Add authenticated internal worker API

This commit is contained in:
ms2k 2026-07-17 09:16:46 +09:00
commit ebf818d23c
20 changed files with 2745 additions and 43 deletions

488
Cargo.lock generated
View file

@ -2,6 +2,216 @@
# It is not intended for manual editing. # It is not intended for manual editing.
version = 4 version = 4
[[package]]
name = "actix-codec"
version = "0.5.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5f7b0a21988c1bf877cf4759ef5ddaac04c1c9fe808c9142ecb78ba97d97a28a"
dependencies = [
"bitflags 2.13.0",
"bytes",
"futures-core",
"futures-sink",
"memchr",
"pin-project-lite",
"tokio",
"tokio-util",
"tracing",
]
[[package]]
name = "actix-http"
version = "3.13.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "48e2faa3e7418ed780cca54829d32782a4008a077230f67457caa063415e99c2"
dependencies = [
"actix-codec",
"actix-rt",
"actix-service",
"actix-tls",
"actix-utils",
"base64",
"bitflags 2.13.0",
"brotli",
"bytes",
"bytestring",
"derive_more",
"encoding_rs",
"flate2",
"foldhash",
"futures-core",
"h2 0.3.27",
"http 0.2.12",
"httparse",
"httpdate",
"itoa",
"language-tags",
"local-channel",
"mime",
"percent-encoding",
"pin-project-lite",
"rand 0.10.2",
"sha1",
"smallvec",
"tokio",
"tokio-util",
"tracing",
"zstd",
]
[[package]]
name = "actix-macros"
version = "0.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e01ed3140b2f8d422c68afa1ed2e85d996ea619c988ac834d255db32138655cb"
dependencies = [
"quote",
"syn",
]
[[package]]
name = "actix-router"
version = "0.5.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "14f8c75c51892f18d9c46150c5ac7beb81c95f78c8b83a634d49f4ca32551fe7"
dependencies = [
"bytestring",
"cfg-if 1.0.4",
"http 0.2.12",
"regex",
"regex-lite",
"serde",
"tracing",
]
[[package]]
name = "actix-rt"
version = "2.11.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "92589714878ca59a7626ea19734f0e07a6a875197eec751bb5d3f99e64998c63"
dependencies = [
"futures-core",
"tokio",
]
[[package]]
name = "actix-server"
version = "2.6.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a65064ea4a457eaf07f2fba30b4c695bf43b721790e9530d26cb6f9019ff7502"
dependencies = [
"actix-rt",
"actix-service",
"actix-utils",
"futures-core",
"futures-util",
"mio",
"socket2 0.5.10",
"tokio",
"tracing",
]
[[package]]
name = "actix-service"
version = "2.0.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9e46f36bf0e5af44bdc4bdb36fbbd421aa98c79a9bce724e1edeb3894e10dc7f"
dependencies = [
"futures-core",
"pin-project-lite",
]
[[package]]
name = "actix-tls"
version = "3.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6176099de3f58fbddac916a7f8c6db297e021d706e7a6b99947785fee14abe9f"
dependencies = [
"actix-rt",
"actix-service",
"actix-utils",
"futures-core",
"impl-more 0.1.9",
"pin-project-lite",
"rustls-pki-types",
"tokio",
"tokio-rustls",
"tokio-util",
"tracing",
]
[[package]]
name = "actix-utils"
version = "3.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "88a1dcdff1466e3c2488e1cb5c36a71822750ad43839937f85d2f4d9f8b705d8"
dependencies = [
"local-waker",
"pin-project-lite",
]
[[package]]
name = "actix-web"
version = "4.14.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "df09e2d9239703dd64056359c920c7f3fba6535ec61a0059e0f44e095ffe02b4"
dependencies = [
"actix-codec",
"actix-http",
"actix-macros",
"actix-router",
"actix-rt",
"actix-server",
"actix-service",
"actix-tls",
"actix-utils",
"actix-web-codegen",
"bytes",
"bytestring",
"cfg-if 1.0.4",
"cookie",
"derive_more",
"encoding_rs",
"foldhash",
"futures-core",
"futures-util",
"impl-more 0.3.1",
"itoa",
"language-tags",
"log",
"mime",
"once_cell",
"pin-project-lite",
"regex",
"regex-lite",
"serde",
"serde_json",
"serde_urlencoded",
"smallvec",
"socket2 0.6.5",
"time",
"tracing",
"url",
]
[[package]]
name = "actix-web-codegen"
version = "4.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f591380e2e68490b5dfaf1dd1aa0ebe78d84ba7067078512b4ea6e4492d622b8"
dependencies = [
"actix-router",
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "adler2"
version = "2.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "320119579fcad9c21884f5c4861d16174d0e06250625266f50fe6898340abefa"
[[package]] [[package]]
name = "ahash" name = "ahash"
version = "0.8.12" version = "0.8.12"
@ -25,6 +235,21 @@ dependencies = [
"memchr", "memchr",
] ]
[[package]]
name = "alloc-no-stdlib"
version = "2.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cc7bb162ec39d46ab1ca8c77bf72e890535becd1751bb45f64c597edb4c8c6b3"
[[package]]
name = "alloc-stdlib"
version = "0.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0e76a019e91224d279006ff972f1e984179a6e9feb050adba6ce8274aef23195"
dependencies = [
"alloc-no-stdlib",
]
[[package]] [[package]]
name = "allocator-api2" name = "allocator-api2"
version = "0.2.21" version = "0.2.21"
@ -421,6 +646,27 @@ version = "0.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "dc0b364ead1874514c8c2855ab558056ebfeb775653e7ae45ff72f28f8f3166c" checksum = "dc0b364ead1874514c8c2855ab558056ebfeb775653e7ae45ff72f28f8f3166c"
[[package]]
name = "brotli"
version = "8.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5cc91aac060a7a1e25823bdccbfb6af1875b88f17c6daac97894eed8207166b3"
dependencies = [
"alloc-no-stdlib",
"alloc-stdlib",
"brotli-decompressor",
]
[[package]]
name = "brotli-decompressor"
version = "5.0.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3a32acac15fe1967bc3986b2a6347dffc965602354ea6f450ad07e8bfd253583"
dependencies = [
"alloc-no-stdlib",
"alloc-stdlib",
]
[[package]] [[package]]
name = "bstr" name = "bstr"
version = "1.13.0" version = "1.13.0"
@ -472,6 +718,15 @@ dependencies = [
"either", "either",
] ]
[[package]]
name = "bytestring"
version = "1.5.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "86566c496f2f47d9b8147a4c8b02ffdb69c919fe0c2b2e7195d22cbba0e635c9"
dependencies = [
"bytes",
]
[[package]] [[package]]
name = "cc" name = "cc"
version = "1.2.67" version = "1.2.67"
@ -626,6 +881,26 @@ version = "0.4.2"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3d52eff69cd5e647efe296129160853a42795992097e8af39800e1060caeea9b" checksum = "3d52eff69cd5e647efe296129160853a42795992097e8af39800e1060caeea9b"
[[package]]
name = "convert_case"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "633458d4ef8c78b72454de2d54fd6ab2e60f9e02be22f3c6104cdc8a4e0fceb9"
dependencies = [
"unicode-segmentation",
]
[[package]]
name = "cookie"
version = "0.16.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e859cd57d0710d9e06c381b550c06e76992472a8c6d527aecd2fc673dcc231fb"
dependencies = [
"percent-encoding",
"time",
"version_check",
]
[[package]] [[package]]
name = "core-foundation" name = "core-foundation"
version = "0.9.4" version = "0.9.4"
@ -886,6 +1161,29 @@ dependencies = [
"serde_core", "serde_core",
] ]
[[package]]
name = "derive_more"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d751e9e49156b02b44f9c1815bcb94b984cdcc4396ecc32521c739452808b134"
dependencies = [
"derive_more-impl",
]
[[package]]
name = "derive_more-impl"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "799a97264921d8623a957f6c3b9011f3b5492f557bbb7a5a19b7fa6d06ba8dcb"
dependencies = [
"convert_case",
"proc-macro2",
"quote",
"rustc_version",
"syn",
"unicode-xid",
]
[[package]] [[package]]
name = "digest" name = "digest"
version = "0.10.7" version = "0.10.7"
@ -1049,6 +1347,16 @@ version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5baebc0774151f905a1a2cc41989300b1e6fbb29aff0ceffa1064fdd3088d582" checksum = "5baebc0774151f905a1a2cc41989300b1e6fbb29aff0ceffa1064fdd3088d582"
[[package]]
name = "flate2"
version = "1.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "843fba2746e448b37e26a819579957415c8cef339bf08564fe8b7ddbd959573c"
dependencies = [
"crc32fast",
"miniz_oxide",
]
[[package]] [[package]]
name = "fluent-uri" name = "fluent-uri"
version = "0.4.1" version = "0.4.1"
@ -1300,6 +1608,25 @@ dependencies = [
"regex-syntax", "regex-syntax",
] ]
[[package]]
name = "h2"
version = "0.3.27"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0beca50380b1fc32983fc1cb4587bfa4bb9e78fc259aad4a0032d2080309222d"
dependencies = [
"bytes",
"fnv",
"futures-core",
"futures-sink",
"futures-util",
"http 0.2.12",
"indexmap",
"slab",
"tokio",
"tokio-util",
"tracing",
]
[[package]] [[package]]
name = "h2" name = "h2"
version = "0.4.15" version = "0.4.15"
@ -1536,7 +1863,7 @@ dependencies = [
"bytes", "bytes",
"futures-channel", "futures-channel",
"futures-core", "futures-core",
"h2", "h2 0.4.15",
"http 1.4.2", "http 1.4.2",
"http-body 1.1.0", "http-body 1.1.0",
"httparse", "httparse",
@ -1580,7 +1907,7 @@ dependencies = [
"libc", "libc",
"percent-encoding", "percent-encoding",
"pin-project-lite", "pin-project-lite",
"socket2", "socket2 0.6.5",
"system-configuration", "system-configuration",
"tokio", "tokio",
"tower-service", "tower-service",
@ -1721,6 +2048,18 @@ dependencies = [
"icu_properties", "icu_properties",
] ]
[[package]]
name = "impl-more"
version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e8a5a9a0ff0086c7a148acb942baaabeadf9504d10400b5a05645853729b9cd2"
[[package]]
name = "impl-more"
version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "35a84fd5aa25fae5c0f4a33d9cac2ca017fc622cbd089be2229993514990f870"
[[package]] [[package]]
name = "indexmap" name = "indexmap"
version = "2.14.0" version = "2.14.0"
@ -1934,6 +2273,12 @@ version = "0.4.1"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a" checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"
[[package]]
name = "language-tags"
version = "0.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d4345964bb142484797b161f473a503a434de77149dd8c7427788c6e13379388"
[[package]] [[package]]
name = "lazy_static" name = "lazy_static"
version = "1.5.0" version = "1.5.0"
@ -1998,6 +2343,23 @@ version = "0.8.2"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "92daf443525c4cce67b150400bc2316076100ce0b3686209eb8cf3c31612e6f0" checksum = "92daf443525c4cce67b150400bc2316076100ce0b3686209eb8cf3c31612e6f0"
[[package]]
name = "local-channel"
version = "0.1.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b6cbc85e69b8df4b8bb8b89ec634e7189099cea8927a276b7384ce5488e53ec8"
dependencies = [
"futures-core",
"futures-sink",
"local-waker",
]
[[package]]
name = "local-waker"
version = "0.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4d873d7c67ce09b42110d801813efbc9364414e356be9935700d368351657487"
[[package]] [[package]]
name = "lock_api" name = "lock_api"
version = "0.4.14" version = "0.4.14"
@ -2090,6 +2452,16 @@ version = "0.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a" checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a"
[[package]]
name = "miniz_oxide"
version = "0.8.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1fa76a2c86f704bdb222d66965fb3d63269ce38518b83cb0575fca855ebb6316"
dependencies = [
"adler2",
"simd-adler32",
]
[[package]] [[package]]
name = "mio" name = "mio"
version = "1.2.2" version = "1.2.2"
@ -2097,6 +2469,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "30d65c71f1ce40ab09135ce117d742b9f8a19ff91a41a8b57ed50bc2de59c427" checksum = "30d65c71f1ce40ab09135ce117d742b9f8a19ff91a41a8b57ed50bc2de59c427"
dependencies = [ dependencies = [
"libc", "libc",
"log",
"wasi 0.11.1+wasi-snapshot-preview1", "wasi 0.11.1+wasi-snapshot-preview1",
"windows-sys 0.61.2", "windows-sys 0.61.2",
] ]
@ -2604,7 +2977,7 @@ dependencies = [
"quinn-udp", "quinn-udp",
"rustc-hash", "rustc-hash",
"rustls", "rustls",
"socket2", "socket2 0.6.5",
"thiserror", "thiserror",
"tokio", "tokio",
"tracing", "tracing",
@ -2643,7 +3016,7 @@ dependencies = [
"cfg_aliases", "cfg_aliases",
"libc", "libc",
"once_cell", "once_cell",
"socket2", "socket2 0.6.5",
"tracing", "tracing",
"windows-sys 0.61.2", "windows-sys 0.61.2",
] ]
@ -2731,6 +3104,7 @@ version = "0.14.8"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "57f6d249aad744e274e682777a50283a225a32705394ee6d5fcc01efa25e4055" checksum = "57f6d249aad744e274e682777a50283a225a32705394ee6d5fcc01efa25e4055"
dependencies = [ dependencies = [
"pem",
"ring", "ring",
"rustls-pki-types", "rustls-pki-types",
"time", "time",
@ -2766,7 +3140,7 @@ dependencies = [
"pin-project-lite", "pin-project-lite",
"ryu", "ryu",
"sha1_smol", "sha1_smol",
"socket2", "socket2 0.6.5",
"tokio", "tokio",
"tokio-util", "tokio-util",
"url", "url",
@ -2852,6 +3226,12 @@ dependencies = [
"regex-syntax", "regex-syntax",
] ]
[[package]]
name = "regex-lite"
version = "0.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cab834c73d247e67f4fae452806d17d3c7501756d98c8808d7c9c7aa7d18f973"
[[package]] [[package]]
name = "regex-syntax" name = "regex-syntax"
version = "0.8.11" version = "0.8.11"
@ -2880,6 +3260,31 @@ dependencies = [
[[package]] [[package]]
name = "remora-api" name = "remora-api"
version = "0.0.0" version = "0.0.0"
dependencies = [
"actix-tls",
"actix-web",
"getrandom 0.3.4",
"rcgen",
"remora-admission",
"remora-attempt",
"remora-contracts",
"remora-coordination",
"remora-credentials",
"remora-domain",
"remora-durable",
"remora-identity",
"remora-result",
"remora-target",
"reqwest 0.13.4",
"rustls",
"serde",
"serde_json",
"sha2 0.11.0",
"sqlx",
"thiserror",
"time",
"tokio",
]
[[package]] [[package]]
name = "remora-attempt" name = "remora-attempt"
@ -3222,7 +3627,7 @@ dependencies = [
"encoding_rs", "encoding_rs",
"futures-core", "futures-core",
"futures-util", "futures-util",
"h2", "h2 0.4.15",
"http 1.4.2", "http 1.4.2",
"http-body 1.1.0", "http-body 1.1.0",
"http-body-util", "http-body-util",
@ -3358,6 +3763,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3c54fcab019b409d04215d3a17cb438fd7fbf192ee61461f20f4fe18704bc138" checksum = "3c54fcab019b409d04215d3a17cb438fd7fbf192ee61461f20f4fe18704bc138"
dependencies = [ dependencies = [
"aws-lc-rs", "aws-lc-rs",
"log",
"once_cell", "once_cell",
"ring", "ring",
"rustls-pki-types", "rustls-pki-types",
@ -3669,6 +4075,16 @@ version = "2.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f8fadd59c855ef2080decdef8ff161eb6661b86933c9d82e5ba29dc602a55aba" checksum = "f8fadd59c855ef2080decdef8ff161eb6661b86933c9d82e5ba29dc602a55aba"
[[package]]
name = "signal-hook-registry"
version = "1.4.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c4db69cba1110affc0e9f7bcd48bbf87b3f4fc7c61fc9155afd4c469eb3d6c1b"
dependencies = [
"errno",
"libc",
]
[[package]] [[package]]
name = "signature" name = "signature"
version = "2.2.0" version = "2.2.0"
@ -3679,6 +4095,12 @@ dependencies = [
"rand_core 0.6.4", "rand_core 0.6.4",
] ]
[[package]]
name = "simd-adler32"
version = "0.3.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3a219298ac11a56ea9a6d2120044824d6f01aeb034955e7af7bc16858527deea"
[[package]] [[package]]
name = "simd_cesu8" name = "simd_cesu8"
version = "1.2.0" version = "1.2.0"
@ -3722,6 +4144,16 @@ dependencies = [
"serde", "serde",
] ]
[[package]]
name = "socket2"
version = "0.5.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e22376abed350d73dd1cd119b57ffccad95b4e585a7cda43e286245ce23c0678"
dependencies = [
"libc",
"windows-sys 0.52.0",
]
[[package]] [[package]]
name = "socket2" name = "socket2"
version = "0.6.5" version = "0.6.5"
@ -4166,8 +4598,10 @@ dependencies = [
"bytes", "bytes",
"libc", "libc",
"mio", "mio",
"parking_lot",
"pin-project-lite", "pin-project-lite",
"socket2", "signal-hook-registry",
"socket2 0.6.5",
"tokio-macros", "tokio-macros",
"windows-sys 0.61.2", "windows-sys 0.61.2",
] ]
@ -4448,6 +4882,18 @@ version = "0.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7df058c713841ad818f1dc5d3fd88063241cc61f49f5fbea4b951e8cf5a8d71d" checksum = "7df058c713841ad818f1dc5d3fd88063241cc61f49f5fbea4b951e8cf5a8d71d"
[[package]]
name = "unicode-segmentation"
version = "1.13.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c6f5d3c3b1bf09027a88a6bc961fc00497d651009560b5463668dc81b0fa87a8"
[[package]]
name = "unicode-xid"
version = "0.2.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
[[package]] [[package]]
name = "unsafe-libyaml" name = "unsafe-libyaml"
version = "0.2.11" version = "0.2.11"
@ -5228,3 +5674,31 @@ name = "zmij"
version = "1.0.23" version = "1.0.23"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "29666d0abbfad1e3dc4dcf6144730dd3a3ab225bbbdac83319345b1b44ccfc1b" checksum = "29666d0abbfad1e3dc4dcf6144730dd3a3ab225bbbdac83319345b1b44ccfc1b"
[[package]]
name = "zstd"
version = "0.13.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e91ee311a569c327171651566e07972200e76fcfe2242a4fa446149a3881c08a"
dependencies = [
"zstd-safe",
]
[[package]]
name = "zstd-safe"
version = "7.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8f49c4d5f0abb602a93fb8736af2a4f4dd9512e36f7f570d66e65ff867ed3b9d"
dependencies = [
"zstd-sys",
]
[[package]]
name = "zstd-sys"
version = "2.0.16+zstd.1.5.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "91e19ebc2adc8f83e43039e79776e3fda8ca919132d68a1fed6a5faca2683748"
dependencies = [
"cc",
"pkg-config",
]

View file

@ -257,6 +257,34 @@ COPY crates crates
COPY --from=kind-attempt-verified /tmp/remora-attempt-tests.tar.zst /tests/remora-attempt-tests.tar.zst COPY --from=kind-attempt-verified /tmp/remora-attempt-tests.tar.zst /tests/remora-attempt-tests.tar.zst
CMD ["bash", "-ec", "test -s /run/remora-fixture/profile.env && test -n \"$DATABASE_URL\" && test -n \"$REMORA_COORDINATION_URL\" && test -n \"$REMORA_TARGET_ENDPOINT\" && case $REMORA_TARGET_PROVIDER in garage|versitygw) ;; *) exit 1 ;; esac && export REMORA_RECOVERY_REQUIRE=1 && cargo-nextest nextest run --archive-file /tests/remora-attempt-tests.tar.zst --workspace-remap /workspace --test-threads=1 --no-fail-fast -E 'binary(recovery)'"] CMD ["bash", "-ec", "test -s /run/remora-fixture/profile.env && test -n \"$DATABASE_URL\" && test -n \"$REMORA_COORDINATION_URL\" && test -n \"$REMORA_TARGET_ENDPOINT\" && case $REMORA_TARGET_PROVIDER in garage|versitygw) ;; *) exit 1 ;; esac && export REMORA_RECOVERY_REQUIRE=1 && cargo-nextest nextest run --archive-file /tests/remora-attempt-tests.tar.zst --workspace-remap /workspace --test-threads=1 --no-fail-fast -E 'binary(recovery)'"]
FROM rust-tools AS kind-worker-api-build
WORKDIR /workspace
COPY Cargo.toml Cargo.lock rust-toolchain.toml ./
COPY .config .config
COPY contracts contracts
COPY crates crates
COPY docs/api docs/api
RUN cargo nextest archive \
--package remora-api \
--build-jobs 4 \
--locked \
--archive-file /tmp/remora-api-tests.tar.zst
FROM kind-worker-api-build AS kind-worker-api-verified
RUN cargo clippy \
--package remora-api \
--all-targets \
--locked \
-- -D warnings \
&& cargo test --package remora-api --doc --locked
FROM rust-tools AS kind-worker-api-runner
WORKDIR /workspace
COPY Cargo.toml ./
COPY crates crates
COPY --from=kind-worker-api-verified /tmp/remora-api-tests.tar.zst /tests/remora-api-tests.tar.zst
CMD ["bash", "-ec", "test -s /run/remora-fixture/profile.env && test -n \"$DATABASE_URL\" && test -n \"$REMORA_COORDINATION_URL\" && export REMORA_WORKER_API_REQUIRE=1 && cargo-nextest nextest run --archive-file /tests/remora-api-tests.tar.zst --workspace-remap /workspace --test-threads=1 --no-fail-fast"]
FROM rust-tools AS kind-http-source-build FROM rust-tools AS kind-http-source-build
WORKDIR /workspace WORKDIR /workspace
COPY Cargo.toml Cargo.lock rust-toolchain.toml ./ COPY Cargo.toml Cargo.lock rust-toolchain.toml ./

View file

@ -1,4 +1,4 @@
.PHONY: check-dependencies check-fixture-lib check-kind-sample check-durable check-coordination check-identity check-credentials check-admission check-attempt check-attempt-mutations check-recovery check-stream check-target check-result check-outbound check-http-source check-object-source check-hf-source check-fast check-full check-fast-host check-full-host rust-check web-install web-check web-test-fast web-test-full .PHONY: check-dependencies check-fixture-lib check-kind-sample check-durable check-coordination check-identity check-credentials check-admission check-attempt check-attempt-mutations check-recovery check-worker-api check-stream check-target check-result check-outbound check-http-source check-object-source check-hf-source check-fast check-full check-fast-host check-full-host rust-check web-install web-check web-test-fast web-test-full
CONTAINER_ENGINE ?= nerdctl CONTAINER_ENGINE ?= nerdctl
CONTAINER_RUN_FLAGS ?= --memory=16g --memory-swap=24g --cpus=4 CONTAINER_RUN_FLAGS ?= --memory=16g --memory-swap=24g --cpus=4
@ -80,6 +80,9 @@ check-attempt-mutations:
check-recovery: check-recovery:
./tests/kind/run-recovery-job.sh ./tests/kind/run-recovery-job.sh
check-worker-api:
./tests/kind/run-worker-api-job.sh
check-stream: check-stream:
cargo fmt --all --check cargo fmt --all --check
cargo clippy --package remora-stream --all-targets --all-features -- -D warnings cargo clippy --package remora-stream --all-targets --all-features -- -D warnings

View file

@ -202,12 +202,15 @@
"kind": "success", "kind": "success",
"result": { "result": {
"kind": "manifest", "kind": "manifest",
"manifest_id": "manifest_0001",
"schema_version": 1,
"manifest": { "manifest": {
"store": "target-main", "store": "target-main",
"bucket": "target-bucket", "bucket": "target-bucket",
"key": "_remora/results/manifest.json", "key": "_remora/results/manifest.json",
"version_id": "version-1", "version_id": "version-1",
"checksum": "sha256:0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef" "checksum": "sha256:0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
"size": 512
}, },
"member_count": 1, "member_count": 1,
"total_bytes": 4096 "total_bytes": 4096

View file

@ -4,3 +4,30 @@ edition.workspace = true
publish = false publish = false
rust-version.workspace = true rust-version.workspace = true
version.workspace = true version.workspace = true
[dependencies]
actix-web = { version = "=4.14.0", features = ["rustls-0_23"] }
actix-tls = "=3.5.0"
getrandom = "0.3.4"
remora-attempt = { path = "../remora-attempt" }
remora-contracts = { path = "../remora-contracts" }
remora-domain.workspace = true
remora-result = { path = "../remora-result" }
remora-target = { path = "../remora-target" }
serde.workspace = true
serde_json.workspace = true
sha2.workspace = true
sqlx.workspace = true
thiserror.workspace = true
time.workspace = true
[dev-dependencies]
rcgen = "=0.14.8"
remora-admission = { path = "../remora-admission" }
remora-coordination = { path = "../remora-coordination" }
remora-credentials = { path = "../remora-credentials" }
remora-durable.workspace = true
remora-identity = { path = "../remora-identity" }
reqwest = { version = "=0.13.4", default-features = false, features = ["json", "rustls"] }
rustls = "=0.23.42"
tokio.workspace = true

View file

@ -1,3 +1,9 @@
mod worker;
pub use worker::{
AuthenticatedWorker, WorkerApiState, authenticate_rustls_worker, configure_worker_api,
};
pub const COMPONENT_NAME: &str = "remora-api"; pub const COMPONENT_NAME: &str = "remora-api";
#[cfg(test)] #[cfg(test)]

View file

@ -0,0 +1,613 @@
use std::{
any::Any,
future::{Ready, ready},
};
use actix_web::{
FromRequest, HttpRequest, HttpResponse, ResponseError, dev::Payload, http::StatusCode, web,
};
use remora_attempt::{
ActivationMutationOutcome, AttemptAuthority, AttemptError, AttemptMutationOutcome,
AttemptService, CheckpointProjection, CheckpointRecord, ClaimOutcome, DurableDescriptor,
LeaseProjection, MutationRejection, OutcomePublication, ProgressRecord, RenewalMutationOutcome,
};
use remora_contracts::{
ActivationResponseDto, ApiErrorCodeDto, ApiErrorDto, AttemptActivateRequestDto,
AttemptAuthorityDto, AttemptCheckpointRequestDto, AttemptOutcomeDto, AttemptOutcomeRequestDto,
AttemptProgressRequestDto, AttemptRenewRequestDto, AttemptResolvedSourceRequestDto,
CheckpointDescriptorKindDto, CheckpointDescriptorRefDto, ClaimRequestDto, ClaimResponseDto,
CredentialGrantVersionDto, DateTimeDto, DurableDescriptorKindDto, DurableDescriptorRefDto,
ExecutionEnvelopeDto, JobPhaseProjectionDto, LeaseProjectionDto, NullableDto,
PositiveIntegerDto, ResolvedSourceDescriptorKindDto, ResolvedSourceDescriptorRefDto,
SafeIntegerDto, SourceProjectionDto, StagingTargetDto, TargetDto,
};
use remora_domain::OpaqueId;
use remora_result::PublicationInput;
use remora_target::{ContentSha256, ObjectIdentity, RESERVED_NAMESPACE_SEGMENT, StoredObject};
use sha2::{Digest, Sha256};
use time::{OffsetDateTime, format_description::well_known::Rfc3339};
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct AuthenticatedWorker {
pub identity: OpaqueId,
}
pub fn authenticate_rustls_worker(
connection: &dyn Any,
extensions: &mut actix_web::dev::Extensions,
) {
type Stream = actix_tls::accept::rustls_0_23::TlsStream<actix_web::rt::net::TcpStream>;
let Some(stream) = connection.downcast_ref::<Stream>() else {
return;
};
let Some(certificate) = stream
.get_ref()
.1
.peer_certificates()
.and_then(|chain| chain.first())
else {
return;
};
let digest = Sha256::digest(certificate.as_ref());
let mut value = String::from("w");
for byte in digest {
use std::fmt::Write as _;
write!(value, "{byte:02x}").expect("writing to String cannot fail");
}
extensions.insert(AuthenticatedWorker {
identity: OpaqueId::parse(value).expect("certificate digest is a valid identity"),
});
}
impl FromRequest for AuthenticatedWorker {
type Error = WorkerApiError;
type Future = Ready<Result<Self, Self::Error>>;
fn from_request(request: &HttpRequest, _payload: &mut Payload) -> Self::Future {
ready(
request
.conn_data::<Self>()
.cloned()
.ok_or_else(WorkerApiError::unauthenticated),
)
}
}
#[derive(Clone)]
pub struct WorkerApiState {
attempts: AttemptService,
}
impl WorkerApiState {
#[must_use]
pub fn new(attempts: AttemptService) -> Self {
Self { attempts }
}
}
pub fn configure_worker_api(config: &mut web::ServiceConfig) {
config.service(
web::scope("/internal/v1")
.app_data(
web::JsonConfig::default()
.limit(1024 * 1024)
.error_handler(|_, _| WorkerApiError::invalid_request().into()),
)
.route("/jobs/{job_id}/claim", web::post().to(claim))
.route("/attempts/{attempt_id}/renew", web::post().to(renew))
.route("/attempts/{attempt_id}/activate", web::post().to(activate))
.route(
"/attempts/{attempt_id}/resolved-source",
web::post().to(pin_resolved_source),
)
.route(
"/attempts/{attempt_id}/checkpoint",
web::post().to(persist_checkpoint),
)
.route(
"/attempts/{attempt_id}/progress",
web::post().to(report_progress),
)
.route(
"/attempts/{attempt_id}/outcome",
web::post().to(report_outcome),
),
);
}
async fn claim(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<ClaimRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let job_id = parse_id(path.into_inner())?;
let outcome = state
.attempts
.claim(&job_id, &request.worker_incarnation, &request.operation_id)
.await
.map_err(WorkerApiError::from_attempt)?;
let ClaimOutcome::Claimed { authority, .. } = outcome else {
return Err(WorkerApiError::precondition_failed());
};
let execution = load_execution_envelope(&state.attempts, &authority).await?;
Ok(HttpResponse::Ok().json(ClaimResponseDto {
authority: authority_dto(&authority)?,
execution,
}))
}
async fn renew(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptRenewRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
match state
.attempts
.renew_operation(&authority, &request.operation_id)
.await
.map_err(WorkerApiError::from_attempt)?
{
RenewalMutationOutcome::Renewed { lease, .. } => {
Ok(HttpResponse::Ok().json(lease_dto(lease)?))
}
RenewalMutationOutcome::AuthorityLost { .. } => Err(WorkerApiError::precondition_failed()),
}
}
async fn activate(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptActivateRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
let lease = match state
.attempts
.activate_operation(&authority, &request.operation_id)
.await
.map_err(WorkerApiError::from_attempt)?
{
ActivationMutationOutcome::Activated { lease, .. } => lease,
ActivationMutationOutcome::PreconditionFailed => {
return Err(WorkerApiError::precondition_failed());
}
};
let execution = load_execution_envelope(&state.attempts, &authority).await?;
Ok(HttpResponse::Ok().json(ActivationResponseDto {
authority: authority_dto(&authority)?,
execution,
lease: lease_dto(lease)?,
}))
}
async fn pin_resolved_source(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptResolvedSourceRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
accepted(
state
.attempts
.pin_resolved_source(
&authority,
&request.operation_id,
descriptor_from_resolved(&request.resolved_source_ref)?,
)
.await
.map_err(WorkerApiError::from_attempt)?,
)
}
async fn persist_checkpoint(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptCheckpointRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
accepted(
state
.attempts
.persist_checkpoint(
&authority,
&request.operation_id,
CheckpointRecord {
descriptor: descriptor_from_checkpoint(&request.checkpoint_ref)?,
resolved_source_id: request.resolved_source_id.clone(),
validated_bytes: request.validated_bytes.get(),
validated_members: request.validated_members.get(),
},
)
.await
.map_err(WorkerApiError::from_attempt)?,
)
}
async fn report_progress(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptProgressRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
accepted(
state
.attempts
.report_progress(
&authority,
&request.operation_id,
ProgressRecord {
completed_bytes: request.progress.completed_bytes.get(),
total_bytes: request.progress.total_bytes.map(SafeIntegerDto::get),
completed_members: request.progress.completed_members.get(),
total_members: request.progress.total_members.map(SafeIntegerDto::get),
},
)
.await
.map_err(WorkerApiError::from_attempt)?,
)
}
async fn report_outcome(
_: AuthenticatedWorker,
state: web::Data<WorkerApiState>,
path: web::Path<String>,
request: web::Json<AttemptOutcomeRequestDto>,
) -> Result<HttpResponse, WorkerApiError> {
let authority = request_authority(
&state,
path.into_inner(),
request.fence.get(),
&request.worker_incarnation,
)
.await?;
let outcome = match &request.outcome {
AttemptOutcomeDto::Success(success) => {
let result = &success.result;
state
.attempts
.publish_success_operation(
&authority,
&request.operation_id,
PublicationInput {
attempt_id: authority.attempt_id.clone(),
manifest_id: result.manifest_id.clone(),
schema_version: u32::try_from(result.schema_version.get())
.map_err(|_| WorkerApiError::invalid_request())?,
manifest: StoredObject {
identity: ObjectIdentity {
target_store_id: result.manifest.store.clone(),
bucket: result.manifest.bucket.clone(),
key: result.manifest.key.clone(),
version_id: result.manifest.version_id.clone(),
},
content_sha256: ContentSha256::parse(&result.manifest.checksum)
.map_err(|_| WorkerApiError::invalid_request())?,
size: result.manifest.size.get(),
},
member_count: result.member_count.get(),
total_bytes: result.total_bytes.get(),
},
)
.await
}
AttemptOutcomeDto::Failure(failure) => {
state
.attempts
.publish_failure_operation(&authority, &request.operation_id, failure.failure.class)
.await
}
}
.map_err(WorkerApiError::from_attempt)?;
let phase = match outcome {
OutcomePublication::Accepted { phase, .. } => phase,
OutcomePublication::Rejected { reason, .. } => {
return Err(WorkerApiError::from_rejection(reason));
}
};
Ok(HttpResponse::Ok().json(JobPhaseProjectionDto { phase }))
}
fn accepted(outcome: AttemptMutationOutcome) -> Result<HttpResponse, WorkerApiError> {
match outcome {
AttemptMutationOutcome::Accepted(_) => Ok(HttpResponse::NoContent().finish()),
AttemptMutationOutcome::Rejected { reason, .. } => {
Err(WorkerApiError::from_rejection(reason))
}
}
}
async fn request_authority(
state: &WorkerApiState,
attempt_id: String,
fence: u64,
worker: &OpaqueId,
) -> Result<AttemptAuthority, WorkerApiError> {
let attempt_id = parse_id(attempt_id)?;
state
.attempts
.historical_authority(&attempt_id)
.await
.map_err(WorkerApiError::from_attempt)?
.filter(|authority| authority.fence == fence && authority.worker_incarnation == *worker)
.ok_or_else(WorkerApiError::precondition_failed)
}
async fn load_execution_envelope(
attempts: &AttemptService,
authority: &AttemptAuthority,
) -> Result<ExecutionEnvelopeDto, WorkerApiError> {
let row = attempts
.execution_envelope(&authority.attempt_id)
.await
.map_err(WorkerApiError::from_attempt)?
.ok_or_else(WorkerApiError::precondition_failed)?;
if row.job_id != authority.job_id
|| row.attempt_id != authority.attempt_id
|| row.fence != authority.fence
|| row.worker_incarnation != authority.worker_incarnation
{
return Err(WorkerApiError::precondition_failed());
}
let source: SourceProjectionDto =
serde_json::from_value(row.source).map_err(|_| WorkerApiError::unavailable())?;
let target_prefix = row.target_prefix.join("/");
let resolved_source_ref = row.resolved_source.map(|descriptor| {
NullableDto(Some(descriptor_ref(
descriptor,
DurableDescriptorKindDto::ResolvedSource,
)))
});
let checkpoint_ref = row.checkpoint.and_then(|checkpoint| match checkpoint {
CheckpointProjection::Reusable { checkpoint } => Some(NullableDto(Some(descriptor_ref(
checkpoint.descriptor,
DurableDescriptorKindDto::Checkpoint,
)))),
CheckpointProjection::Complete { .. } => None,
});
Ok(ExecutionEnvelopeDto {
id: row.id,
job_id: authority.job_id.clone(),
attempt_id: authority.attempt_id.clone(),
fence: positive(authority.fence)?,
worker_incarnation: authority.worker_incarnation.clone(),
source,
sensitive_source_ref: row.sensitive_source_ref,
credential_version: NullableDto(match (row.credential_id, row.credential_version_id) {
(Some(credential_id), Some(version_id)) => Some(CredentialGrantVersionDto {
credential_id,
version_id,
}),
(None, None) => None,
_ => return Err(WorkerApiError::unavailable()),
}),
resolved_source_ref: resolved_source_ref.unwrap_or(NullableDto(None)),
target: TargetDto {
store: row.target_store.clone(),
bucket: row.target_bucket.clone(),
prefix: target_prefix,
},
staging_target: StagingTargetDto {
store: row.target_store,
bucket: row.target_bucket,
prefix: format!(
"{RESERVED_NAMESPACE_SEGMENT}/a/{}/",
authority.attempt_id.as_str()
),
},
execution_grant_ref: row.execution_grant_id,
checkpoint_ref: checkpoint_ref.unwrap_or(NullableDto(None)),
})
}
fn descriptor_ref(
descriptor: DurableDescriptor,
kind: DurableDescriptorKindDto,
) -> DurableDescriptorRefDto {
DurableDescriptorRefDto {
id: descriptor.id,
kind,
schema_version: positive(descriptor.schema_version.into()).expect("validated descriptor"),
byte_length: safe(descriptor.byte_length).expect("validated descriptor"),
item_count: descriptor
.item_count
.map(|value| safe(value).expect("validated descriptor")),
checksum: descriptor.checksum,
}
}
fn descriptor_from_resolved(
value: &ResolvedSourceDescriptorRefDto,
) -> Result<DurableDescriptor, WorkerApiError> {
let ResolvedSourceDescriptorKindDto::ResolvedSource = value.kind;
Ok(DurableDescriptor {
id: value.id.clone(),
schema_version: u32::try_from(value.schema_version.get())
.map_err(|_| WorkerApiError::invalid_request())?,
byte_length: value.byte_length.get(),
item_count: value.item_count.map(SafeIntegerDto::get),
checksum: value.checksum.clone(),
})
}
fn descriptor_from_checkpoint(
value: &CheckpointDescriptorRefDto,
) -> Result<DurableDescriptor, WorkerApiError> {
let CheckpointDescriptorKindDto::Checkpoint = value.kind;
Ok(DurableDescriptor {
id: value.id.clone(),
schema_version: u32::try_from(value.schema_version.get())
.map_err(|_| WorkerApiError::invalid_request())?,
byte_length: value.byte_length.get(),
item_count: value.item_count.map(SafeIntegerDto::get),
checksum: value.checksum.clone(),
})
}
fn authority_dto(authority: &AttemptAuthority) -> Result<AttemptAuthorityDto, WorkerApiError> {
Ok(AttemptAuthorityDto {
attempt_id: authority.attempt_id.clone(),
fence: positive(authority.fence)?,
worker_incarnation: authority.worker_incarnation.clone(),
})
}
fn lease_dto(lease: LeaseProjection) -> Result<LeaseProjectionDto, WorkerApiError> {
let observed = timestamp(lease.observed_at_ms)?;
let expires = timestamp(lease.expires_at_ms)?;
let renew_at = observed + (expires - observed) / 2;
Ok(LeaseProjectionDto {
expires_at: datetime(expires)?,
renew_after: datetime(renew_at)?,
})
}
fn timestamp(milliseconds: u64) -> Result<OffsetDateTime, WorkerApiError> {
let nanoseconds = i128::from(milliseconds) * 1_000_000;
OffsetDateTime::from_unix_timestamp_nanos(nanoseconds)
.map_err(|_| WorkerApiError::unavailable())
}
fn datetime(value: OffsetDateTime) -> Result<DateTimeDto, WorkerApiError> {
let value = value
.format(&Rfc3339)
.map_err(|_| WorkerApiError::unavailable())?;
DateTimeDto::parse(value).map_err(|_| WorkerApiError::unavailable())
}
fn positive(value: u64) -> Result<PositiveIntegerDto, WorkerApiError> {
PositiveIntegerDto::new(value).map_err(|_| WorkerApiError::unavailable())
}
fn safe(value: u64) -> Result<SafeIntegerDto, WorkerApiError> {
SafeIntegerDto::new(value).map_err(|_| WorkerApiError::unavailable())
}
fn parse_id(value: impl Into<String>) -> Result<OpaqueId, WorkerApiError> {
OpaqueId::parse(value).map_err(|_| WorkerApiError::invalid_request())
}
#[derive(Debug, thiserror::Error)]
#[error("worker API request failed")]
pub struct WorkerApiError {
status: StatusCode,
code: ApiErrorCodeDto,
message: &'static str,
}
impl WorkerApiError {
fn invalid_request() -> Self {
Self::new(
StatusCode::BAD_REQUEST,
ApiErrorCodeDto::InvalidRequest,
"invalid request",
)
}
fn unauthenticated() -> Self {
Self::new(
StatusCode::UNAUTHORIZED,
ApiErrorCodeDto::Unauthenticated,
"authentication required",
)
}
fn precondition_failed() -> Self {
Self::new(
StatusCode::CONFLICT,
ApiErrorCodeDto::PreconditionFailed,
"precondition failed",
)
}
fn unavailable() -> Self {
Self::new(
StatusCode::SERVICE_UNAVAILABLE,
ApiErrorCodeDto::Unavailable,
"service unavailable",
)
}
fn new(status: StatusCode, code: ApiErrorCodeDto, message: &'static str) -> Self {
Self {
status,
code,
message,
}
}
fn from_attempt(error: AttemptError) -> Self {
match error {
AttemptError::Invalid(_) => Self::invalid_request(),
AttemptError::OperationConflict => Self::precondition_failed(),
AttemptError::Coordination(_) | AttemptError::Durable(_) => Self::unavailable(),
}
}
fn from_rejection(_reason: MutationRejection) -> Self {
Self::precondition_failed()
}
}
impl ResponseError for WorkerApiError {
fn status_code(&self) -> StatusCode {
self.status
}
fn error_response(&self) -> HttpResponse {
HttpResponse::build(self.status)
.content_type("application/problem+json")
.json(ApiErrorDto {
code: self.code,
message: self.message.into(),
request_id: request_id(),
fields: None,
})
}
}
fn request_id() -> OpaqueId {
let mut bytes = [0_u8; 16];
getrandom::fill(&mut bytes).expect("operating system randomness is required");
let mut value = String::from("r");
for byte in bytes {
use std::fmt::Write as _;
write!(value, "{byte:02x}").expect("writing to String cannot fail");
}
OpaqueId::parse(value).expect("generated request identifier is valid")
}

View file

@ -0,0 +1,735 @@
use std::{net::TcpListener, sync::Arc, time::Duration};
use actix_web::{App, HttpResponse, HttpServer, web};
use rcgen::{
BasicConstraints, Certificate, CertificateParams, ExtendedKeyUsagePurpose, IsCa, Issuer,
KeyPair, KeyUsagePurpose,
};
use remora_admission::{
AdmissionConfig, AdmissionDecision, AdmissionService, migrate as admission_migrate,
};
use remora_api::{
AuthenticatedWorker, WorkerApiState, authenticate_rustls_worker, configure_worker_api,
};
use remora_attempt::{AttemptConfig, AttemptService, RecoveryConfig, migrate as attempt_migrate};
use remora_contracts::{
ApiErrorDto, AttemptActivateRequestDto, AttemptCheckpointRequestDto, AttemptFailureOutcomeDto,
AttemptOutcomeDto, AttemptOutcomeRequestDto, AttemptProgressRequestDto, AttemptRenewRequestDto,
AttemptResolvedSourceRequestDto, AttemptRuntimeFailureDto, CheckpointDescriptorKindDto,
CheckpointDescriptorRefDto, ClaimRequestDto, ClaimResponseDto, HttpSourceSubmissionDto,
JobPhaseProjectionDto, JobSubmissionDto, PositiveIntegerDto, ProgressValueDto,
ResolvedSourceDescriptorKindDto, ResolvedSourceDescriptorRefDto, ResultDescriptorDto,
ResultKindDto, SafeIntegerDto, SourceSubmissionDto,
};
use remora_coordination::{CoordinationStore, StoreConfig};
use remora_credentials::migrate as credentials_migrate;
use remora_domain::{
BucketName, JobPhase, OpaqueId, RuntimeFailureClass, StorageScope, TargetPrefix, TargetStoreId,
};
use remora_durable::{PoolConfig, connect, migrate as durable_migrate};
use remora_identity::{DefaultTarget, Principal};
use remora_result::migrate as result_migrate;
use remora_target::attempt_manifest_key;
use reqwest::{Certificate as ClientCertificate, Client, Identity};
use rustls::{
RootCertStore, ServerConfig,
pki_types::{PrivateKeyDer, PrivatePkcs8KeyDer},
server::WebPkiClientVerifier,
};
use sqlx::PgPool;
struct Harness {
pool: PgPool,
coordination: CoordinationStore,
}
impl Harness {
async fn new(name: &str) -> Option<Self> {
let database_url = required_env("DATABASE_URL")?;
let coordination_url = required_env("REMORA_COORDINATION_URL")?;
let pool = connect(
&database_url,
PoolConfig {
max_connections: 16,
acquire_timeout: Duration::from_secs(10),
idle_timeout: Some(Duration::from_secs(30)),
},
)
.await
.unwrap();
durable_migrate(&pool).await.unwrap();
credentials_migrate(&pool).await.unwrap();
admission_migrate(&pool).await.unwrap();
attempt_migrate(&pool).await.unwrap();
result_migrate(&pool).await.unwrap();
sqlx::query("TRUNCATE TABLE jobs CASCADE")
.execute(&pool)
.await
.unwrap();
let coordination = CoordinationStore::connect(
&coordination_url,
format!("{}wp18:{name}:{}:", coordination_namespace(), nonce()),
StoreConfig {
global_limit: 16,
lease_ttl: Duration::from_secs(10),
reserve_ttl: Duration::from_secs(10),
progress_ttl: Duration::from_secs(10),
cancel_ttl: None,
},
)
.await
.unwrap();
Some(Self { pool, coordination })
}
fn service(&self) -> AttemptService {
AttemptService::new(
self.pool.clone(),
self.coordination.clone(),
AttemptConfig {
max_attempts: 3,
local_limit: 8,
},
)
.unwrap()
}
async fn job(&self) -> OpaqueId {
let store = TargetStoreId::parse("primary").unwrap();
let bucket = BucketName::parse("remora-results").unwrap();
let prefix = TargetPrefix::parse("users/alice", ["_remora"]).unwrap();
let principal = Principal {
id: unique_id("owner"),
issuer: "https://issuer.example".into(),
subject: "subject".into(),
administrator: false,
scopes: [StorageScope {
store: store.clone(),
bucket: bucket.clone(),
prefix: prefix.clone(),
}]
.into_iter()
.collect(),
};
let service = AdmissionService::new(
self.pool.clone(),
AdmissionConfig {
default_target: DefaultTarget {
store,
bucket,
prefix,
},
reserved_target_segments: vec!["_remora".into()],
policy_version: "worker-api-v1".into(),
},
)
.unwrap();
let outcome = service
.submit(
&principal,
unique_id("submit").as_str(),
JobSubmissionDto {
source: SourceSubmissionDto::Http(HttpSourceSubmissionDto {
url: "https://example.invalid/artifact".into(),
result_path: "artifact.bin".into(),
credential_id: None,
}),
target: None,
label: Some("worker-api-test".into()),
},
)
.await
.unwrap();
let AdmissionDecision::Accepted(job) = outcome.decision else {
panic!("job rejected")
};
self.coordination
.enqueue(job.job_id.as_str())
.await
.unwrap();
job.job_id
}
async fn seal_result(&self, attempt_id: &OpaqueId) -> ResultDescriptorDto {
let manifest_id = unique_id("manifest");
let checksum = format!("sha256:{}", "11".repeat(32));
let key = attempt_manifest_key(attempt_id, &manifest_id);
sqlx::query(
"INSERT INTO result_inventories (
attempt_id, state, manifest_schema_version, manifest_id, manifest_store,
manifest_bucket, manifest_key, manifest_version_id, manifest_sha256,
manifest_size, member_count, total_bytes
) VALUES ($1, 'sealed', 1, $2, 'primary', 'remora-results', $3,
'version-1', $4, 128, 1, 4096)",
)
.bind(attempt_id.as_str())
.bind(manifest_id.as_str())
.bind(&key)
.bind(&checksum)
.execute(&self.pool)
.await
.unwrap();
ResultDescriptorDto {
kind: ResultKindDto::Manifest,
manifest_id,
schema_version: positive(1),
manifest: remora_contracts::ObjectIdentityDto {
store: "primary".into(),
bucket: "remora-results".into(),
key,
version_id: Some("version-1".into()),
checksum,
size: safe(128),
},
member_count: positive(1),
total_bytes: safe(4096),
}
}
}
#[tokio::test]
async fn worker_operations_are_authenticated_fenced_replayable_and_redacted() {
let Some(harness) = Harness::new("flow").await else {
return;
};
let state = WorkerApiState::new(harness.service());
let (base, server, client) = start_authenticated_server(state).await;
let job_id = harness.job().await;
let worker = unique_id("worker");
let claim_operation = unique_id("operation");
let claim = ClaimRequestDto {
operation_id: claim_operation.clone(),
worker_incarnation: worker.clone(),
};
let unauthenticated = actix_web::test::init_service(App::new().route(
"/",
web::post().to(|_: AuthenticatedWorker| async { HttpResponse::NoContent().finish() }),
))
.await;
let response = actix_web::test::call_service(
&unauthenticated,
actix_web::test::TestRequest::post().uri("/").to_request(),
)
.await;
assert_eq!(response.status(), 401);
assert_eq!(
response.headers().get("content-type").unwrap(),
"application/problem+json"
);
let claim_url = format!("{base}/internal/v1/jobs/{job_id}/claim");
let response = client.post(&claim_url).json(&claim).send().await.unwrap();
assert_eq!(response.status(), 200);
let claimed: ClaimResponseDto = response.json().await.unwrap();
assert_eq!(claimed.authority.worker_incarnation, worker);
assert_eq!(claimed.execution.attempt_id, claimed.authority.attempt_id);
assert_eq!(claimed.execution.job_id, job_id);
assert!(
claimed
.execution
.staging_target
.prefix
.contains(claimed.authority.attempt_id.as_str())
);
let changed_claim = client
.post(&claim_url)
.json(&ClaimRequestDto {
operation_id: claim_operation,
worker_incarnation: unique_id("worker"),
})
.send()
.await
.unwrap();
assert_eq!(changed_claim.status(), 409);
assert_eq!(
changed_claim.headers().get("content-type").unwrap(),
"application/problem+json"
);
let attempt_url = format!(
"{base}/internal/v1/attempts/{}",
claimed.authority.attempt_id
);
let activation_operation = unique_id("operation");
let activation = AttemptActivateRequestDto {
operation_id: activation_operation,
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
};
let response = client
.post(format!("{attempt_url}/activate"))
.json(&activation)
.send()
.await
.unwrap();
assert_eq!(response.status(), 200);
let activation_body: serde_json::Value = response.json().await.unwrap();
let activated: remora_contracts::ActivationResponseDto =
serde_json::from_value(activation_body.clone()).unwrap();
assert!(activated.lease.renew_after.as_str() < activated.lease.expires_at.as_str());
let renewal_operation = unique_id("operation");
let renewal = AttemptRenewRequestDto {
operation_id: renewal_operation.clone(),
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
};
let first_renewal = client
.post(format!("{attempt_url}/renew"))
.json(&renewal)
.send()
.await
.unwrap();
assert_eq!(first_renewal.status(), 200);
let first_lease: serde_json::Value = first_renewal.json().await.unwrap();
let replayed_lease: serde_json::Value = client
.post(format!("{attempt_url}/renew"))
.json(&renewal)
.send()
.await
.unwrap()
.json()
.await
.unwrap();
assert_eq!(replayed_lease, first_lease);
let replayed_activation: serde_json::Value = client
.post(format!("{attempt_url}/activate"))
.json(&activation)
.send()
.await
.unwrap()
.json()
.await
.unwrap();
assert_eq!(replayed_activation, activation_body);
let authority_mismatches = [
(
format!("{base}/internal/v1/attempts/{}/renew", unique_id("attempt")),
claimed.authority.fence,
worker.clone(),
),
(
format!("{attempt_url}/renew"),
positive(claimed.authority.fence.get() + 1),
worker.clone(),
),
(
format!("{attempt_url}/renew"),
claimed.authority.fence,
unique_id("worker"),
),
];
for (url, fence, worker_incarnation) in authority_mismatches {
let response = client
.post(url)
.json(&AttemptRenewRequestDto {
operation_id: unique_id("operation"),
fence,
worker_incarnation,
})
.send()
.await
.unwrap();
assert_eq!(response.status(), 409);
}
let resolved_id = unique_id("resolved");
let resolved = AttemptResolvedSourceRequestDto {
operation_id: unique_id("operation"),
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
resolved_source_ref: ResolvedSourceDescriptorRefDto {
id: resolved_id.clone(),
kind: ResolvedSourceDescriptorKindDto::ResolvedSource,
schema_version: positive(1),
byte_length: safe(64),
item_count: Some(safe(1)),
checksum: format!("sha256:{}", "22".repeat(32)),
},
};
assert_eq!(
client
.post(format!("{attempt_url}/resolved-source"))
.json(&resolved)
.send()
.await
.unwrap()
.status(),
204
);
let checkpoint = AttemptCheckpointRequestDto {
operation_id: unique_id("operation"),
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
checkpoint_ref: CheckpointDescriptorRefDto {
id: unique_id("checkpoint"),
kind: CheckpointDescriptorKindDto::Checkpoint,
schema_version: positive(1),
byte_length: safe(32),
item_count: Some(safe(1)),
checksum: format!("sha256:{}", "33".repeat(32)),
},
resolved_source_id: resolved_id,
validated_bytes: safe(1024),
validated_members: safe(1),
};
assert_eq!(
client
.post(format!("{attempt_url}/checkpoint"))
.json(&checkpoint)
.send()
.await
.unwrap()
.status(),
204
);
let progress = AttemptProgressRequestDto {
operation_id: unique_id("operation"),
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
progress: ProgressValueDto {
completed_bytes: safe(2048),
total_bytes: Some(safe(4096)),
completed_members: safe(1),
total_members: Some(safe(1)),
},
};
assert_eq!(
client
.post(format!("{attempt_url}/progress"))
.json(&progress)
.send()
.await
.unwrap()
.status(),
204
);
let replayed_claim: ClaimResponseDto = client
.post(&claim_url)
.json(&claim)
.send()
.await
.unwrap()
.json()
.await
.unwrap();
assert_eq!(
serde_json::to_value(replayed_claim).unwrap(),
serde_json::to_value(&claimed).unwrap()
);
let failure = AttemptOutcomeRequestDto {
operation_id: unique_id("operation"),
fence: claimed.authority.fence,
worker_incarnation: worker.clone(),
outcome: AttemptOutcomeDto::Failure(AttemptFailureOutcomeDto {
failure: AttemptRuntimeFailureDto {
class: RuntimeFailureClass::TransientSource,
message: "secret-value-must-not-escape".into(),
safe_details: None,
},
}),
};
let outcome_url = format!("{attempt_url}/outcome");
let response = client
.post(&outcome_url)
.json(&failure)
.send()
.await
.unwrap();
assert_eq!(response.status(), 200);
let bytes = response.bytes().await.unwrap();
assert!(!String::from_utf8_lossy(&bytes).contains("secret-value-must-not-escape"));
let projection: JobPhaseProjectionDto = serde_json::from_slice(&bytes).unwrap();
assert_eq!(projection.phase, JobPhase::RetryWait);
let report = harness
.service()
.reconcile_once(&RecoveryConfig {
retry_delay: Duration::ZERO,
cleanup_claim_ttl: Duration::from_secs(1),
batch_size: 128,
})
.await
.unwrap();
assert_eq!(report.retries_scheduled, 1);
let replay = client
.post(&outcome_url)
.json(&failure)
.send()
.await
.unwrap();
assert_eq!(replay.status(), 200);
assert_eq!(replay.bytes().await.unwrap(), bytes);
let cross_kind = client
.post(format!("{attempt_url}/progress"))
.json(&AttemptProgressRequestDto {
operation_id: renewal_operation,
fence: claimed.authority.fence,
worker_incarnation: worker,
progress: ProgressValueDto {
completed_bytes: safe(2048),
total_bytes: Some(safe(4096)),
completed_members: safe(1),
total_members: Some(safe(1)),
},
})
.send()
.await
.unwrap();
assert_eq!(cross_kind.status(), 409);
assert_eq!(
cross_kind.headers().get("content-type").unwrap(),
"application/problem+json"
);
let error: ApiErrorDto = cross_kind.json().await.unwrap();
assert_eq!(error.message, "precondition failed");
server.stop(true).await;
}
#[tokio::test]
async fn success_outcome_publishes_the_exact_manifest_and_replays() {
let Some(harness) = Harness::new("success").await else {
return;
};
let state = WorkerApiState::new(harness.service());
let (base, server, client) = start_authenticated_server(state).await;
let job_id = harness.job().await;
let worker = unique_id("worker");
let claim: ClaimResponseDto = client
.post(format!("{base}/internal/v1/jobs/{job_id}/claim"))
.json(&ClaimRequestDto {
operation_id: unique_id("operation"),
worker_incarnation: worker.clone(),
})
.send()
.await
.unwrap()
.json()
.await
.unwrap();
let attempt_url = format!("{base}/internal/v1/attempts/{}", claim.authority.attempt_id);
assert_eq!(
client
.post(format!("{attempt_url}/activate"))
.json(&AttemptActivateRequestDto {
operation_id: unique_id("operation"),
fence: claim.authority.fence,
worker_incarnation: worker.clone(),
})
.send()
.await
.unwrap()
.status(),
200
);
let result = harness.seal_result(&claim.authority.attempt_id).await;
let operation = unique_id("operation");
let outcome = AttemptOutcomeRequestDto {
operation_id: operation,
fence: claim.authority.fence,
worker_incarnation: worker,
outcome: AttemptOutcomeDto::Success(remora_contracts::AttemptSuccessOutcomeDto { result }),
};
let outcome_url = format!("{attempt_url}/outcome");
for _ in 0..2 {
let response = client
.post(&outcome_url)
.json(&outcome)
.send()
.await
.unwrap();
assert_eq!(response.status(), 200);
assert_eq!(
response
.json::<JobPhaseProjectionDto>()
.await
.unwrap()
.phase,
JobPhase::Succeeded
);
}
server.stop(true).await;
}
#[tokio::test]
async fn mutual_tls_requires_a_trusted_client_certificate() {
let certificates = certificates();
let listener = TcpListener::bind("127.0.0.1:0").unwrap();
let address = listener.local_addr().unwrap();
let server = HttpServer::new(|| {
App::new().route(
"/identity",
web::get().to(|identity: AuthenticatedWorker| async move {
HttpResponse::Ok().body(identity.identity.to_string())
}),
)
})
.on_connect(authenticate_rustls_worker)
.listen_rustls_0_23(listener, certificates.server_config)
.unwrap()
.run();
let handle = server.handle();
tokio::spawn(server);
let root = ClientCertificate::from_der(certificates.ca.der()).unwrap();
let untrusted = Client::builder()
.add_root_certificate(root.clone())
.build()
.unwrap();
assert!(
untrusted
.get(format!("https://localhost:{}/identity", address.port()))
.send()
.await
.is_err()
);
let mut client_pem = certificates.client.pem();
client_pem.push_str(&certificates.client_key.serialize_pem());
let trusted = Client::builder()
.add_root_certificate(root)
.identity(Identity::from_pem(client_pem.as_bytes()).unwrap())
.build()
.unwrap();
let response = trusted
.get(format!("https://localhost:{}/identity", address.port()))
.send()
.await
.unwrap();
assert_eq!(response.status(), 200);
let expected = certificate_identity(certificates.client.der().as_ref());
assert_eq!(response.text().await.unwrap(), expected);
handle.stop(true).await;
}
async fn start_authenticated_server(
state: WorkerApiState,
) -> (String, actix_web::dev::ServerHandle, Client) {
let certificates = certificates();
let client = trusted_client(&certificates);
let listener = TcpListener::bind("127.0.0.1:0").unwrap();
let address = listener.local_addr().unwrap();
let server = HttpServer::new(move || {
App::new()
.app_data(web::Data::new(state.clone()))
.configure(configure_worker_api)
})
.on_connect(authenticate_rustls_worker)
.listen_rustls_0_23(listener, certificates.server_config)
.unwrap()
.run();
let handle = server.handle();
tokio::spawn(server);
(
format!("https://localhost:{}", address.port()),
handle,
client,
)
}
fn trusted_client(certificates: &Certificates) -> Client {
let root = ClientCertificate::from_der(certificates.ca.der()).unwrap();
let mut client_pem = certificates.client.pem();
client_pem.push_str(&certificates.client_key.serialize_pem());
Client::builder()
.add_root_certificate(root)
.identity(Identity::from_pem(client_pem.as_bytes()).unwrap())
.build()
.unwrap()
}
fn certificate_identity(certificate: &[u8]) -> String {
use sha2::{Digest, Sha256};
let mut value = String::from("w");
for byte in Sha256::digest(certificate) {
use std::fmt::Write as _;
write!(value, "{byte:02x}").unwrap();
}
value
}
struct Certificates {
ca: Certificate,
client: Certificate,
client_key: KeyPair,
server_config: ServerConfig,
}
fn certificates() -> Certificates {
let _ = rustls::crypto::ring::default_provider().install_default();
let mut ca_params = CertificateParams::new(Vec::new()).unwrap();
ca_params.is_ca = IsCa::Ca(BasicConstraints::Unconstrained);
ca_params.key_usages = vec![
KeyUsagePurpose::DigitalSignature,
KeyUsagePurpose::KeyCertSign,
KeyUsagePurpose::CrlSign,
];
let ca_key = KeyPair::generate().unwrap();
let ca = ca_params.self_signed(&ca_key).unwrap();
let issuer = Issuer::new(ca_params, ca_key);
let server_key = KeyPair::generate().unwrap();
let mut server_params = CertificateParams::new(vec!["localhost".into()]).unwrap();
server_params.key_usages = vec![KeyUsagePurpose::DigitalSignature];
server_params.extended_key_usages = vec![ExtendedKeyUsagePurpose::ServerAuth];
let server = server_params.signed_by(&server_key, &issuer).unwrap();
let client_key = KeyPair::generate().unwrap();
let mut client_params = CertificateParams::new(Vec::new()).unwrap();
client_params.key_usages = vec![KeyUsagePurpose::DigitalSignature];
client_params.extended_key_usages = vec![ExtendedKeyUsagePurpose::ClientAuth];
let client = client_params.signed_by(&client_key, &issuer).unwrap();
let mut roots = RootCertStore::empty();
roots.add(ca.der().clone()).unwrap();
let verifier = WebPkiClientVerifier::builder(Arc::new(roots))
.build()
.unwrap();
let server_key_der = PrivateKeyDer::Pkcs8(PrivatePkcs8KeyDer::from(server_key.serialize_der()));
let server_config = ServerConfig::builder()
.with_client_cert_verifier(verifier)
.with_single_cert(vec![server.der().clone()], server_key_der)
.unwrap();
Certificates {
ca,
client,
client_key,
server_config,
}
}
fn required_env(name: &str) -> Option<String> {
match std::env::var(name) {
Ok(value) if !value.is_empty() => Some(value),
_ if std::env::var_os("REMORA_WORKER_API_REQUIRE").is_some() => {
panic!("{name} is required")
}
_ => None,
}
}
fn coordination_namespace() -> String {
std::env::var("REMORA_COORDINATION_NAMESPACE").unwrap_or_else(|_| "remora:test:".into())
}
fn nonce() -> u128 {
use std::sync::atomic::{AtomicU64, Ordering};
static COUNTER: AtomicU64 = AtomicU64::new(0);
(u128::from(std::process::id()) << 64) | u128::from(COUNTER.fetch_add(1, Ordering::Relaxed))
}
fn unique_id(prefix: &str) -> OpaqueId {
OpaqueId::parse(format!("{prefix}{:032x}", nonce())).unwrap()
}
fn positive(value: u64) -> PositiveIntegerDto {
PositiveIntegerDto::new(value).unwrap()
}
fn safe(value: u64) -> SafeIntegerDto {
SafeIntegerDto::new(value).unwrap()
}

View file

@ -0,0 +1,2 @@
ALTER TABLE attempts
ADD COLUMN execution_envelope JSONB;

View file

@ -20,6 +20,7 @@ pub use recovery_types::{
}; };
pub use service::{AttemptConfig, AttemptService}; pub use service::{AttemptConfig, AttemptService};
pub use types::{ pub use types::{
ActivationOutcome, AttemptAuthority, AttemptFaultInjector, AttemptFaultPoint, AttemptState, ActivationMutationOutcome, ActivationOutcome, AttemptAuthority, AttemptFaultInjector,
ClaimOutcome, ClaimRejection, RecoveryOutcome, RenewalOutcome, AttemptFaultPoint, AttemptState, ClaimOutcome, ClaimRejection, ExecutionEnvelopeRecord,
LeaseProjection, OutcomePublication, RecoveryOutcome, RenewalMutationOutcome, RenewalOutcome,
}; };

View file

@ -14,7 +14,7 @@ use time::OffsetDateTime;
use crate::{ use crate::{
AttemptAuthority, AttemptError, AttemptMutationOutcome, AttemptService, CheckpointProjection, AttemptAuthority, AttemptError, AttemptMutationOutcome, AttemptService, CheckpointProjection,
CheckpointRecord, CommittedResultRecord, DurableDescriptor, FailureRecord, CheckpointRecord, CommittedResultRecord, DurableDescriptor, FailureRecord,
JobMutationProjection, MutationRejection, ProgressRecord, JobMutationProjection, MutationRejection, OutcomePublication, ProgressRecord,
}; };
impl AttemptService { impl AttemptService {
@ -64,6 +64,18 @@ impl AttemptService {
.await .await
} }
pub async fn publish_success_operation(
&self,
authority: &AttemptAuthority,
operation_id: &OpaqueId,
result: PublicationInput,
) -> Result<OutcomePublication, AttemptError> {
let outcome = self
.publish_success(authority, operation_id, result)
.await?;
self.publication_outcome(operation_id, outcome).await
}
pub async fn persist_checkpoint( pub async fn persist_checkpoint(
&self, &self,
authority: &AttemptAuthority, authority: &AttemptAuthority,
@ -124,6 +136,16 @@ impl AttemptService {
.await .await
} }
pub async fn publish_failure_operation(
&self,
authority: &AttemptAuthority,
operation_id: &OpaqueId,
class: RuntimeFailureClass,
) -> Result<OutcomePublication, AttemptError> {
let outcome = self.publish_failure(authority, operation_id, class).await?;
self.publication_outcome(operation_id, outcome).await
}
pub async fn report_progress( pub async fn report_progress(
&self, &self,
authority: &AttemptAuthority, authority: &AttemptAuthority,
@ -227,7 +249,7 @@ impl AttemptService {
.execute(tx.as_mut()) .execute(tx.as_mut())
.await?; .await?;
enqueue(tx, "job.updated", json!({ "job_id": authority.job_id })).await?; enqueue(tx, "job.updated", json!({ "job_id": authority.job_id })).await?;
accepted() accepted_phase(JobPhase::Succeeded)
}) })
}, },
) )
@ -307,7 +329,11 @@ impl AttemptService {
json!({ "job_id": authority.job_id }), json!({ "job_id": authority.job_id }),
) )
.await?; .await?;
accepted() accepted_phase(if retry {
JobPhase::RetryWait
} else {
JobPhase::Failed
})
}) })
}, },
) )
@ -329,6 +355,43 @@ impl AttemptService {
row.map(ProjectionRow::try_into_projection).transpose() row.map(ProjectionRow::try_into_projection).transpose()
} }
async fn publication_outcome(
&self,
operation_id: &OpaqueId,
outcome: AttemptMutationOutcome,
) -> Result<OutcomePublication, AttemptError> {
match outcome {
AttemptMutationOutcome::Rejected {
reason,
disposition,
} => Ok(OutcomePublication::Rejected {
reason,
disposition,
}),
AttemptMutationOutcome::Accepted(disposition) => {
let response = sqlx::query_scalar::<_, Json<Value>>(
"SELECT response_body FROM mutation_records
WHERE scope_subject = $1 AND scope_operation = 'attempt.mutation'
AND operation_key = $2",
)
.bind(self.operation_subject.as_str())
.bind(operation_id.as_str())
.fetch_one(&self.pool)
.await
.map_err(DurableError::from)?;
let phase = serde_json::from_value(
response
.0
.get("phase")
.cloned()
.ok_or_else(|| AttemptError::Invalid("missing outcome phase".into()))?,
)
.map_err(|error| AttemptError::Invalid(error.to_string()))?;
Ok(OutcomePublication::Accepted { disposition, phase })
}
}
}
async fn execute_durable_mutation<T, D>( async fn execute_durable_mutation<T, D>(
&self, &self,
authority: &AttemptAuthority, authority: &AttemptAuthority,
@ -423,6 +486,13 @@ fn accepted() -> Result<(MutationStatus, Value), DurableError> {
Ok((MutationStatus::Accepted, json!({ "accepted": true }))) Ok((MutationStatus::Accepted, json!({ "accepted": true })))
} }
fn accepted_phase(phase: JobPhase) -> Result<(MutationStatus, Value), DurableError> {
Ok((
MutationStatus::Accepted,
json!({ "accepted": true, "phase": phase }),
))
}
fn rejected(reason: MutationRejection) -> Result<(MutationStatus, Value), DurableError> { fn rejected(reason: MutationRejection) -> Result<(MutationStatus, Value), DurableError> {
Ok((MutationStatus::Rejected, json!({ "reason": reason }))) Ok((MutationStatus::Rejected, json!({ "reason": reason })))
} }

View file

@ -14,9 +14,10 @@ use sqlx::{PgPool, types::Json};
use time::OffsetDateTime; use time::OffsetDateTime;
use crate::{ use crate::{
ActivationOutcome, AttemptAuthority, AttemptError, AttemptFaultInjector, AttemptFaultPoint, ActivationMutationOutcome, ActivationOutcome, AttemptAuthority, AttemptError,
AttemptState, CheckpointProjection, ClaimOutcome, ClaimRejection, RecoveryOutcome, AttemptFaultInjector, AttemptFaultPoint, AttemptState, CheckpointProjection, ClaimOutcome,
RenewalOutcome, ClaimRejection, ExecutionEnvelopeRecord, LeaseProjection, RecoveryOutcome,
RenewalMutationOutcome, RenewalOutcome,
}; };
struct NoFaults; struct NoFaults;
@ -179,6 +180,23 @@ impl AttemptService {
authority: &AttemptAuthority, authority: &AttemptAuthority,
operation_id: &OpaqueId, operation_id: &OpaqueId,
) -> Result<ActivationOutcome, AttemptError> { ) -> Result<ActivationOutcome, AttemptError> {
Ok(
match self.activate_operation(authority, operation_id).await? {
ActivationMutationOutcome::Activated { disposition, .. } => {
ActivationOutcome::Activated(disposition)
}
ActivationMutationOutcome::PreconditionFailed => {
ActivationOutcome::PreconditionFailed
}
},
)
}
pub async fn activate_operation(
&self,
authority: &AttemptAuthority,
operation_id: &OpaqueId,
) -> Result<ActivationMutationOutcome, AttemptError> {
validate_authority(authority)?; validate_authority(authority)?;
let request_fingerprint = fingerprint(&ActivationRequest { let request_fingerprint = fingerprint(&ActivationRequest {
job_id: &authority.job_id, job_id: &authority.job_id,
@ -203,16 +221,17 @@ impl AttemptService {
request_fingerprint, request_fingerprint,
move |tx| { move |tx| {
Box::pin(async move { Box::pin(async move {
let accepted = let lease =
activate_attempt(tx, &coordination, &authority_for_effect, faults.as_ref()) activate_attempt(tx, &coordination, &authority_for_effect, faults.as_ref())
.await?; .await?;
Ok(( Ok((
if accepted { if lease.is_some() {
MutationStatus::Accepted MutationStatus::Accepted
} else { } else {
MutationStatus::Rejected MutationStatus::Rejected
}, },
json!({ "activated": accepted }), serde_json::to_value(ActivationStored { lease })
.map_err(DurableError::effect)?,
)) ))
}) })
}, },
@ -225,7 +244,14 @@ impl AttemptService {
.exact_authority_is_executing_and_live(authority) .exact_authority_is_executing_and_live(authority)
.await? .await?
{ {
return Ok(ActivationOutcome::Activated(MutationDisposition::First)); let lease = self
.live_lease(authority)
.await?
.ok_or_else(|| AttemptError::Invalid("live lease disappeared".into()))?;
return Ok(ActivationMutationOutcome::Activated {
disposition: MutationDisposition::First,
lease,
});
} }
self.release_exact(authority).await; self.release_exact(authority).await;
return Err(error.into()); return Err(error.into());
@ -237,16 +263,24 @@ impl AttemptService {
} }
if outcome.status == MutationStatus::Rejected { if outcome.status == MutationStatus::Rejected {
self.release_exact(authority).await; self.release_exact(authority).await;
return Ok(ActivationOutcome::PreconditionFailed); return Ok(ActivationMutationOutcome::PreconditionFailed);
} }
if !self if !self
.exact_authority_is_executing_and_live(authority) .exact_authority_is_executing_and_live(authority)
.await? .await?
{ {
self.release_exact(authority).await; self.release_exact(authority).await;
return Ok(ActivationOutcome::PreconditionFailed); return Ok(ActivationMutationOutcome::PreconditionFailed);
} }
Ok(ActivationOutcome::Activated(outcome.disposition)) let stored: ActivationStored = serde_json::from_value(outcome.response)
.map_err(|error| AttemptError::Invalid(error.to_string()))?;
let lease = stored
.lease
.ok_or_else(|| AttemptError::Invalid("accepted activation omitted its lease".into()))?;
Ok(ActivationMutationOutcome::Activated {
disposition: outcome.disposition,
lease,
})
} }
pub async fn renew( pub async fn renew(
@ -269,6 +303,103 @@ impl AttemptService {
Ok(RenewalOutcome::Renewed) Ok(RenewalOutcome::Renewed)
} }
pub async fn renew_operation(
&self,
authority: &AttemptAuthority,
operation_id: &OpaqueId,
) -> Result<RenewalMutationOutcome, AttemptError> {
validate_authority(authority)?;
let request_fingerprint = fingerprint(&ActivationRequest {
job_id: &authority.job_id,
attempt_id: &authority.attempt_id,
fence: authority.fence,
worker_incarnation: &authority.worker_incarnation,
})
.map_err(|error| AttemptError::Invalid(error.to_string()))?;
let eligibility = authority_binding(authority);
let authority = authority.clone();
let coordination = self.coordination.clone();
let faults = self.faults.clone();
let outcome = execute_mutation(
&self.pool,
MutationScope {
subject: &self.operation_subject,
operation: "attempt.mutation",
key: operation_id.as_str(),
},
"renew_attempt",
&eligibility,
request_fingerprint,
move |tx| {
Box::pin(async move {
let Some(row) = lock_authority(tx, &authority.attempt_id).await? else {
return Ok((
MutationStatus::Rejected,
serde_json::to_value(RenewalMutationOutcome::AuthorityLost {
disposition: None,
})
.map_err(DurableError::effect)?,
));
};
if row.state != AttemptState::Executing
|| !row.matches(
&authority.job_id,
authority.fence,
&authority.worker_incarnation,
)
{
return Ok((
MutationStatus::Rejected,
serde_json::to_value(RenewalMutationOutcome::AuthorityLost {
disposition: None,
})
.map_err(DurableError::effect)?,
));
}
faults.hit(AttemptFaultPoint::RenewalBeforeLease);
let expected = lease_owner(&authority).map_err(DurableError::effect)?;
let renewed = coordination
.renew(authority.job_id.as_str(), &expected)
.await
.map_err(DurableError::effect)?;
faults.hit(AttemptFaultPoint::RenewalAfterLease);
let live = coordination
.get_live_lease(authority.job_id.as_str())
.await
.map_err(DurableError::effect)?;
let response = match live
.filter(|lease| renewed.ok() && lease.record.owner() == expected)
{
Some(lease) => RenewalMutationOutcome::Renewed {
lease: LeaseProjection {
expires_at_ms: lease.record.expires_at,
observed_at_ms: lease.now_ms,
},
disposition: None,
},
None => RenewalMutationOutcome::AuthorityLost { disposition: None },
};
let status = if matches!(response, RenewalMutationOutcome::Renewed { .. }) {
MutationStatus::Accepted
} else {
MutationStatus::Rejected
};
Ok((
status,
serde_json::to_value(response).map_err(DurableError::effect)?,
))
})
},
)
.await?;
if outcome.disposition == MutationDisposition::Conflict {
return Err(AttemptError::OperationConflict);
}
let response: RenewalMutationOutcome = serde_json::from_value(outcome.response)
.map_err(|error| AttemptError::Invalid(error.to_string()))?;
Ok(response.with_disposition(outcome.disposition))
}
pub async fn recover_inactive( pub async fn recover_inactive(
&self, &self,
authority: &AttemptAuthority, authority: &AttemptAuthority,
@ -372,6 +503,45 @@ impl AttemptService {
.map_err(Into::into) .map_err(Into::into)
} }
pub async fn historical_authority(
&self,
attempt_id: &OpaqueId,
) -> Result<Option<AttemptAuthority>, AttemptError> {
load_historical_authority(&self.pool, attempt_id)
.await
.map_err(Into::into)
}
pub async fn execution_envelope(
&self,
attempt_id: &OpaqueId,
) -> Result<Option<ExecutionEnvelopeRecord>, AttemptError> {
let value = sqlx::query_scalar::<_, Option<Json<ExecutionEnvelopeRecord>>>(
"SELECT execution_envelope FROM attempts WHERE attempt_id = $1",
)
.bind(attempt_id.as_str())
.fetch_optional(&self.pool)
.await?
.flatten();
Ok(value.map(|value| value.0))
}
pub async fn live_lease(
&self,
authority: &AttemptAuthority,
) -> Result<Option<LeaseProjection>, AttemptError> {
let expected = lease_owner(authority)?;
Ok(self
.coordination
.get_live_lease(authority.job_id.as_str())
.await?
.filter(|lease| lease.record.owner() == expected)
.map(|lease| LeaseProjection {
expires_at_ms: lease.record.expires_at,
observed_at_ms: lease.now_ms,
}))
}
pub fn release_local(&self, job_id: &OpaqueId) -> bool { pub fn release_local(&self, job_id: &OpaqueId) -> bool {
self.local_permits.release_all(job_id.as_str()) self.local_permits.release_all(job_id.as_str())
} }
@ -451,6 +621,11 @@ struct ActivationRequest<'a> {
worker_incarnation: &'a OpaqueId, worker_incarnation: &'a OpaqueId,
} }
#[derive(serde::Deserialize, Serialize)]
struct ActivationStored {
lease: Option<LeaseProjection>,
}
struct AuthorityRow { struct AuthorityRow {
job_id: OpaqueId, job_id: OpaqueId,
fence: u64, fence: u64,
@ -534,11 +709,14 @@ async fn authorize_attempt(
_ => 0, _ => 0,
}; };
let now = OffsetDateTime::now_utc(); let now = OffsetDateTime::now_utc();
let envelope =
load_execution_envelope_record(tx, input.job_id, input.attempt_id, fence, input.worker)
.await?;
sqlx::query( sqlx::query(
"INSERT INTO attempts ( "INSERT INTO attempts (
attempt_id, job_id, fence, worker_incarnation, state, attempt_id, job_id, fence, worker_incarnation, state,
initial_progress, authorized_at, executing_at, reaped_at initial_progress, authorized_at, executing_at, reaped_at, execution_envelope
) VALUES ($1, $2, $3, $4, 'authorized', $5, $6, NULL, NULL)", ) VALUES ($1, $2, $3, $4, 'authorized', $5, $6, NULL, NULL, $7)",
) )
.bind(input.attempt_id.as_str()) .bind(input.attempt_id.as_str())
.bind(input.job_id.as_str()) .bind(input.job_id.as_str())
@ -546,6 +724,7 @@ async fn authorize_attempt(
.bind(input.worker.as_str()) .bind(input.worker.as_str())
.bind(i64::try_from(initial_progress).map_err(DurableError::effect)?) .bind(i64::try_from(initial_progress).map_err(DurableError::effect)?)
.bind(now) .bind(now)
.bind(Json(&envelope))
.execute(tx.as_mut()) .execute(tx.as_mut())
.await?; .await?;
sqlx::query( sqlx::query(
@ -572,21 +751,79 @@ async fn authorize_attempt(
}) })
} }
async fn load_execution_envelope_record(
tx: &mut remora_durable::PgTx<'_>,
job_id: &OpaqueId,
attempt_id: &OpaqueId,
fence: u64,
worker: &OpaqueId,
) -> Result<ExecutionEnvelopeRecord, DurableError> {
type Row = (
Json<serde_json::Value>,
String,
String,
String,
Json<Vec<String>>,
String,
Option<String>,
Option<String>,
Option<Json<crate::DurableDescriptor>>,
Option<Json<CheckpointProjection>>,
);
let row = sqlx::query_as::<_, Row>(
"SELECT j.source_projection, j.sensitive_source_ref, j.target_store,
j.target_bucket, j.target_prefix, j.execution_grant_id,
g.credential_id, g.credential_version_id, j.resolved_source, j.checkpoint
FROM jobs j
JOIN execution_grants g ON g.grant_id = j.execution_grant_id
WHERE j.job_id = $1",
)
.bind(job_id.as_str())
.fetch_one(tx.as_mut())
.await?;
let grant_id = OpaqueId::parse(row.5).map_err(DurableError::effect)?;
Ok(ExecutionEnvelopeRecord {
id: grant_id.clone(),
job_id: job_id.clone(),
attempt_id: attempt_id.clone(),
fence,
worker_incarnation: worker.clone(),
source: row.0.0,
sensitive_source_ref: OpaqueId::parse(row.1).map_err(DurableError::effect)?,
credential_id: row
.6
.map(OpaqueId::parse)
.transpose()
.map_err(DurableError::effect)?,
credential_version_id: row
.7
.map(OpaqueId::parse)
.transpose()
.map_err(DurableError::effect)?,
resolved_source: row.8.map(|value| value.0),
target_store: row.2,
target_bucket: row.3,
target_prefix: row.4.0,
execution_grant_id: grant_id,
checkpoint: row.9.map(|value| value.0),
})
}
async fn activate_attempt( async fn activate_attempt(
tx: &mut remora_durable::PgTx<'_>, tx: &mut remora_durable::PgTx<'_>,
coordination: &CoordinationStore, coordination: &CoordinationStore,
authority: &AttemptAuthority, authority: &AttemptAuthority,
faults: &dyn AttemptFaultInjector, faults: &dyn AttemptFaultInjector,
) -> Result<bool, DurableError> { ) -> Result<Option<LeaseProjection>, DurableError> {
let Some(row) = lock_authority(tx, &authority.attempt_id).await? else { let Some(row) = lock_authority(tx, &authority.attempt_id).await? else {
return Ok(false); return Ok(None);
}; };
if !row.matches( if !row.matches(
&authority.job_id, &authority.job_id,
authority.fence, authority.fence,
&authority.worker_incarnation, &authority.worker_incarnation,
) { ) {
return Ok(false); return Ok(None);
} }
if row.state == AttemptState::Executing { if row.state == AttemptState::Executing {
let live = coordination let live = coordination
@ -595,16 +832,28 @@ async fn activate_attempt(
.map_err(DurableError::effect)?; .map_err(DurableError::effect)?;
let expected = lease_owner(authority).map_err(DurableError::effect)?; let expected = lease_owner(authority).map_err(DurableError::effect)?;
if !live.is_some_and(|lease| lease.record.owner() == expected) { if !live.is_some_and(|lease| lease.record.owner() == expected) {
return Ok(false); return Ok(None);
} }
return Ok(coordination if !coordination
.renew(authority.job_id.as_str(), &expected) .renew(authority.job_id.as_str(), &expected)
.await .await
.map_err(DurableError::effect)? .map_err(DurableError::effect)?
.ok()); .ok()
{
return Ok(None);
}
return Ok(coordination
.get_live_lease(authority.job_id.as_str())
.await
.map_err(DurableError::effect)?
.filter(|lease| lease.record.owner() == expected)
.map(|lease| LeaseProjection {
expires_at_ms: lease.record.expires_at,
observed_at_ms: lease.now_ms,
}));
} }
if row.state != AttemptState::Authorized { if row.state != AttemptState::Authorized {
return Ok(false); return Ok(None);
} }
faults.hit(AttemptFaultPoint::ActivationBeforeLease); faults.hit(AttemptFaultPoint::ActivationBeforeLease);
let owner = lease_owner(authority).map_err(DurableError::effect)?; let owner = lease_owner(authority).map_err(DurableError::effect)?;
@ -620,7 +869,7 @@ async fn activate_attempt(
.await .await
.map_err(DurableError::effect)?; .map_err(DurableError::effect)?;
if !coordinated.ok() { if !coordinated.ok() {
return Ok(false); return Ok(None);
} }
faults.hit(AttemptFaultPoint::ActivationAfterLease); faults.hit(AttemptFaultPoint::ActivationAfterLease);
let live = coordination let live = coordination
@ -628,9 +877,9 @@ async fn activate_attempt(
.await .await
.map_err(DurableError::effect)?; .map_err(DurableError::effect)?;
let expected = lease_owner(authority).map_err(DurableError::effect)?; let expected = lease_owner(authority).map_err(DurableError::effect)?;
if !live.is_some_and(|lease| lease.record.owner() == expected) { let Some(live) = live.filter(|lease| lease.record.owner() == expected) else {
return Ok(false); return Ok(None);
} };
let now = OffsetDateTime::now_utc(); let now = OffsetDateTime::now_utc();
sqlx::query( sqlx::query(
"UPDATE attempts SET state = 'executing', executing_at = $2 "UPDATE attempts SET state = 'executing', executing_at = $2
@ -640,7 +889,10 @@ async fn activate_attempt(
.bind(now) .bind(now)
.execute(tx.as_mut()) .execute(tx.as_mut())
.await?; .await?;
Ok(true) Ok(Some(LeaseProjection {
expires_at_ms: live.record.expires_at,
observed_at_ms: live.now_ms,
}))
} }
async fn lock_authority( async fn lock_authority(
@ -692,6 +944,30 @@ async fn load_authority(
.transpose() .transpose()
} }
async fn load_historical_authority(
pool: &PgPool,
attempt_id: &OpaqueId,
) -> Result<Option<AttemptAuthority>, DurableError> {
let row = sqlx::query_as::<_, (String, String, i64, String, String)>(
"SELECT job_id, attempt_id, fence, worker_incarnation, state
FROM attempts WHERE attempt_id = $1",
)
.bind(attempt_id.as_str())
.fetch_optional(pool)
.await?;
row.map(|(job, attempt, fence, worker, state)| {
Ok(AttemptAuthority {
job_id: OpaqueId::parse(job).map_err(DurableError::effect)?,
attempt_id: OpaqueId::parse(attempt).map_err(DurableError::effect)?,
fence: u64::try_from(fence).map_err(DurableError::effect)?,
worker_incarnation: OpaqueId::parse(worker).map_err(DurableError::effect)?,
state: AttemptState::parse(&state)
.ok_or_else(|| DurableError::Config(format!("unknown attempt state '{state}'")))?,
})
})
.transpose()
}
fn authority_row(row: (String, i64, String, String)) -> Result<AuthorityRow, DurableError> { fn authority_row(row: (String, i64, String, String)) -> Result<AuthorityRow, DurableError> {
Ok(AuthorityRow { Ok(AuthorityRow {
job_id: OpaqueId::parse(row.0).map_err(DurableError::effect)?, job_id: OpaqueId::parse(row.0).map_err(DurableError::effect)?,

View file

@ -1,4 +1,4 @@
use remora_domain::OpaqueId; use remora_domain::{JobPhase, OpaqueId};
use remora_durable::MutationDisposition; use remora_durable::MutationDisposition;
use serde::{Deserialize, Serialize}; use serde::{Deserialize, Serialize};
@ -95,12 +95,83 @@ pub enum ActivationOutcome {
PreconditionFailed, PreconditionFailed,
} }
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum ActivationMutationOutcome {
Activated {
disposition: MutationDisposition,
lease: LeaseProjection,
},
PreconditionFailed,
}
#[derive(Clone, Copy, Debug, Eq, PartialEq)] #[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum RenewalOutcome { pub enum RenewalOutcome {
Renewed, Renewed,
AuthorityLost, AuthorityLost,
} }
#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
pub struct LeaseProjection {
pub expires_at_ms: u64,
pub observed_at_ms: u64,
}
#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(tag = "status", rename_all = "snake_case")]
pub enum RenewalMutationOutcome {
Renewed {
lease: LeaseProjection,
#[serde(skip)]
disposition: Option<MutationDisposition>,
},
AuthorityLost {
#[serde(skip)]
disposition: Option<MutationDisposition>,
},
}
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum OutcomePublication {
Accepted {
disposition: MutationDisposition,
phase: JobPhase,
},
Rejected {
reason: crate::MutationRejection,
disposition: MutationDisposition,
},
}
#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
pub struct ExecutionEnvelopeRecord {
pub id: OpaqueId,
pub job_id: OpaqueId,
pub attempt_id: OpaqueId,
pub fence: u64,
pub worker_incarnation: OpaqueId,
pub source: serde_json::Value,
pub sensitive_source_ref: OpaqueId,
pub credential_id: Option<OpaqueId>,
pub credential_version_id: Option<OpaqueId>,
pub resolved_source: Option<crate::DurableDescriptor>,
pub target_store: String,
pub target_bucket: String,
pub target_prefix: Vec<String>,
pub execution_grant_id: OpaqueId,
pub checkpoint: Option<crate::CheckpointProjection>,
}
impl RenewalMutationOutcome {
pub(crate) fn with_disposition(mut self, value: MutationDisposition) -> Self {
match &mut self {
Self::Renewed { disposition, .. } | Self::AuthorityLost { disposition } => {
*disposition = Some(value);
}
}
self
}
}
#[derive(Clone, Copy, Debug, Eq, PartialEq)] #[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub enum RecoveryOutcome { pub enum RecoveryOutcome {
StillOwned, StillOwned,

View file

@ -315,6 +315,8 @@ pub enum SafeDetailValueDto {
#[serde(deny_unknown_fields)] #[serde(deny_unknown_fields)]
pub struct ResultDescriptorDto { pub struct ResultDescriptorDto {
pub kind: ResultKindDto, pub kind: ResultKindDto,
pub manifest_id: OpaqueId,
pub schema_version: PositiveIntegerDto,
pub manifest: ObjectIdentityDto, pub manifest: ObjectIdentityDto,
pub member_count: PositiveIntegerDto, pub member_count: PositiveIntegerDto,
pub total_bytes: SafeIntegerDto, pub total_bytes: SafeIntegerDto,
@ -339,6 +341,7 @@ pub struct ObjectIdentityDto {
)] )]
pub version_id: Option<String>, pub version_id: Option<String>,
pub checksum: String, pub checksum: String,
pub size: SafeIntegerDto,
} }
#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)] #[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]

View file

@ -157,6 +157,9 @@ pub struct AttemptCheckpointRequestDto {
pub fence: PositiveIntegerDto, pub fence: PositiveIntegerDto,
pub worker_incarnation: OpaqueId, pub worker_incarnation: OpaqueId,
pub checkpoint_ref: CheckpointDescriptorRefDto, pub checkpoint_ref: CheckpointDescriptorRefDto,
pub resolved_source_id: OpaqueId,
pub validated_bytes: SafeIntegerDto,
pub validated_members: SafeIntegerDto,
} }
#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)] #[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]

View file

@ -1,5 +1,12 @@
use std::{collections::BTreeSet, fs}; use std::{collections::BTreeSet, fs};
use remora_contracts::{
ActivationResponseDto, AttemptActivateRequestDto, AttemptCheckpointRequestDto,
AttemptOutcomeRequestDto, AttemptProgressRequestDto, AttemptRenewRequestDto,
AttemptResolvedSourceRequestDto, ClaimRequestDto, ClaimResponseDto, JobPhaseProjectionDto,
LeaseProjectionDto,
};
use serde::{Serialize, de::DeserializeOwned};
use serde_json::Value; use serde_json::Value;
#[test] #[test]
@ -48,6 +55,273 @@ fn api_contract_documents_parse_and_local_references_resolve() {
validate_openapi_corpus(&component_schema, &corpus, "JobFailure", "job_failures"); validate_openapi_corpus(&component_schema, &corpus, "JobFailure", "job_failures");
} }
#[test]
fn worker_routes_and_dtos_conform_to_the_mutable_openapi_contract() {
let openapi: Value =
serde_yaml_ng::from_str(&fs::read_to_string("../../docs/api/openapi.yaml").unwrap())
.unwrap();
let component_schema = openapi_component_schema(&openapi);
let mutation = serde_json::json!({
"operation_id": "operation_0001",
"fence": 1,
"worker_incarnation": "worker_0001"
});
let descriptor = serde_json::json!({
"id": "descriptor_0001",
"kind": "resolved_source",
"schema_version": 1,
"byte_length": 64,
"item_count": 1,
"checksum": format!("sha256:{}", "22".repeat(32))
});
let execution = serde_json::json!({
"id": "grant_0001",
"job_id": "job_00000001",
"attempt_id": "attempt_0001",
"fence": 1,
"worker_incarnation": "worker_0001",
"source": {
"kind": "http",
"display_url": "https://example.invalid/artifact",
"result_path": "artifact.bin"
},
"sensitive_source_ref": "source_00001",
"credential_version": null,
"resolved_source_ref": null,
"target": {
"store": "primary",
"bucket": "remora-results",
"prefix": "users/alice"
},
"staging_target": {
"store": "primary",
"bucket": "remora-results",
"prefix": "_remora/a/attempt_0001/"
},
"execution_grant_ref": "grant_0001",
"checkpoint_ref": null
});
let authority = serde_json::json!({
"attempt_id": "attempt_0001",
"fence": 1,
"worker_incarnation": "worker_0001"
});
let lease = serde_json::json!({
"expires_at": "2026-07-17T00:01:00Z",
"renew_after": "2026-07-17T00:00:30Z"
});
let claim_request = serde_json::json!({
"operation_id": "operation_0001",
"worker_incarnation": "worker_0001"
});
assert_dto_and_schema::<ClaimRequestDto>(&component_schema, "ClaimRequest", claim_request);
let claim_response = serde_json::json!({
"authority": authority,
"execution": execution
});
assert_dto_and_schema::<ClaimResponseDto>(
&component_schema,
"ClaimResponse",
claim_response.clone(),
);
assert_dto_and_schema::<AttemptRenewRequestDto>(
&component_schema,
"AttemptRenewRequest",
mutation.clone(),
);
assert_dto_and_schema::<AttemptActivateRequestDto>(
&component_schema,
"AttemptActivateRequest",
mutation.clone(),
);
assert_dto_and_schema::<ActivationResponseDto>(
&component_schema,
"ActivationResponse",
serde_json::json!({
"authority": claim_response["authority"],
"execution": claim_response["execution"],
"lease": lease.clone()
}),
);
assert_dto_and_schema::<LeaseProjectionDto>(
&component_schema,
"LeaseProjection",
lease.clone(),
);
assert_dto_and_schema::<AttemptResolvedSourceRequestDto>(
&component_schema,
"AttemptResolvedSourceRequest",
merged(
&mutation,
serde_json::json!({ "resolved_source_ref": descriptor.clone() }),
),
);
let mut checkpoint_descriptor = descriptor.clone();
checkpoint_descriptor["kind"] = Value::String("checkpoint".into());
assert_dto_and_schema::<AttemptCheckpointRequestDto>(
&component_schema,
"AttemptCheckpointRequest",
merged(
&mutation,
serde_json::json!({
"checkpoint_ref": checkpoint_descriptor,
"resolved_source_id": "descriptor_0001",
"validated_bytes": 1024,
"validated_members": 1
}),
),
);
assert_dto_and_schema::<AttemptProgressRequestDto>(
&component_schema,
"AttemptProgressRequest",
merged(
&mutation,
serde_json::json!({
"progress": {
"completed_bytes": 1024,
"total_bytes": 4096,
"completed_members": 1,
"total_members": 1
}
}),
),
);
assert_dto_and_schema::<AttemptOutcomeRequestDto>(
&component_schema,
"AttemptOutcomeRequest",
merged(
&mutation,
serde_json::json!({
"outcome": {
"kind": "failure",
"failure": {
"class": "transient_source",
"message": "temporarily unavailable"
}
}
}),
),
);
assert_dto_and_schema::<JobPhaseProjectionDto>(
&component_schema,
"JobPhaseProjection",
serde_json::json!({ "phase": "RetryWait" }),
);
let routes = [
(
"/internal/v1/jobs/{job_id}/claim",
"ClaimRequest",
"200",
Some("ClaimResponse"),
),
(
"/internal/v1/attempts/{attempt_id}/renew",
"AttemptRenewRequest",
"200",
Some("LeaseProjection"),
),
(
"/internal/v1/attempts/{attempt_id}/activate",
"AttemptActivateRequest",
"200",
Some("ActivationResponse"),
),
(
"/internal/v1/attempts/{attempt_id}/resolved-source",
"AttemptResolvedSourceRequest",
"204",
None,
),
(
"/internal/v1/attempts/{attempt_id}/checkpoint",
"AttemptCheckpointRequest",
"204",
None,
),
(
"/internal/v1/attempts/{attempt_id}/progress",
"AttemptProgressRequest",
"204",
None,
),
(
"/internal/v1/attempts/{attempt_id}/outcome",
"AttemptOutcomeRequest",
"200",
Some("JobPhaseProjection"),
),
];
for (path, request, status, response) in routes {
assert_worker_route(&openapi, path, request, status, response);
}
}
fn assert_dto_and_schema<T>(component_schema: &Value, definition: &str, value: Value)
where
T: DeserializeOwned + Serialize,
{
let dto: T = serde_json::from_value(value.clone()).unwrap();
assert_eq!(serde_json::to_value(dto).unwrap(), value);
validate_openapi_value(component_schema, definition, &value);
}
fn validate_openapi_value(component_schema: &Value, definition: &str, value: &Value) {
let mut schema = component_schema.clone();
schema.as_object_mut().unwrap().insert(
"$ref".to_owned(),
Value::String(format!("#/$defs/{definition}")),
);
jsonschema::draft202012::options()
.should_validate_formats(true)
.build(&schema)
.unwrap()
.validate(value)
.unwrap();
}
fn merged(base: &Value, extension: Value) -> Value {
let mut value = base.clone();
value
.as_object_mut()
.unwrap()
.extend(extension.as_object().unwrap().clone());
value
}
fn assert_worker_route(
openapi: &Value,
path: &str,
request_schema: &str,
success_status: &str,
response_schema: Option<&str>,
) {
let operation = &openapi["paths"][path]["post"];
assert_eq!(
operation.pointer("/security/0/serviceAuth"),
Some(&serde_json::json!([]))
);
assert_eq!(
operation.pointer("/requestBody/content/application~1json/schema/$ref"),
Some(&Value::String(format!(
"#/components/schemas/{request_schema}"
)))
);
assert!(operation["responses"].get("409").is_some());
assert!(operation["responses"].get("default").is_some());
let success = &operation["responses"][success_status];
match response_schema {
Some(response_schema) => assert_eq!(
success.pointer("/content/application~1json/schema/$ref"),
Some(&Value::String(format!(
"#/components/schemas/{response_schema}"
)))
),
None => assert!(success.get("content").is_none()),
}
}
fn openapi_component_schema(openapi: &Value) -> Value { fn openapi_component_schema(openapi: &Value) -> Value {
let mut definitions = openapi let mut definitions = openapi
.pointer("/components/schemas") .pointer("/components/schemas")

View file

@ -42,6 +42,7 @@ const PATH_SET: &AsciiSet = &CONTROLS
.add(b'}'); .add(b'}');
/// One version or delete marker listed under a key/prefix. /// One version or delete marker listed under a key/prefix.
#[cfg(feature = "test-utils")]
#[derive(Clone, Debug, Eq, PartialEq)] #[derive(Clone, Debug, Eq, PartialEq)]
pub struct ObjectVersion { pub struct ObjectVersion {
pub key: String, pub key: String,
@ -81,6 +82,7 @@ impl MultipartClient {
/// Enable bucket versioning when the peer supports it. /// Enable bucket versioning when the peer supports it.
/// ///
/// Peers that return NotImplemented are reported as `false` without error. /// Peers that return NotImplemented are reported as `false` without error.
#[cfg(feature = "test-utils")]
pub async fn enable_bucket_versioning(&self) -> Result<bool, TargetError> { pub async fn enable_bucket_versioning(&self) -> Result<bool, TargetError> {
let body = r#"<?xml version="1.0" encoding="UTF-8"?> let body = r#"<?xml version="1.0" encoding="UTF-8"?>
<VersioningConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/"> <VersioningConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/">
@ -341,6 +343,7 @@ impl MultipartClient {
} }
/// List object versions and delete markers under an optional key prefix. /// List object versions and delete markers under an optional key prefix.
#[cfg(feature = "test-utils")]
pub async fn list_object_versions( pub async fn list_object_versions(
&self, &self,
prefix: &str, prefix: &str,
@ -410,6 +413,7 @@ impl MultipartClient {
} }
/// Delete one exact object version or delete marker. /// Delete one exact object version or delete marker.
#[cfg(feature = "test-utils")]
pub async fn delete_object_version( pub async fn delete_object_version(
&self, &self,
key: &str, key: &str,
@ -556,6 +560,7 @@ struct CompletePart {
etag: String, etag: String,
} }
#[cfg(feature = "test-utils")]
#[derive(Debug, Deserialize)] #[derive(Debug, Deserialize)]
struct ListVersionsResult { struct ListVersionsResult {
#[serde(rename = "Version", default)] #[serde(rename = "Version", default)]
@ -570,6 +575,7 @@ struct ListVersionsResult {
next_version_id_marker: Option<String>, next_version_id_marker: Option<String>,
} }
#[cfg(feature = "test-utils")]
#[derive(Debug, Deserialize)] #[derive(Debug, Deserialize)]
struct ListedVersion { struct ListedVersion {
#[serde(rename = "Key", default)] #[serde(rename = "Key", default)]

View file

@ -551,7 +551,7 @@ paths:
content: content:
application/json: application/json:
schema: schema:
$ref: '#/components/schemas/Job' $ref: '#/components/schemas/JobPhaseProjection'
'409': '409':
$ref: '#/components/responses/PreconditionFailed' $ref: '#/components/responses/PreconditionFailed'
default: default:
@ -1109,6 +1109,13 @@ components:
JobPhase: JobPhase:
type: string type: string
enum: [Queued, Running, RetryWait, Cancelling, Succeeded, Failed, Cancelled] enum: [Queued, Running, RetryWait, Cancelling, Succeeded, Failed, Cancelled]
JobPhaseProjection:
type: object
additionalProperties: false
required: [phase]
properties:
phase:
$ref: '#/components/schemas/JobPhase'
TerminalJobPhase: TerminalJobPhase:
type: string type: string
enum: [Succeeded, Failed, Cancelled] enum: [Succeeded, Failed, Cancelled]
@ -1247,10 +1254,15 @@ components:
ResultDescriptor: ResultDescriptor:
type: object type: object
additionalProperties: false additionalProperties: false
required: [kind, manifest, member_count, total_bytes] required: [kind, manifest_id, schema_version, manifest, member_count, total_bytes]
properties: properties:
kind: kind:
const: manifest const: manifest
manifest_id:
$ref: '#/components/schemas/OpaqueId'
schema_version:
type: integer
minimum: 1
manifest: manifest:
$ref: '#/components/schemas/ObjectIdentity' $ref: '#/components/schemas/ObjectIdentity'
member_count: member_count:
@ -1262,7 +1274,7 @@ components:
ObjectIdentity: ObjectIdentity:
type: object type: object
additionalProperties: false additionalProperties: false
required: [store, bucket, key, checksum] required: [store, bucket, key, checksum, size]
properties: properties:
store: store:
type: string type: string
@ -1274,6 +1286,9 @@ components:
type: string type: string
checksum: checksum:
$ref: '#/components/schemas/Sha256Digest' $ref: '#/components/schemas/Sha256Digest'
size:
type: integer
minimum: 0
Sha256Digest: Sha256Digest:
type: string type: string
pattern: '^sha256:[a-f0-9]{64}$' pattern: '^sha256:[a-f0-9]{64}$'
@ -1657,7 +1672,7 @@ components:
allOf: allOf:
- $ref: '#/components/schemas/AttemptMutation' - $ref: '#/components/schemas/AttemptMutation'
- type: object - type: object
required: [checkpoint_ref] required: [checkpoint_ref, resolved_source_id, validated_bytes, validated_members]
properties: properties:
checkpoint_ref: checkpoint_ref:
allOf: allOf:
@ -1666,6 +1681,14 @@ components:
properties: properties:
kind: kind:
const: checkpoint const: checkpoint
resolved_source_id:
$ref: '#/components/schemas/OpaqueId'
validated_bytes:
type: integer
minimum: 0
validated_members:
type: integer
minimum: 0
AttemptProgressRequest: AttemptProgressRequest:
unevaluatedProperties: false unevaluatedProperties: false
allOf: allOf:

View file

@ -0,0 +1,50 @@
#!/usr/bin/env bash
set -euo pipefail
ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
source "$ROOT/tests/kind/lib.sh"
USER_KUBECONFIG_BEFORE="$(user_kubeconfig_digest)"
SUCCEEDED=false
OWNS_FIXTURE=false
cleanup() {
local status=$?
trap - EXIT INT TERM
if [[ "${REMORA_TESTBED_RETAIN:-0}" = 1 ]] && [[ "$SUCCEEDED" != true ]]; then
printf 'retained kubeconfig: %s\n' "$KUBECONFIG_PATH"
printf 'retained context: %s\n' "$CONTEXT"
exit "$status"
fi
if [[ "$OWNS_FIXTURE" = true ]]; then
destroy_fixture || status=1
fi
if [[ "$(user_kubeconfig_digest)" != "$USER_KUBECONFIG_BEFORE" ]]; then
printf 'user kubeconfig changed unexpectedly\n' >&2
status=1
fi
exit "$status"
}
trap cleanup EXIT
trap 'exit 130' INT
trap 'exit 143' TERM
acquire_fixture_lock
REMORA_REAL_NERDCTL="$(command -v nerdctl)"
export REMORA_REAL_NERDCTL
"$REMORA_REAL_NERDCTL" --namespace default build \
--file "$ROOT/Dockerfile.test" \
--target kind-worker-api-runner \
--tag remora-kind-worker-api:local \
"$ROOT"
OWNS_FIXTURE=true
setup_fixture durable coordination
load_local_image remora-kind-worker-api:local
run_job_matrix \
worker-api \
"$TESTBED/worker-api-job.yaml.in" \
remora-kind-worker-api:local \
durable coordination
SUCCEEDED=true
printf 'worker API package test Jobs passed\n'

View file

@ -0,0 +1,34 @@
apiVersion: batch/v1
kind: Job
metadata:
name: @@JOB_NAME@@
namespace: remora-testbed
spec:
activeDeadlineSeconds: 600
backoffLimit: 0
ttlSecondsAfterFinished: 300
template:
spec:
automountServiceAccountToken: false
enableServiceLinks: false
restartPolicy: Never
containers:
- name: tests
image: @@TEST_IMAGE@@
imagePullPolicy: Never
envFrom:
- configMapRef: {name: @@PROFILE_NAME@@}
env:
- name: REMORA_WORKER_API_REQUIRE
value: "1"
volumeMounts:
- {name: fixture-profile, mountPath: /run/remora-fixture, readOnly: true}
resources:
requests: {cpu: 250m, memory: 256Mi}
limits: {cpu: "2", memory: 1Gi}
volumes:
- name: fixture-profile
configMap:
name: @@PROFILE_NAME@@-file
items:
- {key: profile.env, path: profile.env}