"once_cell",
"rustix 0.38.44",
]
[[package]]
name = "whoami"
version = "2.1.2"
@@ -12687,7 +12353,7 @@ version = "0.1.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@@ -12814,7 +12480,7 @@ version = "0.52.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "282be5f36a8ce781fad8c8ae18fa3f9beff57ec1b52cb3de0789201425d9a33d"
dependencies = [
"windows-targets 0.52.6",
"windows-targets",
]
[[package]]
@@ -12823,16 +12489,7 @@ version = "0.59.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1e38bc4d79ed67fd075bcc251a1c39b32a1776bbe92e5bef1f0bf1f8c531853b"
dependencies = [
"windows-targets 0.52.6",
]
[[package]]
name = "windows-sys"
version = "0.60.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f2f500e4d28234f72040990ec9d39e3a6b950f9f22d3dba18416c35882612bcb"
dependencies = [
"windows-targets 0.53.5",
"windows-targets",
]
[[package]]
@@ -12850,31 +12507,14 @@ version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9b724f72796e036ab90c1021d4780d4d3d648aca59e491e6b98e725b84e99973"
dependencies = [
"windows_aarch64_gnullvm 0.52.6",
"windows_aarch64_msvc 0.52.6",
"windows_i686_gnu 0.52.6",
"windows_i686_gnullvm 0.52.6",
"windows_i686_msvc 0.52.6",
"windows_x86_64_gnu 0.52.6",
"windows_x86_64_gnullvm 0.52.6",
"windows_x86_64_msvc 0.52.6",
]
[[package]]
name = "windows-targets"
version = "0.53.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4945f9f551b88e0d65f3db0bc25c33b8acea4d9e41163edf90dcd0b19f9069f3"
dependencies = [
"windows-link",
"windows_aarch64_gnullvm 0.53.1",
"windows_aarch64_msvc 0.53.1",
"windows_i686_gnu 0.53.1",
"windows_i686_gnullvm 0.53.1",
"windows_i686_msvc 0.53.1",
"windows_x86_64_gnu 0.53.1",
"windows_x86_64_gnullvm 0.53.1",
"windows_x86_64_msvc 0.53.1",
"windows_aarch64_gnullvm",
"windows_aarch64_msvc",
"windows_i686_gnu",
"windows_i686_gnullvm",
"windows_i686_msvc",
"windows_x86_64_gnu",
"windows_x86_64_gnullvm",
"windows_x86_64_msvc",
]
[[package]]
@@ -12892,96 +12532,48 @@ version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "32a4622180e7a0ec044bb555404c800bc9fd9ec262ec147edd5989ccd0c02cd3"
[[package]]
name = "windows_aarch64_gnullvm"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a9d8416fa8b42f5c947f8482c43e7d89e73a173cead56d044f6a56104a6d1b53"
[[package]]
name = "windows_aarch64_msvc"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "09ec2a7bb152e2252b53fa7803150007879548bc709c039df7627cabbd05d469"
[[package]]
name = "windows_aarch64_msvc"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b9d782e804c2f632e395708e99a94275910eb9100b2114651e04744e9b125006"
[[package]]
name = "windows_i686_gnu"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8e9b5ad5ab802e97eb8e295ac6720e509ee4c243f69d781394014ebfe8bbfa0b"
[[package]]
name = "windows_i686_gnu"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "960e6da069d81e09becb0ca57a65220ddff016ff2d6af6a223cf372a506593a3"
[[package]]
name = "windows_i686_gnullvm"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0eee52d38c090b3caa76c563b86c3a4bd71ef1a819287c19d586d7334ae8ed66"
[[package]]
name = "windows_i686_gnullvm"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fa7359d10048f68ab8b09fa71c3daccfb0e9b559aed648a8f95469c27057180c"
[[package]]
name = "windows_i686_msvc"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "240948bc05c5e7c6dabba28bf89d89ffce3e303022809e73deaefe4f6ec56c66"
[[package]]
name = "windows_i686_msvc"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1e7ac75179f18232fe9c285163565a57ef8d3c89254a30685b57d83a38d326c2"
[[package]]
name = "windows_x86_64_gnu"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "147a5c80aabfbf0c7d901cb5895d1de30ef2907eb21fbbab29ca94c5b08b1a78"
[[package]]
name = "windows_x86_64_gnu"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9c3842cdd74a865a8066ab39c8a7a473c0778a3f29370b5fd6b4b9aa7df4a499"
[[package]]
name = "windows_x86_64_gnullvm"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "24d5b23dc417412679681396f2b49f3de8c1473deb516bd34410872eff51ed0d"
[[package]]
name = "windows_x86_64_gnullvm"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0ffa179e2d07eee8ad8f57493436566c7cc30ac536a3379fdf008f47f6bb7ae1"
[[package]]
name = "windows_x86_64_msvc"
version = "0.52.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "589f6da84c646204747d1270a2a5661ea66ed1cced2631d546fdfb155959f9ec"
[[package]]
name = "windows_x86_64_msvc"
version = "0.53.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d6bbff5f0aada427a1e5a6da5f1f98158182f26556f345ac9e04d36d0ebed650"
[[package]]
name = "winnow"
version = "1.0.3"
+2 -14
View File
@@ -279,7 +279,7 @@ redis = { version = "1.3.0", features = ["connection-manager", "tokio-rustls-com
rustix = { version = "1.1.4", features = ["fs"] }
rust-embed = { version = "8.11.0" }
rustc-hash = { version = "2.1.2" }
s3s = { git = "https://github.com/rustfs/s3s", rev = "300c657fb895b87e01e30b06e41a7996e57191f9", features = ["minio"] }
s3s = { git = "https://github.com/rustfs/s3s", rev = "921e9c5911951ae9efb91dc237e11be9f2f10a14", features = ["minio"] }
serial_test = "3.5.0"
shadow-rs = { version = "2.0.0", default-features = false }
siphasher = "1.0.3"
@@ -320,7 +320,7 @@ opentelemetry-otlp = { version = "0.32.0", features = ["gzip-http", "reqwest-rus
opentelemetry_sdk = { version = "0.32.1", features = ["rt-tokio"] }
opentelemetry-semantic-conventions = { version = "0.32.1", features = ["semconv_experimental"] }
opentelemetry-stdout = { version = "0.32.0" }
pyroscope = { version = "=2.0.5", features = ["backend-pprof-rs", "backend-jemalloc"] }
pyroscope = { version = "2.0.6", features = ["backend-pprof-rs"] }
# FTP and SFTP
libunftp = { version = "0.23.0", features = ["experimental"] }
@@ -335,20 +335,8 @@ dav-server = "0.11.0"
# Performance Analysis and Memory Profiling
mimalloc = "0.1"
# Use tikv-jemallocator as memory allocator and enable performance analysis
tikv-jemallocator = { version = "0.6", features = ["profiling", "stats", "unprefixed_malloc_on_supported_platforms", "background_threads"] }
# Used to control and obtain statistics for jemalloc at runtime
tikv-jemalloc-ctl = { version = "0.6", features = ["use_std", "stats", "profiling"] }
# Snapshot testing for output format regression detection
insta = { version = "1.48", features = ["yaml", "json"] }
# Used to generate pprof-compatible memory profiling data and support symbolization and flame graphs
jemalloc_pprof = { version = "0.8.2", features = ["symbolize", "flamegraph"] }
# Used to generate CPU performance analysis data and flame diagrams
# pprof = { version = "0.15.0", features = ["flamegraph", "protobuf-codec"] }
# Use pprof-pyroscope-fork to match pyroscope 2.0.5's internal pprof dependency (v0.1500.4).
# Cargo unifies them into a single instance, avoiding duplicate perf_signal_handler symbols.
# NOTE: Do NOT upgrade pyroscope to >=2.0.6 — it vendors pprof-rs internally, causing symbol collision.
pprof = { package = "pprof-pyroscope-fork", version = "0.1500.4", features = ["flamegraph", "protobuf-codec"] }
[workspace.metadata.cargo-shear]
ignored = ["rustfs"]
+5 -4
View File
@@ -15,7 +15,8 @@
/// Profiler related environment variable names and default values
pub const ENV_ENABLE_PROFILING: &str = "RUSTFS_ENABLE_PROFILING";
// CPU profiling
// Legacy local CPU pprof controls. Local pprof dumps are currently unsupported;
// Pyroscope export is configured through observability settings.
pub const ENV_CPU_MODE: &str = "RUSTFS_PROF_CPU_MODE"; // off|continuous|periodic
/// Frequency of CPU profiling samples
pub const ENV_CPU_FREQ: &str = "RUSTFS_PROF_CPU_FREQ";
@@ -24,7 +25,7 @@ pub const ENV_CPU_INTERVAL_SECS: &str = "RUSTFS_PROF_CPU_INTERVAL_SECS";
/// Duration of each CPU profiling session (for periodic mode)
pub const ENV_CPU_DURATION_SECS: &str = "RUSTFS_PROF_CPU_DURATION_SECS";
/// Memory profiling (jemalloc)
/// Periodic memory pprof dumps (currently unsupported with mimalloc)
pub const ENV_MEM_PERIODIC: &str = "RUSTFS_PROF_MEM_PERIODIC";
/// Interval between memory profiling snapshots (for periodic mode)
pub const ENV_MEM_INTERVAL_SECS: &str = "RUSTFS_PROF_MEM_INTERVAL_SECS";
@@ -34,7 +35,7 @@ pub const ENV_OUTPUT_DIR: &str = "RUSTFS_PROF_OUTPUT_DIR";
/// Defaults for profiler settings
pub const DEFAULT_ENABLE_PROFILING: bool = false;
/// CPU profiling
/// Legacy local CPU pprof mode
pub const DEFAULT_CPU_MODE: &str = "off";
/// Frequency of CPU profiling samples
pub const DEFAULT_CPU_FREQ: usize = 100;
@@ -42,7 +43,7 @@ pub const DEFAULT_CPU_FREQ: usize = 100;
pub const DEFAULT_CPU_INTERVAL_SECS: u64 = 300;
/// Duration of each CPU profiling session (for periodic mode)
pub const DEFAULT_CPU_DURATION_SECS: u64 = 60;
/// Memory profiling (jemalloc)
/// Periodic memory pprof dumps (currently unsupported with mimalloc)
pub const DEFAULT_MEM_PERIODIC: bool = false;
/// Interval between memory profiling snapshots (for periodic mode)
pub const DEFAULT_MEM_INTERVAL_SECS: u64 = 300;
+19 -19
View File
@@ -782,7 +782,7 @@ mod tests {
let contract = OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![
profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true),
profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true),
profiler_backend("memory_pprof", OpsProfilerBackendStatus::Disabled, false),
profiler_backend("ebpf", OpsProfilerBackendStatus::Unsupported, false),
profiler_backend("future_kernel_profiler", OpsProfilerBackendStatus::Unknown, false),
@@ -803,7 +803,7 @@ mod tests {
fn ops_profiler_schema_serializes_stable_json_shape() {
let contract = OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
};
let value = serde_json::to_value(contract).expect("ops profiler schema should serialize");
@@ -813,7 +813,7 @@ mod tests {
json!({
"mode": "capability_description",
"backends": [{
"backend": "cpu_pprof",
"backend": "pyroscope",
"status": "enabled",
"supports_profile_export": true,
"redaction_required": ["secret", "token", "local_path", "host"],
@@ -860,7 +860,7 @@ mod tests {
contract: OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![
profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true),
profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true),
profiler_backend("memory_pprof", OpsProfilerBackendStatus::Disabled, false),
profiler_backend("ebpf", OpsProfilerBackendStatus::Unsupported, false),
],
@@ -898,7 +898,7 @@ mod tests {
},
contract: OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
},
};
@@ -916,7 +916,7 @@ mod tests {
"contract": {
"mode": "capability_description",
"backends": [{
"backend": "cpu_pprof",
"backend": "pyroscope",
"status": "enabled",
"supports_profile_export": true,
"redaction_required": ["secret", "token", "local_path", "host"],
@@ -942,7 +942,7 @@ mod tests {
},
contract: OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
},
};
@@ -976,7 +976,7 @@ mod tests {
fn rejects_ops_profiler_execution_requests_and_empty_backends() {
let mut contract = OpsProfilerContract {
mode: OpsProfilerContractMode::ExecutionRequest,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
};
assert_eq!(
@@ -998,7 +998,7 @@ mod tests {
let err = serde_json::from_value::<OpsProfilerContract>(json!({
"mode": "capability_description",
"backends": [{
"backend": "cpu_pprof",
"backend": "pyroscope",
"status": "enabled",
"supports_profile_export": true,
"redaction_required": ["local_path"]
@@ -1014,7 +1014,7 @@ mod tests {
let err = serde_json::from_value::<OpsProfilerContract>(json!({
"mode": "capability_description",
"backends": [{
"backend": "cpu_pprof",
"backend": "pyroscope",
"status": "enabled",
"supports_profile_export": true,
"redaction_required": ["local_path"],
@@ -1035,7 +1035,7 @@ mod tests {
fn rejects_ops_profiler_invalid_backend_and_redaction_shape() {
let mut contract = OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
};
contract.backends[0].backend = OpsProfilerBackendName::new(" ");
@@ -1044,15 +1044,15 @@ mod tests {
ExtensionContractError::EmptyOpsProfilerBackend
);
contract.backends[0] = profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true);
contract.backends[0] = profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true);
contract
.backends
.push(profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Disabled, false));
.push(profiler_backend("pyroscope", OpsProfilerBackendStatus::Disabled, false));
assert_eq!(
validate_ops_profiler_contract(&contract).expect_err("duplicate backends should fail validation"),
ExtensionContractError::DuplicateOpsProfilerBackend {
backend: "cpu_pprof".to_string()
backend: "pyroscope".to_string()
}
);
@@ -1062,7 +1062,7 @@ mod tests {
assert_eq!(
validate_ops_profiler_contract(&contract).expect_err("duplicate redaction fields should fail validation"),
ExtensionContractError::DuplicateOpsProfilerRedactionField {
backend: "cpu_pprof".to_string(),
backend: "pyroscope".to_string(),
field: OpsProfilerRedactionField::Host
}
);
@@ -1072,7 +1072,7 @@ mod tests {
assert_eq!(
validate_ops_profiler_contract(&contract).expect_err("profile export requires local path redaction"),
ExtensionContractError::OpsProfilerMissingLocalPathRedaction {
backend: "cpu_pprof".to_string()
backend: "pyroscope".to_string()
}
);
}
@@ -1081,14 +1081,14 @@ mod tests {
fn rejects_ops_profiler_missing_provenance_boundary() {
let mut contract = OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true)],
backends: vec![profiler_backend("pyroscope", OpsProfilerBackendStatus::Enabled, true)],
};
contract.backends[0].provenance.source = " ".to_string();
assert_eq!(
validate_ops_profiler_contract(&contract).expect_err("provenance source should be required"),
ExtensionContractError::EmptyOpsProfilerProvenanceSource {
backend: "cpu_pprof".to_string()
backend: "pyroscope".to_string()
}
);
@@ -1098,7 +1098,7 @@ mod tests {
assert_eq!(
validate_ops_profiler_contract(&contract).expect_err("collection boundary should be required"),
ExtensionContractError::EmptyOpsProfilerCollectionBoundary {
backend: "cpu_pprof".to_string()
backend: "pyroscope".to_string()
}
);
}
+1 -2
View File
@@ -28,7 +28,7 @@ documentation = "https://docs.rs/rustfs-obs/latest/rustfs_obs/"
[features]
default = []
gpu = ["dep:nvml-wrapper"]
pyroscope = ["dep:jemalloc_pprof", "dep:pyroscope"]
pyroscope = ["dep:pyroscope"]
[lints]
workspace = true
@@ -83,7 +83,6 @@ nvml-wrapper = { workspace = true, optional = true }
[target.'cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))'.dependencies]
pyroscope = { workspace = true, features = ["backend-pprof-rs"], optional = true }
jemalloc_pprof = { workspace = true, optional = true }
[dev-dependencies]
-30
View File
@@ -39,10 +39,6 @@ pub(crate) type ProfilingAgent = pyroscope::PyroscopeAgent<pyroscope::pyroscope:
any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))
)))]
pub(crate) type ProfilingAgent = ();
#[cfg(all(feature = "pyroscope", target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
pub(crate) type MemoryProfilingAgent = pyroscope::PyroscopeAgent<pyroscope::pyroscope::PyroscopeAgentRunning>;
#[cfg(not(all(feature = "pyroscope", target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
pub(crate) type MemoryProfilingAgent = ();
const LOG_COMPONENT_OBS: &str = "obs";
const LOG_SUBSYSTEM_GUARD: &str = "guard";
@@ -67,7 +63,6 @@ pub struct OtelGuard {
/// Optional logger provider for OTLP log export.
pub(crate) logger_provider: Option<SdkLoggerProvider>,
pub(crate) profiling_agent: Option<ProfilingAgent>,
pub(crate) memory_profiling_agent: Option<MemoryProfilingAgent>,
/// Handle to the background log-cleanup task; aborted on drop.
pub(crate) cleanup_handle: Option<tokio::task::JoinHandle<()>>,
/// Worker guard that keeps the non-blocking `tracing_appender` thread
@@ -84,7 +79,6 @@ impl std::fmt::Debug for OtelGuard {
.field("meter_provider", &self.meter_provider.is_some())
.field("logger_provider", &self.logger_provider.is_some())
.field("profiling_agent", &self.profiling_agent.is_some())
.field("memory_profiling_agent", &self.memory_profiling_agent.is_some())
.field("cleanup_handle", &self.cleanup_handle.is_some())
.field("tracing_guard", &self.tracing_guard.is_some())
.field("stdout_guard", &self.stdout_guard.is_some())
@@ -161,30 +155,6 @@ impl Drop for OtelGuard {
}
}
#[cfg(all(feature = "pyroscope", target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
if let Some(agent) = self.memory_profiling_agent.take() {
match agent.stop() {
Err(err) => {
if tracing::dispatcher::has_been_set() {
error!(
event = EVENT_OBS_GUARD_SHUTDOWN,
component = LOG_COMPONENT_OBS,
subsystem = LOG_SUBSYSTEM_GUARD,
resource = "memory_profiling_agent",
result = "shutdown_failed",
error = %err,
"observability guard shutdown failed"
);
} else {
eprintln!("{}", format_guard_shutdown_stderr_message("memory_profiling_agent", err));
}
}
Ok(stopped) => {
stopped.shutdown();
}
}
}
if let Some(handle) = self.cleanup_handle.take() {
debug!(
event = EVENT_OBS_GUARD_SHUTDOWN,
-2
View File
@@ -271,7 +271,6 @@ fn init_stdout_only(_config: &OtelConfig, logger_level: &str, is_production: boo
meter_provider: None,
logger_provider: None,
profiling_agent: None,
memory_profiling_agent: None,
tracing_guard: Some(guard),
stdout_guard: None,
cleanup_handle: None,
@@ -389,7 +388,6 @@ fn init_file_logging_internal(
meter_provider: None,
logger_provider: None,
profiling_agent: None,
memory_profiling_agent: None,
tracing_guard: Some(guard),
stdout_guard,
cleanup_handle: Some(cleanup_handle),
+1 -62
View File
@@ -40,7 +40,7 @@ use crate::cleaner::types::FileMatchMode;
use crate::config::OtelConfig;
use crate::global::set_observability_metric_enabled;
use crate::telemetry::filter::build_env_filter;
use crate::telemetry::guard::{MemoryProfilingAgent, OtelGuard, ProfilingAgent};
use crate::telemetry::guard::{OtelGuard, ProfilingAgent};
use crate::telemetry::local::{build_json_log_layer, spawn_cleanup_task};
use crate::telemetry::recorder::Recorder;
use crate::telemetry::resource::build_resource;
@@ -173,7 +173,6 @@ pub(super) fn init_observability_http(
let meter_provider = build_meter_provider(&metric_ep, config, res.clone(), &service_name, use_stdout)?;
let profiling_agent = init_profiler(config);
let memory_profiling_agent = init_memory_profiler(config);
// ── Logger Logic ──────────────────────────────────────────────────────────
// Logging is the only signal that may intentionally route to either OTLP
@@ -317,7 +316,6 @@ pub(super) fn init_observability_http(
meter_provider,
logger_provider,
profiling_agent,
memory_profiling_agent,
tracing_guard,
stdout_guard,
cleanup_handle,
@@ -563,65 +561,6 @@ fn init_profiler(_config: &OtelConfig) -> Option<ProfilingAgent> {
None
}
/// Initialise a Pyroscope agent for continuous **memory** profiling via jemalloc.
///
/// This is only available on `linux + gnu + x86_64` where tikv-jemallocator
/// is the global allocator and `jemalloc_pprof::PROF_CTL` is accessible.
///
/// Returns `None` when profiling export is disabled, the endpoint is missing,
/// jemalloc profiling is not activated, or the agent fails to build/start.
#[cfg(all(feature = "pyroscope", target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
fn init_memory_profiler(config: &OtelConfig) -> Option<MemoryProfilingAgent> {
use pyroscope::backend::jemalloc_backend;
use pyroscope::pyroscope::PyroscopeAgentBuilder;
use rustfs_config::VERSION;
if !config
.profiling_export_enabled
.unwrap_or(rustfs_config::DEFAULT_OBS_PROFILING_EXPORT_ENABLED)
{
return None;
}
let endpoint = config.profiling_endpoint.as_ref()?.as_str();
if endpoint.is_empty() {
return None;
}
// Verify jemalloc profiling is available and activated
{
let prof_ctl = jemalloc_pprof::PROF_CTL.as_ref()?;
let ctl = prof_ctl.try_lock().ok()?;
if !ctl.activated() {
eprintln!("Memory profiling skipped: jemalloc profiling is not activated");
return None;
}
}
let backend = jemalloc_backend();
let service_name = config.service_name.as_deref().unwrap_or(APP_NAME);
let version = config.service_version.as_deref().unwrap_or(VERSION);
let sample_rate = 100;
let agent = PyroscopeAgentBuilder::new(endpoint, service_name, sample_rate, "pyroscope-rs", "1.0.1", backend)
.tags(vec![("version", version), ("profile_type", "memory")])
.build()
.ok()?;
match agent.start() {
Ok(agent) => Some(agent),
Err(err) => {
eprintln!("Memory profiling agent start error: {err:?}");
None
}
}
}
#[cfg(not(all(feature = "pyroscope", target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
fn init_memory_profiler(_config: &OtelConfig) -> Option<MemoryProfilingAgent> {
None
}
/// Create a stdout periodic metrics reader for the given interval.
///
/// This helper is primarily used for local development and diagnostics when
+4 -4
View File
@@ -123,8 +123,8 @@ pub fn builtin_ops_profiler_contract() -> OpsProfilerContract {
OpsProfilerContract {
mode: OpsProfilerContractMode::CapabilityDescription,
backends: vec![
builtin_ops_profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Enabled, true),
builtin_ops_profiler_backend("memory_pprof", OpsProfilerBackendStatus::Disabled, true),
builtin_ops_profiler_backend("cpu_pprof", OpsProfilerBackendStatus::Unsupported, false),
builtin_ops_profiler_backend("memory_pprof", OpsProfilerBackendStatus::Unsupported, false),
builtin_ops_profiler_backend("ebpf", OpsProfilerBackendStatus::Unsupported, false),
],
}
@@ -391,8 +391,8 @@ mod tests {
assert_eq!(
backends,
vec![
("cpu_pprof", OpsProfilerBackendStatus::Enabled, true),
("memory_pprof", OpsProfilerBackendStatus::Disabled, true),
("cpu_pprof", OpsProfilerBackendStatus::Unsupported, false),
("memory_pprof", OpsProfilerBackendStatus::Unsupported, false),
("ebpf", OpsProfilerBackendStatus::Unsupported, false),
]
);
+2 -2
View File
@@ -172,13 +172,13 @@ mod tests {
assert_eq!(registry.registered_backend_count(), contract.backends.len());
assert_eq!(
registry.registration_for("cpu_pprof").map(|registration| registration.status),
Some(OpsProfilerBackendStatus::Enabled)
Some(OpsProfilerBackendStatus::Unsupported)
);
assert_eq!(
registry
.registration_for("memory_pprof")
.map(|registration| registration.supports_profile_export),
Some(true)
Some(false)
);
assert_eq!(
registry.authorize_read(OpsProfilerReadRequest {
+2 -11
View File
@@ -186,25 +186,16 @@ opentelemetry = { workspace = true }
tracing-opentelemetry = { workspace = true }
# Data structures
hashbrown = { workspace = true }
mimalloc = { workspace = true }
[target.'cfg(target_os = "linux")'.dependencies]
libsystemd.workspace = true
# io-uring is Linux-only. Scope the feature to Linux targets.
tokio = { workspace = true, features = ["io-uring"] }
[target.'cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))'.dependencies]
mimalloc = { workspace = true }
[target.'cfg(not(target_os = "windows"))'.dependencies]
libmimalloc-sys = { version = "0.1.49", features = ["extended"] }
[target.'cfg(any(target_os = "linux", target_os = "macos"))'.dependencies]
pprof = { workspace = true }
# jemalloc-based allocator and memory profiling are only enabled on validated linux-gnu-x86_64 targets.
[target.'cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))'.dependencies]
tikv-jemallocator = { workspace = true }
tikv-jemalloc-ctl = { workspace = true }
jemalloc_pprof = { workspace = true }
[dev-dependencies]
uuid = { workspace = true, features = ["v4"] }
serial_test = { workspace = true }
+8 -47
View File
@@ -15,11 +15,7 @@
use crate::admin::{auth::validate_admin_request, router::Operation};
use crate::auth::{check_key_valid, get_session_token};
use crate::server::RemoteAddr;
#[cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
use http::HeaderMap;
use http::StatusCode;
#[cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
use http::header::CONTENT_TYPE;
use matchit::Params;
use rustfs_policy::policy::action::{Action, AdminAction};
use s3s::{Body, S3Request, S3Response, S3Result, s3_error};
@@ -45,6 +41,10 @@ pub(super) async fn authorize_profile_request(req: &S3Request<Body>) -> S3Result
.await
}
pub(super) fn profile_not_implemented_response(message: String) -> S3Response<(StatusCode, Body)> {
S3Response::new((StatusCode::NOT_IMPLEMENTED, Body::from(message)))
}
pub struct TriggerProfileCPU {}
#[async_trait::async_trait]
impl Operation for TriggerProfileCPU {
@@ -52,30 +52,8 @@ impl Operation for TriggerProfileCPU {
authorize_profile_request(&req).await?;
info!("Triggering CPU profile dump via S3 request...");
#[cfg(not(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))))]
{
return Ok(S3Response::new((
StatusCode::NOT_IMPLEMENTED,
Body::from(
crate::profiling::dump_cpu_pprof_for(std::time::Duration::from_secs(0))
.await
.unwrap_err(),
),
)));
}
#[cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
{
let dur = std::time::Duration::from_secs(60);
match crate::profiling::dump_cpu_pprof_for(dur).await {
Ok(path) => {
let mut header = HeaderMap::new();
header.insert(CONTENT_TYPE, "text/html".parse().expect("operation should succeed"));
Ok(S3Response::with_headers((StatusCode::OK, Body::from(path.display().to_string())), header))
}
Err(e) => Err(s3s::s3_error!(InternalError, "{}", format!("Failed to dump CPU profile: {e}"))),
}
}
crate::profiling::log_cpu_pprof_dump_skipped();
Ok(profile_not_implemented_response(crate::profiling::local_cpu_pprof_unsupported_message()))
}
}
@@ -86,25 +64,8 @@ impl Operation for TriggerProfileMemory {
authorize_profile_request(&req).await?;
info!("Triggering Memory profile dump via S3 request...");
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
{
return Ok(S3Response::new((
StatusCode::NOT_IMPLEMENTED,
Body::from(crate::profiling::dump_memory_pprof_now().await.unwrap_err()),
)));
}
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
{
match crate::profiling::dump_memory_pprof_now().await {
Ok(path) => {
let mut header = HeaderMap::new();
header.insert(CONTENT_TYPE, "text/html".parse().expect("operation should succeed"));
Ok(S3Response::with_headers((StatusCode::OK, Body::from(path.display().to_string())), header))
}
Err(e) => Err(s3s::s3_error!(InternalError, "{}", format!("Failed to dump Memory profile: {e}"))),
}
}
crate::profiling::log_memory_pprof_dump_skipped();
Ok(profile_not_implemented_response(crate::profiling::memory_pprof_unsupported_message()))
}
}
+21 -169
View File
@@ -12,28 +12,23 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use super::profile::authorize_profile_request;
use super::profile::{authorize_profile_request, profile_not_implemented_response};
use crate::admin::router::{AdminOperation, Operation, S3Router};
use crate::server::ADMIN_PREFIX;
use http::{HeaderMap, HeaderValue, Uri};
use http::{HeaderMap, HeaderValue};
use hyper::{Method, StatusCode};
use matchit::Params;
use s3s::header::CONTENT_TYPE;
use s3s::{Body, S3Request, S3Response, S3Result};
use std::collections::HashMap;
use serde::Serialize;
use tracing::error;
#[allow(dead_code)]
fn extract_query_params(uri: &Uri) -> HashMap<String, String> {
let mut params = HashMap::new();
if let Some(query) = uri.query() {
for (key, value) in url::form_urlencoded::parse(query.as_bytes()) {
params.insert(key.into_owned(), value.into_owned());
}
}
params
#[derive(Serialize)]
struct ProfileStatus {
enabled: &'static str,
status: &'static str,
platform: &'static str,
message: &'static str,
}
pub fn register_profiling_route(r: &mut S3Router<AdminOperation>) -> std::io::Result<()> {
@@ -54,114 +49,17 @@ pub fn register_profiling_route(r: &mut S3Router<AdminOperation>) -> std::io::Re
pub struct ProfileHandler {}
#[allow(dead_code)]
fn map_cpu_profile_collect_error_message(err: &str) -> (StatusCode, String) {
if err.contains("start running cpu profiler error") {
return (
StatusCode::CONFLICT,
"CPU profiler is already running. Disable RUSTFS_OBS_PROFILING_EXPORT_ENABLED or retry later.".to_string(),
);
}
(StatusCode::INTERNAL_SERVER_ERROR, format!("Failed to collect CPU profile: {err}"))
}
#[async_trait::async_trait]
impl Operation for ProfileHandler {
async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
authorize_profile_request(&req).await?;
#[cfg(not(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))))]
{
let requested_url = req.uri.to_string();
let target_os = std::env::consts::OS;
let target_arch = std::env::consts::ARCH;
let target_env = option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown");
let msg = format!(
"CPU profiling is not supported on this platform. target_os={target_os}, target_env={target_env}, target_arch={target_arch}, requested_url={requested_url}"
);
return Ok(S3Response::new((StatusCode::NOT_IMPLEMENTED, Body::from(msg))));
}
#[cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
{
use rustfs_config::{DEFAULT_CPU_FREQ, ENV_CPU_FREQ};
use rustfs_utils::get_env_usize;
let queries = extract_query_params(&req.uri);
let seconds = queries.get("seconds").and_then(|s| s.parse::<u64>().ok()).unwrap_or(30);
let format = queries.get("format").cloned().unwrap_or_else(|| "protobuf".to_string());
if seconds > 300 {
return Ok(S3Response::new((
StatusCode::BAD_REQUEST,
Body::from("Profile duration cannot exceed 300 seconds".to_string()),
)));
}
match format.as_str() {
"protobuf" | "pb" => match crate::profiling::dump_cpu_pprof_for(std::time::Duration::from_secs(seconds)).await {
Ok(path) => match tokio::fs::read(&path).await {
Ok(bytes) => {
let mut headers = HeaderMap::new();
headers.insert(CONTENT_TYPE, HeaderValue::from_static("application/octet-stream"));
Ok(S3Response::with_headers((StatusCode::OK, Body::from(bytes)), headers))
}
Err(e) => {
error!("Failed to read profile file {}: {}", path.display(), e);
Ok(S3Response::new((
StatusCode::INTERNAL_SERVER_ERROR,
Body::from(format!("Failed to read profile file: {e}")),
)))
}
},
Err(e) => {
let (status, message) = map_cpu_profile_collect_error_message(&e);
error!("CPU protobuf profile collection failed: {}", e);
Ok(S3Response::new((status, Body::from(message))))
}
},
"flamegraph" | "svg" => {
let freq = get_env_usize(ENV_CPU_FREQ, DEFAULT_CPU_FREQ) as i32;
let guard = match pprof::ProfilerGuard::new(freq) {
Ok(g) => g,
Err(e) => {
return Ok(S3Response::new((
StatusCode::INTERNAL_SERVER_ERROR,
Body::from(format!("Failed to create profiler: {e}")),
)));
}
};
tokio::time::sleep(std::time::Duration::from_secs(seconds)).await;
let report = match guard.report().build() {
Ok(r) => r,
Err(e) => {
return Ok(S3Response::new((
StatusCode::INTERNAL_SERVER_ERROR,
Body::from(format!("Failed to build profile report: {e}")),
)));
}
};
let mut flamegraph_buf = Vec::new();
if let Err(e) = report.flamegraph(&mut flamegraph_buf) {
return Ok(S3Response::new((
StatusCode::INTERNAL_SERVER_ERROR,
Body::from(format!("Failed to generate flamegraph: {e}")),
)));
}
let mut headers = HeaderMap::new();
headers.insert(CONTENT_TYPE, HeaderValue::from_static("image/svg+xml"));
Ok(S3Response::with_headers((StatusCode::OK, Body::from(flamegraph_buf)), headers))
}
_ => Ok(S3Response::new((
StatusCode::BAD_REQUEST,
Body::from("Unsupported format. Use 'protobuf' or 'flamegraph'".to_string()),
))),
}
}
let requested_url = req.uri.to_string();
crate::profiling::log_cpu_pprof_dump_skipped();
Ok(profile_not_implemented_response(format!(
"{}; requested_url={requested_url}",
crate::profiling::local_cpu_pprof_unsupported_message()
)))
}
}
@@ -172,37 +70,11 @@ impl Operation for ProfileStatusHandler {
async fn call(&self, req: S3Request<Body>, _params: Params<'_, '_>) -> S3Result<S3Response<(StatusCode, Body)>> {
authorize_profile_request(&req).await?;
#[cfg(not(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))))]
let message = format!("CPU profiling is not supported on {} platform", std::env::consts::OS);
#[cfg(not(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))))]
let status = HashMap::from([
("enabled", "false"),
("status", "not_supported"),
("platform", std::env::consts::OS),
("message", message.as_str()),
]);
#[cfg(any(target_os = "macos", all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
let status = {
use rustfs_config::{DEFAULT_ENABLE_PROFILING, ENV_ENABLE_PROFILING};
use rustfs_utils::get_env_bool;
let enabled = get_env_bool(ENV_ENABLE_PROFILING, DEFAULT_ENABLE_PROFILING);
if enabled {
HashMap::from([
("enabled", "true"),
("status", "running"),
("supported_formats", "protobuf, flamegraph"),
("max_duration_seconds", "300"),
("endpoint", "/rustfs/admin/debug/pprof/profile"),
])
} else {
HashMap::from([
("enabled", "false"),
("status", "disabled"),
("message", "Set RUSTFS_ENABLE_PROFILING=true to enable profiling"),
])
}
let status = ProfileStatus {
enabled: "false",
status: "not_supported",
platform: std::env::consts::OS,
message: crate::profiling::LOCAL_CPU_PPROF_UNSUPPORTED_SUMMARY,
};
match serde_json::to_string(&status) {
@@ -224,11 +96,10 @@ impl Operation for ProfileStatusHandler {
#[cfg(test)]
mod tests {
use super::{ProfileHandler, ProfileStatusHandler, extract_query_params};
use super::{ProfileHandler, ProfileStatusHandler};
use crate::admin::router::Operation;
use http::{Extensions, HeaderMap, Uri};
use hyper::Method;
use hyper::StatusCode;
use matchit::Params;
use s3s::{Body, S3ErrorCode, S3Request};
@@ -246,17 +117,6 @@ mod tests {
}
}
#[test]
fn test_extract_query_params_decodes_percent_encoded_values() {
let uri: Uri = "/rustfs/admin/debug/pprof/profile?format=flamegraph&note=a%2Bb+value"
.parse()
.expect("uri should parse");
let params = extract_query_params(&uri);
assert_eq!(params.get("format"), Some(&"flamegraph".to_string()));
assert_eq!(params.get("note"), Some(&"a+b value".to_string()));
}
#[tokio::test]
async fn profile_handler_rejects_missing_credentials() {
let result = ProfileHandler {}
@@ -284,12 +144,4 @@ mod tests {
assert_eq!(err.code(), &S3ErrorCode::AccessDenied);
assert_eq!(err.message(), Some("Signature is required"));
}
#[test]
fn cpu_profile_collect_error_maps_profiler_conflict_to_409() {
let (status, message) =
super::map_cpu_profile_collect_error_message("create profiler failed: start running cpu profiler error");
assert_eq!(status, StatusCode::CONFLICT);
assert!(message.contains("CPU profiler is already running"));
}
}
+5 -30
View File
@@ -111,15 +111,7 @@ impl AllocatorReclaimController {
}
pub fn allocator_backend() -> &'static str {
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
{
"jemalloc"
}
#[cfg(all(
not(target_os = "windows"),
not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))
))]
#[cfg(not(target_os = "windows"))]
{
"mimalloc"
}
@@ -206,7 +198,7 @@ fn configured_allocator_reclaim_interval_secs() -> u64 {
}
fn effective_allocator_reclaim_force(backend: &str, configured_force: bool) -> bool {
configured_force && backend != "jemalloc"
configured_force && backend != "mimalloc-windows"
}
fn build_allocator_reclaim_desired_snapshot(
@@ -302,10 +294,7 @@ pub fn allocator_reclaim_controller_snapshot(ctx: &CancellationToken) -> Allocat
)
}
#[cfg(all(
not(target_os = "windows"),
not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))
))]
#[cfg(not(target_os = "windows"))]
#[allow(unsafe_code)]
fn collect_allocator_memory(force: bool) -> Result<(), String> {
// SAFETY: `mi_collect` is provided by the active global allocator backend
@@ -317,13 +306,6 @@ fn collect_allocator_memory(force: bool) -> Result<(), String> {
Ok(())
}
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
fn collect_allocator_memory(_force: bool) -> Result<(), String> {
let _ = tikv_jemalloc_ctl::background_thread::write(true);
tikv_jemalloc_ctl::epoch::advance().map_err(|err| err.to_string())?;
Ok(())
}
#[cfg(target_os = "windows")]
fn collect_allocator_memory(_force: bool) -> Result<(), String> {
Err("allocator reclaim is not supported on Windows".to_string())
@@ -368,13 +350,6 @@ pub fn init_allocator_reclaim(ctx: CancellationToken) {
}
let configured_force = configured_allocator_reclaim_force();
if backend == "jemalloc" && configured_force {
warn!(
backend,
env = rustfs_config::ENV_ALLOCATOR_RECLAIM_FORCE,
"allocator reclaim force mode is not supported on jemalloc backend; ignoring configured force flag"
);
}
let force = effective_allocator_reclaim_force(backend, configured_force);
let idle_intervals = configured_allocator_reclaim_idle_intervals();
let interval = Duration::from_secs(configured_allocator_reclaim_interval_secs());
@@ -473,8 +448,8 @@ mod tests {
}
#[test]
fn allocator_reclaim_force_preserves_jemalloc_override() {
assert!(!effective_allocator_reclaim_force("jemalloc", true));
fn allocator_reclaim_force_is_disabled_only_on_windows_backend() {
assert!(!effective_allocator_reclaim_force("mimalloc-windows", true));
assert!(effective_allocator_reclaim_force("mimalloc", true));
assert!(!effective_allocator_reclaim_force("mimalloc", false));
}
-5
View File
@@ -12,11 +12,6 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
#[global_allocator]
static GLOBAL: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
#[global_allocator]
static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
+73 -638
View File
@@ -12,647 +12,82 @@
// See the License for the specific language governing permissions and
// limitations under the License.
#[cfg(not(any(target_os = "linux", target_os = "macos")))]
mod unsupported_impl {
use std::path::PathBuf;
use std::time::Duration;
use tracing::{debug, info};
use tracing::{debug, info};
const LOG_COMPONENT_PROFILING: &str = "profiling";
const LOG_SUBSYSTEM_PLATFORM: &str = "platform";
const LOG_COMPONENT_PROFILING: &str = "profiling";
const LOG_SUBSYSTEM_CPU: &str = "cpu";
const LOG_SUBSYSTEM_MEMORY: &str = "memory";
const LOG_SUBSYSTEM_RUNTIME: &str = "runtime";
const LOCAL_CPU_PPROF_UNSUPPORTED_REASON: &str = "local_cpu_pprof_unsupported";
const MEMORY_PPROF_UNSUPPORTED_REASON: &str = "mimalloc_memory_pprof_unsupported";
pub const LOCAL_CPU_PPROF_UNSUPPORTED_SUMMARY: &str = "local CPU pprof dumps are not supported; use Pyroscope export instead";
pub const MEMORY_PPROF_UNSUPPORTED_SUMMARY: &str = "memory pprof dumps are not supported with the mimalloc allocator";
pub async fn init_from_env() {
let target_env = option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown");
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_PLATFORM,
event = "profiling_runtime_skipped",
reason = "unsupported_platform",
target_os = std::env::consts::OS,
target_env,
target_arch = std::env::consts::ARCH,
"Profiling runtime skipped"
);
}
/// Stop all background profiling tasks
pub fn shutdown_profiling() {
let target_env = option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown");
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_PLATFORM,
event = "profiling_shutdown_skipped",
reason = "unsupported_platform",
target_os = std::env::consts::OS,
target_env,
target_arch = std::env::consts::ARCH,
"Profiling shutdown skipped"
);
}
pub async fn dump_cpu_pprof_for(_duration: Duration) -> Result<PathBuf, String> {
Err(unsupported_message("CPU profiling"))
}
pub async fn dump_memory_pprof_now() -> Result<PathBuf, String> {
Err(unsupported_message("Memory profiling"))
}
fn unsupported_message(feature: &str) -> String {
let target_env = option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown");
format!(
"{feature} is only supported on linux x86_64 gnu. target_os={}, target_env={target_env}, target_arch={}",
std::env::consts::OS,
std::env::consts::ARCH
)
}
pub async fn init_from_env() {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_RUNTIME,
event = "profiling_runtime_skipped",
reason = LOCAL_CPU_PPROF_UNSUPPORTED_REASON,
target_os = std::env::consts::OS,
target_env = target_env(),
target_arch = std::env::consts::ARCH,
"Local pprof profiling runtime skipped"
);
}
#[cfg(not(any(target_os = "linux", target_os = "macos")))]
pub use unsupported_impl::{dump_cpu_pprof_for, dump_memory_pprof_now, init_from_env, shutdown_profiling};
#[cfg(any(target_os = "linux", target_os = "macos"))]
mod linux_impl {
use pprof::protos::Message;
use rustfs_config::{
DEFAULT_CPU_DURATION_SECS, DEFAULT_CPU_FREQ, DEFAULT_CPU_INTERVAL_SECS, DEFAULT_CPU_MODE, DEFAULT_ENABLE_PROFILING,
DEFAULT_MEM_INTERVAL_SECS, DEFAULT_MEM_PERIODIC, DEFAULT_OUTPUT_DIR, ENV_CPU_DURATION_SECS, ENV_CPU_FREQ,
ENV_CPU_INTERVAL_SECS, ENV_CPU_MODE, ENV_ENABLE_PROFILING, ENV_MEM_INTERVAL_SECS, ENV_MEM_PERIODIC, ENV_OUTPUT_DIR,
};
use rustfs_utils::{get_env_bool, get_env_str, get_env_u64, get_env_usize};
use std::fs::{File, create_dir_all};
use std::io::Write;
use std::path::{Path, PathBuf};
use std::sync::{Arc, OnceLock};
use std::time::Duration;
use tokio::sync::Mutex;
use tokio::time::sleep;
use tokio_util::sync::CancellationToken;
use tracing::{debug, info, warn};
const LOG_COMPONENT_PROFILING: &str = "profiling";
const LOG_SUBSYSTEM_CPU: &str = "cpu";
const LOG_SUBSYSTEM_MEMORY: &str = "memory";
const LOG_SUBSYSTEM_JEMALLOC: &str = "jemalloc";
const LOG_SUBSYSTEM_RUNTIME: &str = "runtime";
static CPU_CONT_GUARD: OnceLock<Arc<Mutex<Option<pprof::ProfilerGuard<'static>>>>> = OnceLock::new();
static PROFILING_CANCEL_TOKEN: OnceLock<CancellationToken> = OnceLock::new();
/// CPU profiling mode
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum CpuMode {
Off,
Continuous,
Periodic,
}
/// Get or create output directory
fn output_dir() -> PathBuf {
let dir = get_env_str(ENV_OUTPUT_DIR, DEFAULT_OUTPUT_DIR);
let p = PathBuf::from(dir);
if let Err(e) = create_dir_all(&p) {
warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_RUNTIME,
event = "profiling_output_dir_fallback",
path = %p.display(),
error = %e,
fallback = ".",
"Profiling output directory fallback applied"
);
return PathBuf::from(".");
}
p
}
/// Read CPU profiling mode from env
fn read_cpu_mode() -> CpuMode {
match get_env_str(ENV_CPU_MODE, DEFAULT_CPU_MODE).to_lowercase().as_str() {
"continuous" => CpuMode::Continuous,
"periodic" => CpuMode::Periodic,
_ => CpuMode::Off,
}
}
/// Generate timestamp string for filenames
fn ts() -> String {
jiff::Zoned::now().strftime("%Y%m%dT%H%M%S").to_string()
}
/// Write pprof report to file in protobuf format
fn write_pprof_report_pb(report: &pprof::Report, path: &Path) -> Result<(), String> {
let profile = report.pprof().map_err(|e| format!("pprof() failed: {e}"))?;
let mut buf = Vec::with_capacity(512 * 1024);
profile.write_to_vec(&mut buf).map_err(|e| format!("encode failed: {e}"))?;
let mut f = File::create(path).map_err(|e| format!("create file failed: {e}"))?;
f.write_all(&buf).map_err(|e| format!("write file failed: {e}"))?;
Ok(())
}
/// Internal: dump CPU pprof from existing guard
async fn dump_cpu_with_guard(guard: &pprof::ProfilerGuard<'_>) -> Result<PathBuf, String> {
let report = guard.report().build().map_err(|e| format!("build report failed: {e}"))?;
let out = output_dir().join(format!("cpu_profile_{}.pb", ts()));
write_pprof_report_pb(&report, &out)?;
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_dump_exported",
profile_type = "cpu",
path = %out.display(),
"Profiling dump exported"
);
Ok(out)
}
// Public API: dump CPU for a duration; if continuous guard exists, snapshot immediately.
pub async fn dump_cpu_pprof_for(duration: Duration) -> Result<PathBuf, String> {
if let Some(cell) = CPU_CONT_GUARD.get() {
let guard_slot = cell.lock().await;
if let Some(ref guard) = *guard_slot {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_dump_source",
profile_type = "cpu",
source = "continuous_guard",
"Using continuous CPU profiling guard for dump"
);
return dump_cpu_with_guard(guard).await;
}
}
let freq = get_env_usize(ENV_CPU_FREQ, DEFAULT_CPU_FREQ) as i32;
let guard = pprof::ProfilerGuard::new(freq).map_err(|e| format!("create profiler failed: {e}"))?;
sleep(duration).await;
dump_cpu_with_guard(&guard).await
}
// Public API: dump memory pprof now (jemalloc)
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
pub async fn dump_memory_pprof_now() -> Result<PathBuf, String> {
let out = output_dir().join(format!("mem_profile_{}.pb", ts()));
let mut f = File::create(&out).map_err(|e| format!("create file failed: {e}"))?;
let prof_ctl_cell = jemalloc_pprof::PROF_CTL
.as_ref()
.ok_or_else(|| "jemalloc profiling control not available".to_string())?;
let mut prof_ctl = prof_ctl_cell.lock().await;
if !prof_ctl.activated() {
return Err("jemalloc profiling is not active".to_string());
}
let bytes = prof_ctl.dump_pprof().map_err(|e| format!("dump pprof failed: {e}"))?;
f.write_all(&bytes).map_err(|e| format!("write file failed: {e}"))?;
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_exported",
profile_type = "memory",
path = %out.display(),
"Profiling dump exported"
);
Ok(out)
}
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
pub async fn dump_memory_pprof_now() -> Result<PathBuf, String> {
Err(memory_profiling_unsupported_message())
}
// Jemalloc status check (No forced placement, only status observation)
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
pub async fn check_jemalloc_profiling() {
use tikv_jemalloc_ctl::{config, epoch, stats};
if let Err(e) = epoch::advance() {
warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_epoch_advance_failed",
error = %e,
"Jemalloc profiling state changed"
);
}
match config::malloc_conf::read() {
Ok(conf) => debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_malloc_conf",
result = "ok",
malloc_conf = %conf,
"Jemalloc profiling state checked"
),
Err(e) => debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_malloc_conf",
result = "read_failed",
error = %e,
"Jemalloc profiling state checked"
),
}
match std::env::var("MALLOC_CONF") {
Ok(v) => debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_malloc_conf_env",
state = "set",
malloc_conf = %v,
"Jemalloc profiling state checked"
),
Err(_) => debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_malloc_conf_env",
state = "unset",
"Jemalloc profiling state checked"
),
}
if let Some(lock) = jemalloc_pprof::PROF_CTL.as_ref() {
let ctl = lock.lock().await;
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_profiling_status",
activated = ctl.activated(),
"Jemalloc profiling status checked"
);
} else {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_profiling_status",
state = "unavailable",
"Jemalloc profiling status checked"
);
}
let _ = epoch::advance();
macro_rules! show {
($name:literal, $reader:expr) => {
match $reader {
Ok(v) => debug!(concat!($name, "={}"), v),
Err(e) => debug!(concat!($name, " read failed: {}"), e),
}
};
}
show!("allocated", stats::allocated::read());
show!("resident", stats::resident::read());
show!("mapped", stats::mapped::read());
show!("metadata", stats::metadata::read());
show!("active", stats::active::read());
}
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
pub async fn check_jemalloc_profiling() {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_JEMALLOC,
event = "jemalloc_profiling_status",
result = "skipped",
reason = "unsupported_target",
"Jemalloc profiling status checked"
);
}
// Internal: start continuous CPU profiling
async fn start_cpu_continuous(freq_hz: i32) {
let cell = CPU_CONT_GUARD.get_or_init(|| Arc::new(Mutex::new(None))).clone();
let mut slot = cell.lock().await;
if slot.is_some() {
warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_continuous",
state = "already_running",
"CPU profiling already running"
);
return;
}
match pprof::ProfilerGuardBuilder::default()
.frequency(freq_hz)
.blocklist(&["libc", "libgcc", "pthread", "vdso"])
.build()
{
Ok(guard) => {
*slot = Some(guard);
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_continuous",
state = "started",
freq_hz,
"CPU profiling started"
);
}
Err(e) => warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_continuous",
state = "start_failed",
error = %e,
"CPU profiling failed to start"
),
}
}
// Internal: start periodic CPU sampling loop
async fn start_cpu_periodic(freq_hz: i32, interval: Duration, duration: Duration, token: CancellationToken) {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_periodic",
state = "started",
freq_hz,
?interval,
?duration,
"Periodic CPU profiling started"
);
tokio::spawn(async move {
loop {
tokio::select! {
_ = token.cancelled() => {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_periodic",
state = "cancelled",
"Periodic CPU profiling cancelled"
);
break;
}
_ = sleep(interval) => {}
}
if token.is_cancelled() {
break;
}
let guard = match pprof::ProfilerGuard::new(freq_hz) {
Ok(g) => g,
Err(e) => {
warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_capture_failed",
profile_type = "cpu_periodic",
stage = "create_guard",
error = %e,
"Profiling capture failed"
);
continue;
}
};
tokio::select! {
_ = token.cancelled() => {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_state",
profile_type = "cpu_periodic",
state = "cancelled_during_capture",
"Periodic CPU profiling cancelled during capture"
);
break;
}
_ = sleep(duration) => {}
}
match guard.report().build() {
Ok(report) => {
let out = output_dir().join(format!("cpu_profile_{}.pb", ts()));
if let Err(e) = write_pprof_report_pb(&report, &out) {
warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_dump_failed",
profile_type = "cpu_periodic",
stage = "write_dump",
path = %out.display(),
error = %e,
"Periodic CPU dump write failed"
);
} else {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_dump_exported",
profile_type = "cpu_periodic",
path = %out.display(),
"Profiling dump exported"
);
}
}
Err(e) => warn!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_capture_failed",
profile_type = "cpu_periodic",
stage = "build_report",
error = %e,
"Profiling capture failed"
),
}
}
});
}
// Internal: start periodic memory dump when jemalloc profiling is active
#[cfg(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64"))]
async fn start_memory_periodic(interval: Duration, token: CancellationToken) {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_state",
profile_type = "memory_periodic",
state = "started",
?interval,
"Periodic memory profiling started"
);
tokio::spawn(async move {
loop {
tokio::select! {
_ = token.cancelled() => {
info!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_state",
profile_type = "memory_periodic",
state = "cancelled",
"Periodic memory profiling cancelled"
);
break;
}
_ = sleep(interval) => {}
}
let Some(lock) = jemalloc_pprof::PROF_CTL.as_ref() else {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_skipped",
profile_type = "memory_periodic",
reason = "prof_ctl_unavailable",
"Profiling dump skipped"
);
continue;
};
let mut ctl = lock.lock().await;
if !ctl.activated() {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_skipped",
profile_type = "memory_periodic",
reason = "jemalloc_inactive",
"Profiling dump skipped"
);
continue;
}
let out = output_dir().join(format!("mem_profile_periodic_{}.pb", ts()));
match File::create(&out) {
Err(e) => {
tracing::error!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_failed",
profile_type = "memory_periodic",
stage = "create_file",
path = %out.display(),
error = %e,
"Periodic memory dump file creation failed"
);
continue;
}
Ok(mut f) => match ctl.dump_pprof() {
Ok(bytes) => {
if let Err(e) = f.write_all(&bytes) {
tracing::error!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_failed",
profile_type = "memory_periodic",
stage = "write_dump",
path = %out.display(),
error = %e,
"Periodic memory dump write failed"
);
} else {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_exported",
profile_type = "memory_periodic",
path = %out.display(),
"Profiling dump exported"
);
}
}
Err(e) => tracing::error!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_failed",
profile_type = "memory_periodic",
stage = "dump_pprof",
error = %e,
"Periodic memory dump export failed"
),
},
}
}
});
}
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
async fn start_memory_periodic(_interval: Duration, _token: CancellationToken) {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_runtime_skipped",
profile_type = "memory_periodic",
reason = "unsupported_target",
"Profiling runtime skipped"
);
}
// Public: unified init entry, avoid duplication/conflict
pub async fn init_from_env() {
let enabled = get_env_bool(ENV_ENABLE_PROFILING, DEFAULT_ENABLE_PROFILING);
if !enabled {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_RUNTIME,
event = "profiling_runtime_disabled",
reason = "env_flag",
"Profiling runtime disabled"
);
return;
}
// Jemalloc state check once (no dump)
check_jemalloc_profiling().await;
// Initialize cancellation token
let token = PROFILING_CANCEL_TOKEN.get_or_init(CancellationToken::new).clone();
// CPU
let cpu_mode = read_cpu_mode();
let cpu_freq = get_env_usize(ENV_CPU_FREQ, DEFAULT_CPU_FREQ) as i32;
let cpu_interval = Duration::from_secs(get_env_u64(ENV_CPU_INTERVAL_SECS, DEFAULT_CPU_INTERVAL_SECS));
let cpu_duration = Duration::from_secs(get_env_u64(ENV_CPU_DURATION_SECS, DEFAULT_CPU_DURATION_SECS));
match cpu_mode {
CpuMode::Off => debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_mode_selected",
profile_type = "cpu",
state = "off",
"Profiling mode selected"
),
CpuMode::Continuous => start_cpu_continuous(cpu_freq).await,
CpuMode::Periodic => start_cpu_periodic(cpu_freq, cpu_interval, cpu_duration, token.clone()).await,
}
// Memory
let mem_periodic = get_env_bool(ENV_MEM_PERIODIC, DEFAULT_MEM_PERIODIC);
let mem_interval = Duration::from_secs(get_env_u64(ENV_MEM_INTERVAL_SECS, DEFAULT_MEM_INTERVAL_SECS));
if mem_periodic {
start_memory_periodic(mem_interval, token).await;
}
}
#[cfg(not(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")))]
fn memory_profiling_unsupported_message() -> String {
let target_env = option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown");
format!(
"Memory profiling is only supported on linux x86_64 gnu. target_os={}, target_env={target_env}, target_arch={}",
std::env::consts::OS,
std::env::consts::ARCH
)
}
/// Stop all background profiling tasks
pub fn shutdown_profiling() {
if let Some(token) = PROFILING_CANCEL_TOKEN.get() {
token.cancel();
}
}
pub fn shutdown_profiling() {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_RUNTIME,
event = "profiling_shutdown_skipped",
reason = LOCAL_CPU_PPROF_UNSUPPORTED_REASON,
target_os = std::env::consts::OS,
target_env = target_env(),
target_arch = std::env::consts::ARCH,
"Local pprof profiling shutdown skipped"
);
}
#[cfg(any(target_os = "linux", target_os = "macos"))]
pub use linux_impl::{dump_cpu_pprof_for, dump_memory_pprof_now, init_from_env, shutdown_profiling};
pub fn log_cpu_pprof_dump_skipped() {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_CPU,
event = "profiling_dump_skipped",
profile_type = "cpu",
reason = LOCAL_CPU_PPROF_UNSUPPORTED_REASON,
"Local CPU pprof dump skipped"
);
}
pub fn log_memory_pprof_dump_skipped() {
debug!(
component = LOG_COMPONENT_PROFILING,
subsystem = LOG_SUBSYSTEM_MEMORY,
event = "profiling_dump_skipped",
profile_type = "memory",
reason = MEMORY_PPROF_UNSUPPORTED_REASON,
"Memory pprof dump skipped"
);
}
pub fn local_cpu_pprof_unsupported_message() -> String {
unsupported_message(LOCAL_CPU_PPROF_UNSUPPORTED_SUMMARY)
}
pub fn memory_pprof_unsupported_message() -> String {
unsupported_message(MEMORY_PPROF_UNSUPPORTED_SUMMARY)
}
fn unsupported_message(summary: &str) -> String {
format!(
"{summary}. target_os={}, target_env={}, target_arch={}",
std::env::consts::OS,
target_env(),
std::env::consts::ARCH
)
}
fn target_env() -> &'static str {
option_env!("CARGO_CFG_TARGET_ENV").unwrap_or("unknown")
}
+2 -10
View File
@@ -143,11 +143,7 @@ fn target_env_name() -> Option<&'static str> {
}
fn cpu_profiling_status() -> CapabilityStatus {
if cfg!(any(target_os = "linux", target_os = "macos")) {
CapabilityStatus::supported()
} else {
CapabilityStatus::unsupported().with_reason("userspace CPU profiling supports linux and macos targets")
}
CapabilityStatus::unsupported().with_reason(crate::profiling::LOCAL_CPU_PPROF_UNSUPPORTED_SUMMARY)
}
fn ebpf_status() -> CapabilityStatus {
@@ -167,11 +163,7 @@ fn numa_status() -> CapabilityStatus {
}
fn memory_profiling_status() -> CapabilityStatus {
if cfg!(all(target_os = "linux", target_env = "gnu", target_arch = "x86_64")) {
CapabilityStatus::supported().with_reason("jemalloc memory profiling target")
} else {
CapabilityStatus::unsupported().with_reason("memory profiling supports linux gnu x86_64 targets")
}
CapabilityStatus::unsupported().with_reason(crate::profiling::MEMORY_PPROF_UNSUPPORTED_SUMMARY)
}
fn cgroup_memory_status() -> CapabilityStatus {