Compare commits

..

29 Commits

Author SHA1 Message Date
Yuchen Liang
e565c4fbe9 Merge branch 'yuchen/direct-io-aligned-alloc' into yuchen/direct-io-aligned-alloc-usage-wip 2024-10-07 13:08:55 -04:00
Yuchen Liang
a46757b769 add with_capacity_aligned_zeroed and leak
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-07 13:06:00 -04:00
Yuchen Liang
9c32bfee3b fix put_io_mode to use the correct http endpoint
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 16:47:36 -04:00
Yuchen Liang
69ef8caf58 simplify virtual file wrapper
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 12:31:55 +00:00
Yuchen Liang
b7443dd643 add set_io_mode option to getpage_latest_lsn
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 12:16:42 +00:00
Yuchen Liang
cc433c76a3 fix clippy
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 12:05:01 +00:00
Yuchen Liang
2034ec906a remove unused imports
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 12:04:11 +00:00
Yuchen Liang
f48ab8bcaa use O_DIRECT as preferred
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 03:57:58 +00:00
Yuchen Liang
2607a57990 pageserver: add direct io config to virtual file
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-10-01 03:57:13 +00:00
Yuchen Liang
f04c1c230c incr len in with_capacity_aligned_zeroed add a test
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 23:47:00 +00:00
Yuchen Liang
13f1931a09 fix build
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:18:29 +00:00
Yuchen Liang
e98a4eb5e2 add safety comments
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:00:05 +00:00
Yuchen Liang
e01d145066 remove example; add with_capacity_aligned_zeroed
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:00:05 +00:00
Yuchen Liang
9e9d76d6f2 use IoBufferMut for pagecache
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:00:05 +00:00
Yuchen Liang
14ec379d2b enable O_DIRECT for delta and image layers
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:00:05 +00:00
Yuchen Liang
ebfe88a463 use IoBufferMut for delta and image layers
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-29 21:00:05 +00:00
Yuchen Liang
eb16aa9e81 Merge branch 'main' into yuchen/direct-io-aligned-alloc 2024-09-29 16:59:27 -04:00
Yuchen Liang
f6d0ed6454 implement reserve for IoBufferMut
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-26 14:12:49 +00:00
Yuchen Liang
a2be8a440b Merge branch 'main' into yuchen/direct-io-aligned-alloc 2024-09-24 21:29:33 -04:00
Yuchen Liang
ff4a1db223 make sure we can Send IoBufferMut
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-03 23:52:34 -04:00
Yuchen Liang
29d54ccd20 Merge branch 'main' into yuchen/direct-io-aligned-alloc 2024-09-03 11:41:49 -04:00
Yuchen Liang
68a1fe20f2 review: use doc comments to reference struct in safety comment
Co-authored-by: Christian Schwarz <christian@neon.tech>
2024-09-03 11:41:02 -04:00
Yuchen Liang
e8408c797a remove unused comments
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-03 11:38:58 -04:00
Yuchen Liang
027f28deb9 remove Vec dependency
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-09-03 11:38:00 -04:00
Yuchen Liang
ea6f9798c6 add safety comment
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-08-16 18:06:16 +00:00
Yuchen Liang
253e4d5843 Merge branch 'main' into yuchen/direct-io-aligned-alloc 2024-08-16 13:14:20 -04:00
Yuchen Liang
852099bc83 remove aligned-vec, use ManuallyDrop<Vec<u8>>
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-08-16 17:13:30 +00:00
Yuchen Liang
148e230d11 add mut version marker trait
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-08-15 04:03:13 +00:00
Yuchen Liang
6d664788c1 feat(pageserver): newtype aligned-vec as aligned buffer allocation
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-08-15 03:43:58 +00:00
219 changed files with 2855 additions and 4815 deletions

View File

@@ -557,7 +557,7 @@ jobs:
ln -s /home/nonroot/pg/usr/lib/postgresql/16/bin/psql /tmp/neon/pg_install/v16/bin/psql
ln -s /home/nonroot/pg/usr/lib/$(uname -m)-linux-gnu /tmp/neon/pg_install/v16/lib
LD_LIBRARY_PATH="/home/nonroot/pg/usr/lib/$(uname -m)-linux-gnu:${LD_LIBRARY_PATH:-}"
LD_LIBRARY_PATH="/home/nonroot/pg/usr/lib/$(uname -m)-linux-gnu:${LD_LIBRARY_PATH}"
export LD_LIBRARY_PATH
echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}" >> ${GITHUB_ENV}

View File

@@ -341,7 +341,7 @@ jobs:
PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
TEST_RESULT_CONNSTR: "${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}"
PAGESERVER_VIRTUAL_FILE_IO_ENGINE: tokio-epoll-uring
SYNC_BETWEEN_TESTS: true
SYNC_AFTER_EACH_TEST: true
# XXX: no coverage data handling here, since benchmarks are run on release builds,
# while coverage is currently collected for the debug ones
@@ -773,7 +773,7 @@ jobs:
matrix:
version: [ v14, v15, v16, v17 ]
env:
VM_BUILDER_VERSION: v0.35.0
VM_BUILDER_VERSION: v0.29.3
steps:
- uses: actions/checkout@v4

View File

@@ -102,17 +102,12 @@ jobs:
# Default set of platforms to run e2e tests on
platforms='["docker", "k8s"]'
# If a PR changes anything that affects computes, add k8s-neonvm to the list of platforms.
# If the PR changes vendor/, pgxn/ or libs/vm_monitor/ directories, or compute/Dockerfile.compute-node, add k8s-neonvm to the list of platforms.
# If the workflow run is not a pull request, add k8s-neonvm to the list.
if [ "$GITHUB_EVENT_NAME" == "pull_request" ]; then
for f in $(gh api "/repos/${GITHUB_REPOSITORY}/pulls/${PR_NUMBER}/files" --paginate --jq '.[].filename'); do
case "$f" in
# List of directories that contain code which affect compute images.
#
# This isn't exhaustive, just the paths that are most directly compute-related.
# For example, compute_ctl also depends on libs/utils, but we don't trigger
# an e2e run on that.
vendor/*|pgxn/*|compute_tools/*|libs/vm_monitor/*|compute/Dockerfile.compute-node)
vendor/*|pgxn/*|libs/vm_monitor/*|compute/Dockerfile.compute-node)
platforms=$(echo "${platforms}" | jq --compact-output '. += ["k8s-neonvm"] | unique')
;;
*)

427
Cargo.lock generated
View File

@@ -90,9 +90,9 @@ dependencies = [
[[package]]
name = "anstyle"
version = "1.0.8"
version = "1.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1bec1de6f59aedf83baf9ff929c98f2ad654b97c9510f4e70cf6f661d49fd5b1"
checksum = "41ed9a86bf92ae6580e0a31281f65a1b1d867c0cc68d5346e2ae128dddfa6a7d"
[[package]]
name = "anstyle-parse"
@@ -269,9 +269,9 @@ checksum = "d468802bab17cbc0cc575e9b053f41e72aa36bfa6b7f55e3529ffa43161b97fa"
[[package]]
name = "aws-config"
version = "1.5.1"
version = "1.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2ac9889352d632214df943e26740c46a0f3da6e329fbd28164fe7ae1b061da7b"
checksum = "baaa0be6ee7d90b775ae6ccb6d2ba182b91219ec2001f92338773a094246af1d"
dependencies = [
"aws-credential-types",
"aws-runtime",
@@ -300,9 +300,9 @@ dependencies = [
[[package]]
name = "aws-credential-types"
version = "1.2.1"
version = "1.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "60e8f6b615cb5fc60a98132268508ad104310f0cfb25a1c22eee76efdf9154da"
checksum = "e16838e6c9e12125face1c1eff1343c75e3ff540de98ff7ebd61874a89bcfeb9"
dependencies = [
"aws-smithy-async",
"aws-smithy-runtime-api",
@@ -312,16 +312,15 @@ dependencies = [
[[package]]
name = "aws-runtime"
version = "1.4.3"
version = "1.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a10d5c055aa540164d9561a0e2e74ad30f0dcf7393c3a92f6733ddf9c5762468"
checksum = "785da4a15e7b166b505fd577e4560c7a7cd8fbdf842eb1336cbcbf8944ce56f1"
dependencies = [
"aws-credential-types",
"aws-sigv4",
"aws-smithy-async",
"aws-smithy-eventstream",
"aws-smithy-http",
"aws-smithy-runtime",
"aws-smithy-runtime-api",
"aws-smithy-types",
"aws-types",
@@ -329,7 +328,6 @@ dependencies = [
"fastrand 2.0.0",
"http 0.2.9",
"http-body 0.4.5",
"once_cell",
"percent-encoding",
"pin-project-lite",
"tracing",
@@ -338,9 +336,9 @@ dependencies = [
[[package]]
name = "aws-sdk-iam"
version = "1.46.0"
version = "1.17.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "053df3024ea2ed0431359b3cddecc92dcfadeaedf71dd497292b39e37e597b46"
checksum = "b8ae76026bfb1b80a6aed0bb400c1139cd9c0563e26bce1986cd021c6a968c7b"
dependencies = [
"aws-credential-types",
"aws-runtime",
@@ -361,9 +359,9 @@ dependencies = [
[[package]]
name = "aws-sdk-s3"
version = "1.52.0"
version = "1.26.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f571deb0a80c20d21d9f3e8418c1712af9ff4bf399d057e5549a934eca4844e2"
checksum = "7bc5ce518d4b8d16e0408de7bdf1b3097cec61a7daa979750a208f8d9934386d"
dependencies = [
"ahash",
"aws-credential-types",
@@ -396,9 +394,9 @@ dependencies = [
[[package]]
name = "aws-sdk-sso"
version = "1.30.0"
version = "1.22.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ebb97e44983752cf7e12968c5f569a5d7562dbbc67006755c331d9d9c99580ae"
checksum = "ca3d6c4cba4e009391b72b0fcf12aff04ea3c9c3aa2ecaafa330326a8bd7e601"
dependencies = [
"aws-credential-types",
"aws-runtime",
@@ -418,9 +416,9 @@ dependencies = [
[[package]]
name = "aws-sdk-ssooidc"
version = "1.31.0"
version = "1.22.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ad061d977235898e4a97ecbd5d882786cca41b4828943584dc792dcc35eb3d3c"
checksum = "73400dc239d14f63d932f4ca7b55af5e9ef1f857f7d70655249ccc287adb2570"
dependencies = [
"aws-credential-types",
"aws-runtime",
@@ -440,9 +438,9 @@ dependencies = [
[[package]]
name = "aws-sdk-sts"
version = "1.30.0"
version = "1.22.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "300ce43d1f7f4eb023e57d38b0921d964e8e62bed7f82f6b7849e7eab7a14575"
checksum = "10f8858308af76fba3e5ffcf1bb56af5471574d2bdfaf0159470c25bc2f760e5"
dependencies = [
"aws-credential-types",
"aws-runtime",
@@ -463,9 +461,9 @@ dependencies = [
[[package]]
name = "aws-sigv4"
version = "1.2.4"
version = "1.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cc8db6904450bafe7473c6ca9123f88cc11089e41a025408f992db4e22d3be68"
checksum = "58b56f1cbe6fd4d0c2573df72868f20ab1c125ca9c9dbce17927a463433a2e57"
dependencies = [
"aws-credential-types",
"aws-smithy-eventstream",
@@ -503,9 +501,9 @@ dependencies = [
[[package]]
name = "aws-smithy-checksums"
version = "0.60.12"
version = "0.60.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "598b1689d001c4d4dc3cb386adb07d37786783aee3ac4b324bcadac116bf3d23"
checksum = "83fa43bc04a6b2441968faeab56e68da3812f978a670a5db32accbdcafddd12f"
dependencies = [
"aws-smithy-http",
"aws-smithy-types",
@@ -524,9 +522,9 @@ dependencies = [
[[package]]
name = "aws-smithy-eventstream"
version = "0.60.5"
version = "0.60.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cef7d0a272725f87e51ba2bf89f8c21e4df61b9e49ae1ac367a6d69916ef7c90"
checksum = "e6363078f927f612b970edf9d1903ef5cef9a64d1e8423525ebb1f0a1633c858"
dependencies = [
"aws-smithy-types",
"bytes",
@@ -535,9 +533,9 @@ dependencies = [
[[package]]
name = "aws-smithy-http"
version = "0.60.11"
version = "0.60.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5c8bc3e8fdc6b8d07d976e301c02fe553f72a39b7a9fea820e023268467d7ab6"
checksum = "4a7de001a1b9a25601016d8057ea16e31a45fdca3751304c8edf4ad72e706c08"
dependencies = [
"aws-smithy-eventstream",
"aws-smithy-runtime-api",
@@ -575,9 +573,9 @@ dependencies = [
[[package]]
name = "aws-smithy-runtime"
version = "1.7.1"
version = "1.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d1ce695746394772e7000b39fe073095db6d45a862d0767dd5ad0ac0d7f8eb87"
checksum = "c9ac79e9f3a4d576f3cd4a470a0275b138d9e7b11b1cd514a6858ae0a79dd5bb"
dependencies = [
"aws-smithy-async",
"aws-smithy-http",
@@ -589,7 +587,6 @@ dependencies = [
"http 0.2.9",
"http-body 0.4.5",
"http-body 1.0.0",
"httparse",
"hyper 0.14.30",
"hyper-rustls 0.24.0",
"once_cell",
@@ -602,9 +599,9 @@ dependencies = [
[[package]]
name = "aws-smithy-runtime-api"
version = "1.7.2"
version = "1.6.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e086682a53d3aa241192aa110fa8dfce98f2f5ac2ead0de84d41582c7e8fdb96"
checksum = "04ec42c2f5c0e7796a2848dde4d9f3bf8ce12ccbb3d5aa40c52fa0cdd61a1c47"
dependencies = [
"aws-smithy-async",
"aws-smithy-types",
@@ -619,9 +616,9 @@ dependencies = [
[[package]]
name = "aws-smithy-types"
version = "1.2.7"
version = "1.1.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "147100a7bea70fa20ef224a6bad700358305f5dc0f84649c53769761395b355b"
checksum = "baf98d97bba6ddaba180f1b1147e202d8fe04940403a95a3f826c790f931bbd1"
dependencies = [
"base64-simd",
"bytes",
@@ -645,23 +642,24 @@ dependencies = [
[[package]]
name = "aws-smithy-xml"
version = "0.60.9"
version = "0.60.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ab0b0166827aa700d3dc519f72f8b3a91c35d0b8d042dc5d643a91e6f80648fc"
checksum = "d123fbc2a4adc3c301652ba8e149bf4bc1d1725affb9784eb20c953ace06bf55"
dependencies = [
"xmlparser",
]
[[package]]
name = "aws-types"
version = "1.3.3"
version = "1.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5221b91b3e441e6675310829fd8984801b772cb1546ef6c0e54dec9f1ac13fef"
checksum = "5a43b56df2c529fe44cb4d92bd64d0479883fb9608ff62daede4df5405381814"
dependencies = [
"aws-credential-types",
"aws-smithy-async",
"aws-smithy-runtime-api",
"aws-smithy-types",
"http 0.2.9",
"rustc_version",
"tracing",
]
@@ -673,7 +671,8 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3b829e4e32b91e643de6eafe82b1d90675f5874230191a4ffbc1b336dec4d6bf"
dependencies = [
"async-trait",
"axum-core 0.3.4",
"axum-core",
"base64 0.21.1",
"bitflags 1.3.2",
"bytes",
"futures-util",
@@ -688,47 +687,16 @@ dependencies = [
"pin-project-lite",
"rustversion",
"serde",
"sync_wrapper 0.1.2",
"tower",
"tower-layer",
"tower-service",
]
[[package]]
name = "axum"
version = "0.7.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3a6c9af12842a67734c9a2e355436e5d03b22383ed60cf13cd0c18fbfe3dcbcf"
dependencies = [
"async-trait",
"axum-core 0.4.5",
"base64 0.21.1",
"bytes",
"futures-util",
"http 1.1.0",
"http-body 1.0.0",
"http-body-util",
"hyper 1.4.1",
"hyper-util",
"itoa",
"matchit 0.7.0",
"memchr",
"mime",
"percent-encoding",
"pin-project-lite",
"rustversion",
"serde",
"serde_json",
"serde_path_to_error",
"serde_urlencoded",
"sha1",
"sync_wrapper 1.0.1",
"sync_wrapper",
"tokio",
"tokio-tungstenite",
"tower",
"tower-layer",
"tower-service",
"tracing",
]
[[package]]
@@ -748,27 +716,6 @@ dependencies = [
"tower-service",
]
[[package]]
name = "axum-core"
version = "0.4.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "09f2bd6146b97ae3359fa0cc6d6b376d9539582c7b4220f041a33ec24c226199"
dependencies = [
"async-trait",
"bytes",
"futures-util",
"http 1.1.0",
"http-body 1.0.0",
"http-body-util",
"mime",
"pin-project-lite",
"rustversion",
"sync_wrapper 1.0.1",
"tower-layer",
"tower-service",
"tracing",
]
[[package]]
name = "azure_core"
version = "0.19.0"
@@ -1265,7 +1212,6 @@ version = "0.1.0"
dependencies = [
"anyhow",
"bytes",
"camino",
"cfg-if",
"chrono",
"clap",
@@ -1277,7 +1223,6 @@ dependencies = [
"notify",
"num_cpus",
"opentelemetry",
"opentelemetry_sdk",
"postgres",
"regex",
"remote_storage",
@@ -1931,9 +1876,9 @@ dependencies = [
[[package]]
name = "env_logger"
version = "0.10.2"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4cd405aab171cb85d6735e5c8d9db038c17d3ca007a4d2c25f337935c3d90580"
checksum = "85cdab6a89accf66733ad5a1693a4dcced6aeff64602b634530dd73c1f3ee9f0"
dependencies = [
"humantime",
"is-terminal",
@@ -2091,7 +2036,7 @@ dependencies = [
"futures-core",
"futures-sink",
"http-body-util",
"hyper 1.4.1",
"hyper 1.2.0",
"hyper-util",
"pin-project",
"rand 0.8.5",
@@ -2454,15 +2399,6 @@ dependencies = [
"digest",
]
[[package]]
name = "home"
version = "0.5.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e3d1354bf6b7235cb4a0576c2619fd4ed18183f689b12b006a0ee7329eeff9a5"
dependencies = [
"windows-sys 0.52.0",
]
[[package]]
name = "hostname"
version = "0.4.0"
@@ -2519,9 +2455,9 @@ dependencies = [
[[package]]
name = "http-body-util"
version = "0.1.2"
version = "0.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "793429d76616a256bcb62c2a2ec2bed781c8307e797e2598c50010f2bee2544f"
checksum = "41cb79eb393015dadd30fc252023adb0b2400a0caee0fa2a077e6e21a551e840"
dependencies = [
"bytes",
"futures-util",
@@ -2604,9 +2540,9 @@ dependencies = [
[[package]]
name = "hyper"
version = "1.4.1"
version = "1.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "50dfd22e0e76d0f662d429a5f80fcaf3855009297eab6a0a9f8543834744ba05"
checksum = "186548d73ac615b32a73aafe38fb4f56c0d340e110e5a200bcadbaf2e199263a"
dependencies = [
"bytes",
"futures-channel",
@@ -2646,7 +2582,7 @@ checksum = "a0bea761b46ae2b24eb4aef630d8d1c398157b6fc29e6350ecf090a0b70c952c"
dependencies = [
"futures-util",
"http 1.1.0",
"hyper 1.4.1",
"hyper 1.2.0",
"hyper-util",
"rustls 0.22.4",
"rustls-pki-types",
@@ -2669,16 +2605,16 @@ dependencies = [
[[package]]
name = "hyper-util"
version = "0.1.7"
version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cde7055719c54e36e95e8719f95883f22072a48ede39db7fc17a4e1d5281e9b9"
checksum = "ca38ef113da30126bbff9cd1705f9273e15d45498615d138b0c20279ac7a76aa"
dependencies = [
"bytes",
"futures-channel",
"futures-util",
"http 1.1.0",
"http-body 1.0.0",
"hyper 1.4.1",
"hyper 1.2.0",
"pin-project-lite",
"socket2",
"tokio",
@@ -3432,82 +3368,102 @@ checksum = "ff011a302c396a5197692431fc1948019154afc178baf7d8e37367442a4601cf"
[[package]]
name = "opentelemetry"
version = "0.24.0"
version = "0.20.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4c365a63eec4f55b7efeceb724f1336f26a9cf3427b70e59e2cd2a5b947fba96"
checksum = "9591d937bc0e6d2feb6f71a559540ab300ea49955229c347a517a28d27784c54"
dependencies = [
"futures-core",
"futures-sink",
"js-sys",
"once_cell",
"pin-project-lite",
"thiserror",
"opentelemetry_api",
"opentelemetry_sdk",
]
[[package]]
name = "opentelemetry-http"
version = "0.13.0"
version = "0.9.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ad31e9de44ee3538fb9d64fe3376c1362f406162434609e79aea2a41a0af78ab"
checksum = "c7594ec0e11d8e33faf03530a4c49af7064ebba81c1480e01be67d90b356508b"
dependencies = [
"async-trait",
"bytes",
"http 1.1.0",
"opentelemetry",
"reqwest 0.12.4",
"http 0.2.9",
"opentelemetry_api",
"reqwest 0.11.19",
]
[[package]]
name = "opentelemetry-otlp"
version = "0.17.0"
version = "0.13.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6b925a602ffb916fb7421276b86756027b37ee708f9dce2dbdcc51739f07e727"
checksum = "7e5e5a5c4135864099f3faafbe939eb4d7f9b80ebf68a8448da961b32a7c1275"
dependencies = [
"async-trait",
"futures-core",
"http 1.1.0",
"opentelemetry",
"http 0.2.9",
"opentelemetry-http",
"opentelemetry-proto",
"opentelemetry-semantic-conventions",
"opentelemetry_api",
"opentelemetry_sdk",
"prost 0.13.3",
"reqwest 0.12.4",
"prost",
"reqwest 0.11.19",
"thiserror",
"tokio",
"tonic",
]
[[package]]
name = "opentelemetry-proto"
version = "0.7.0"
version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "30ee9f20bff9c984511a02f082dc8ede839e4a9bf15cc2487c8d6fea5ad850d9"
checksum = "b1e3f814aa9f8c905d0ee4bde026afd3b2577a97c10e1699912e3e44f0c4cbeb"
dependencies = [
"opentelemetry",
"opentelemetry_api",
"opentelemetry_sdk",
"prost 0.13.3",
"tonic 0.12.3",
"prost",
"tonic",
]
[[package]]
name = "opentelemetry-semantic-conventions"
version = "0.16.0"
version = "0.12.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1cefe0543875379e47eb5f1e68ff83f45cc41366a92dfd0d073d513bf68e9a05"
checksum = "73c9f9340ad135068800e7f1b24e9e09ed9e7143f5bf8518ded3d3ec69789269"
dependencies = [
"opentelemetry",
]
[[package]]
name = "opentelemetry_api"
version = "0.20.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8a81f725323db1b1206ca3da8bb19874bbd3f57c3bcd59471bfb04525b265b9b"
dependencies = [
"futures-channel",
"futures-util",
"indexmap 1.9.3",
"js-sys",
"once_cell",
"pin-project-lite",
"thiserror",
"urlencoding",
]
[[package]]
name = "opentelemetry_sdk"
version = "0.24.1"
version = "0.20.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "692eac490ec80f24a17828d49b40b60f5aeaccdfe6a503f939713afd22bc28df"
checksum = "fa8e705a0612d48139799fcbaba0d4a90f06277153e43dd2bdc16c6f0edd8026"
dependencies = [
"async-trait",
"crossbeam-channel",
"futures-channel",
"futures-executor",
"futures-util",
"glob",
"once_cell",
"opentelemetry",
"opentelemetry_api",
"ordered-float 3.9.2",
"percent-encoding",
"rand 0.8.5",
"regex",
"serde_json",
"thiserror",
"tokio",
@@ -3523,6 +3479,15 @@ dependencies = [
"num-traits",
]
[[package]]
name = "ordered-float"
version = "3.9.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f1e1c390732d15f1d48471625cd92d154e66db2c56645e29a9cd26f4699f72dc"
dependencies = [
"num-traits",
]
[[package]]
name = "ordered-multimap"
version = "0.7.3"
@@ -4265,17 +4230,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0b82eaa1d779e9a4bc1c3217db8ffbeabaae1dca241bf70183242128d48681cd"
dependencies = [
"bytes",
"prost-derive 0.11.9",
]
[[package]]
name = "prost"
version = "0.13.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7b0487d90e047de87f984913713b85c601c05609aad5b0df4b4573fbf69aa13f"
dependencies = [
"bytes",
"prost-derive 0.13.3",
"prost-derive",
]
[[package]]
@@ -4292,7 +4247,7 @@ dependencies = [
"multimap",
"petgraph",
"prettyplease 0.1.25",
"prost 0.11.9",
"prost",
"prost-types",
"regex",
"syn 1.0.109",
@@ -4313,26 +4268,13 @@ dependencies = [
"syn 1.0.109",
]
[[package]]
name = "prost-derive"
version = "0.13.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e9552f850d5f0964a4e4d0bf306459ac29323ddfbae05e35a7c0d35cb0803cc5"
dependencies = [
"anyhow",
"itertools 0.12.1",
"proc-macro2",
"quote",
"syn 2.0.52",
]
[[package]]
name = "prost-types"
version = "0.11.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "213622a1460818959ac1181aaeb2dc9c7f63df720db7d788b3e24eacd1983e13"
dependencies = [
"prost 0.11.9",
"prost",
]
[[package]]
@@ -4373,7 +4315,7 @@ dependencies = [
"humantime",
"humantime-serde",
"hyper 0.14.30",
"hyper 1.4.1",
"hyper 1.2.0",
"hyper-util",
"indexmap 2.0.1",
"ipnet",
@@ -4429,6 +4371,7 @@ dependencies = [
"tokio-tungstenite",
"tokio-util",
"tracing",
"tracing-opentelemetry",
"tracing-subscriber",
"tracing-utils",
"try-lock",
@@ -4739,7 +4682,7 @@ dependencies = [
"scopeguard",
"serde",
"serde_json",
"sync_wrapper 0.1.2",
"sync_wrapper",
"test-context",
"tokio",
"tokio-stream",
@@ -4804,7 +4747,7 @@ dependencies = [
"http 1.1.0",
"http-body 1.0.0",
"http-body-util",
"hyper 1.4.1",
"hyper 1.2.0",
"hyper-rustls 0.26.0",
"hyper-util",
"ipnet",
@@ -4820,7 +4763,7 @@ dependencies = [
"serde",
"serde_json",
"serde_urlencoded",
"sync_wrapper 0.1.2",
"sync_wrapper",
"tokio",
"tokio-rustls 0.25.0",
"tokio-util",
@@ -4861,7 +4804,7 @@ dependencies = [
"futures",
"getrandom 0.2.11",
"http 1.1.0",
"hyper 1.4.1",
"hyper 1.2.0",
"parking_lot 0.11.2",
"reqwest 0.12.4",
"reqwest-middleware",
@@ -4873,9 +4816,9 @@ dependencies = [
[[package]]
name = "reqwest-tracing"
version = "0.5.3"
version = "0.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bfdd9bfa64c72233d8dd99ab7883efcdefe9e16d46488ecb9228b71a2e2ceb45"
checksum = "b253954a1979e02eabccd7e9c3d61d8f86576108baa160775e7f160bb4e800a3"
dependencies = [
"anyhow",
"async-trait",
@@ -5760,9 +5703,9 @@ dependencies = [
"metrics",
"once_cell",
"parking_lot 0.12.1",
"prost 0.11.9",
"prost",
"tokio",
"tonic 0.9.2",
"tonic",
"tonic-build",
"tracing",
"utils",
@@ -5958,12 +5901,6 @@ dependencies = [
"futures-core",
]
[[package]]
name = "sync_wrapper"
version = "1.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a7065abeca94b6a8a577f9bd45aa0867a2238b74e8eb67cf10d492bc39351394"
[[package]]
name = "synstructure"
version = "0.12.6"
@@ -6092,7 +6029,7 @@ checksum = "7e54bc85fc7faa8bc175c4bab5b92ba8d9a3ce893d0e9f42cc455c8ab16a9e09"
dependencies = [
"byteorder",
"integer-encoding",
"ordered-float",
"ordered-float 2.10.1",
]
[[package]]
@@ -6194,9 +6131,9 @@ checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
[[package]]
name = "tokio"
version = "1.38.1"
version = "1.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eb2caba9f80616f438e09748d5acda951967e1ea58508ef53d9c6402485a46df"
checksum = "1adbebffeca75fcfd058afa480fb6c0b81e165a0323f9c9d39c9697e37c46787"
dependencies = [
"backtrace",
"bytes",
@@ -6238,9 +6175,9 @@ dependencies = [
[[package]]
name = "tokio-macros"
version = "2.3.0"
version = "2.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5f5ae998a069d4b5aba8ee9dad856af7d520c3699e6159b185c2acd48155d39a"
checksum = "5b8a1e28f2deaa14e508979454cb3a223b10b938b45af148bc0986de36f1923b"
dependencies = [
"proc-macro2",
"quote",
@@ -6308,9 +6245,9 @@ dependencies = [
[[package]]
name = "tokio-stream"
version = "0.1.16"
version = "0.1.14"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4f4e6ce100d0eb49a2734f8c0812bcd324cf357d21810932c5df6b96ef2b86f1"
checksum = "397c988d37662c7dda6d2208364a706264bf3d6138b11d436cbac0ad38832842"
dependencies = [
"futures-core",
"pin-project-lite",
@@ -6334,9 +6271,9 @@ dependencies = [
[[package]]
name = "tokio-tungstenite"
version = "0.21.0"
version = "0.20.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c83b561d025642014097b66e6c1bb422783339e0909e4429cde4749d1990bc38"
checksum = "2b2dbec703c26b00d74844519606ef15d09a7d6857860f84ad223dec002ddea2"
dependencies = [
"futures-util",
"log",
@@ -6403,7 +6340,7 @@ checksum = "3082666a3a6433f7f511c7192923fa1fe07c69332d3c6a2e6bb040b569199d5a"
dependencies = [
"async-stream",
"async-trait",
"axum 0.6.20",
"axum",
"base64 0.21.1",
"bytes",
"futures-core",
@@ -6415,7 +6352,7 @@ dependencies = [
"hyper-timeout",
"percent-encoding",
"pin-project",
"prost 0.11.9",
"prost",
"rustls-native-certs 0.6.2",
"rustls-pemfile 1.0.2",
"tokio",
@@ -6427,27 +6364,6 @@ dependencies = [
"tracing",
]
[[package]]
name = "tonic"
version = "0.12.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "877c5b330756d856ffcc4553ab34a5684481ade925ecc54bcd1bf02b1d0d4d52"
dependencies = [
"async-trait",
"base64 0.22.1",
"bytes",
"http 1.1.0",
"http-body 1.0.0",
"http-body-util",
"percent-encoding",
"pin-project",
"prost 0.13.3",
"tokio-stream",
"tower-layer",
"tower-service",
"tracing",
]
[[package]]
name = "tonic-build"
version = "0.9.2"
@@ -6495,10 +6411,11 @@ checksum = "b6bc1c9ce2b5135ac7f93c72918fc37feb872bdc6a5533a8b85eb4b86bfdae52"
[[package]]
name = "tracing"
version = "0.1.40"
version = "0.1.37"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c3523ab5a71916ccf420eebdf5521fcef02141234bbc0b8a49f2fdc4544364ef"
checksum = "8ce8c33a8d48bd45d624a6e523445fd21ec13d3653cd51f681abf67418f54eb8"
dependencies = [
"cfg-if",
"log",
"pin-project-lite",
"tracing-attributes",
@@ -6518,9 +6435,9 @@ dependencies = [
[[package]]
name = "tracing-attributes"
version = "0.1.27"
version = "0.1.24"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "34704c8d6ebcbc939824180af020566b01a7c01f80641264eba0999f6c2b6be7"
checksum = "0f57e3ca2a01450b1a921183a9c9cbfda207fd822cef4ccb00a65402cbba7a74"
dependencies = [
"proc-macro2",
"quote",
@@ -6529,9 +6446,9 @@ dependencies = [
[[package]]
name = "tracing-core"
version = "0.1.32"
version = "0.1.31"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c06d3da6113f116aaee68e4d601191614c9053067f9ab7f6edbcb161237daa54"
checksum = "0955b8137a1df6f1a2e9a37d8a6656291ff0297c1a97c24e0d8425fe2312f79a"
dependencies = [
"once_cell",
"valuable",
@@ -6549,22 +6466,21 @@ dependencies = [
[[package]]
name = "tracing-log"
version = "0.2.0"
version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ee855f1f400bd0e5c02d150ae5de3840039a3f54b025156404e34c23c03f47c3"
checksum = "78ddad33d2d10b1ed7eb9d1f518a5674713876e97e5bb9b7345a7984fbb4f922"
dependencies = [
"lazy_static",
"log",
"once_cell",
"tracing-core",
]
[[package]]
name = "tracing-opentelemetry"
version = "0.25.0"
version = "0.21.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a9784ed4da7d921bc8df6963f8c80a0e4ce34ba6ba76668acadd3edbd985ff3b"
checksum = "75327c6b667828ddc28f5e3f169036cb793c3f588d83bf0f262a7f062ffed3c8"
dependencies = [
"js-sys",
"once_cell",
"opentelemetry",
"opentelemetry_sdk",
@@ -6573,7 +6489,6 @@ dependencies = [
"tracing-core",
"tracing-log",
"tracing-subscriber",
"web-time",
]
[[package]]
@@ -6588,9 +6503,9 @@ dependencies = [
[[package]]
name = "tracing-subscriber"
version = "0.3.18"
version = "0.3.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ad0f048c97dbd9faa9b7df56362b8ebcaa52adb06b498c050d2f4e32f90a7a8b"
checksum = "30a651bc37f915e81f087d86e62a18eec5f79550c7faff886f7090b4ea757c77"
dependencies = [
"matchers",
"once_cell",
@@ -6614,7 +6529,6 @@ dependencies = [
"opentelemetry",
"opentelemetry-otlp",
"opentelemetry-semantic-conventions",
"opentelemetry_sdk",
"tokio",
"tracing",
"tracing-opentelemetry",
@@ -6629,14 +6543,14 @@ checksum = "e421abadd41a4225275504ea4d6566923418b7f05506fbc9c0fe86ba7396114b"
[[package]]
name = "tungstenite"
version = "0.21.0"
version = "0.20.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9ef1a641ea34f399a848dea702823bbecfb4c486f911735368f1f137cb8257e1"
checksum = "9e3dac10fd62eaf6617d3a904ae222845979aec67c615d1c842b4002c7666fb9"
dependencies = [
"byteorder",
"bytes",
"data-encoding",
"http 1.1.0",
"http 0.2.9",
"httparse",
"log",
"rand 0.8.5",
@@ -6864,7 +6778,7 @@ name = "vm_monitor"
version = "0.1.0"
dependencies = [
"anyhow",
"axum 0.7.5",
"axum",
"cgroups-rs",
"clap",
"futures",
@@ -7070,16 +6984,6 @@ dependencies = [
"wasm-bindgen",
]
[[package]]
name = "web-time"
version = "1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5a6580f308b1fad9207618087a65c04e7a10bc77e02c8e84e9b00dd4b12fa0bb"
dependencies = [
"js-sys",
"wasm-bindgen",
]
[[package]]
name = "webpki-roots"
version = "0.25.2"
@@ -7097,14 +7001,13 @@ dependencies = [
[[package]]
name = "which"
version = "4.4.2"
version = "4.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "87ba24419a2078cd2b0f2ede2691b6c66d8e47836da3b6db8265ebad47afbfc7"
checksum = "2441c784c52b289a054b7201fc93253e288f094e2f4be9058343127c4226a269"
dependencies = [
"either",
"home",
"libc",
"once_cell",
"rustix",
]
[[package]]
@@ -7335,15 +7238,22 @@ version = "0.1.0"
dependencies = [
"ahash",
"anyhow",
"aws-config",
"aws-runtime",
"aws-sigv4",
"aws-smithy-async",
"aws-smithy-http",
"aws-smithy-types",
"axum",
"base64 0.21.1",
"base64ct",
"bitflags 2.4.1",
"bytes",
"camino",
"cc",
"chrono",
"clap",
"clap_builder",
"crossbeam-utils",
"crypto-bigint 0.5.5",
"der 0.7.8",
"deranged",
@@ -7362,8 +7272,6 @@ dependencies = [
"hex",
"hmac",
"hyper 0.14.30",
"hyper 1.4.1",
"hyper-util",
"indexmap 1.9.3",
"itertools 0.10.5",
"itertools 0.12.1",
@@ -7378,12 +7286,13 @@ dependencies = [
"once_cell",
"parquet",
"proc-macro2",
"prost 0.11.9",
"prost",
"quote",
"rand 0.8.5",
"regex",
"regex-automata 0.4.3",
"regex-syntax 0.8.2",
"reqwest 0.11.19",
"reqwest 0.12.4",
"rustls 0.21.11",
"scopeguard",
@@ -7396,7 +7305,7 @@ dependencies = [
"subtle",
"syn 1.0.109",
"syn 2.0.52",
"sync_wrapper 0.1.2",
"sync_wrapper",
"tikv-jemalloc-sys",
"time",
"time-macros",
@@ -7404,10 +7313,14 @@ dependencies = [
"tokio-rustls 0.24.0",
"tokio-util",
"toml_edit",
"tonic",
"tower",
"tracing",
"tracing-core",
"tracing-log",
"tracing-subscriber",
"url",
"uuid",
"zeroize",
"zstd",
"zstd-safe",

View File

@@ -53,15 +53,15 @@ azure_storage_blobs = { version = "0.19", default-features = false, features = [
flate2 = "1.0.26"
async-stream = "0.3"
async-trait = "0.1"
aws-config = { version = "1.5", default-features = false, features=["rustls", "sso"] }
aws-sdk-s3 = "1.52"
aws-sdk-iam = "1.46.0"
aws-config = { version = "1.3", default-features = false, features=["rustls"] }
aws-sdk-s3 = "1.26"
aws-sdk-iam = "1.15.0"
aws-smithy-async = { version = "1.2.1", default-features = false, features=["rt-tokio"] }
aws-smithy-types = "1.2"
aws-smithy-types = "1.1.9"
aws-credential-types = "1.2.0"
aws-sigv4 = { version = "1.2", features = ["sign-http"] }
aws-types = "1.3"
axum = { version = "0.7.5", features = ["ws"] }
aws-sigv4 = { version = "1.2.1", features = ["sign-http"] }
aws-types = "1.2.0"
axum = { version = "0.6.20", features = ["ws"] }
base64 = "0.13.0"
bincode = "1.3"
bindgen = "0.70"
@@ -96,13 +96,10 @@ hmac = "0.12.1"
hostname = "0.4"
http = {version = "1.1.0", features = ["std"]}
http-types = { version = "2", default-features = false }
http-body-util = "0.1.2"
humantime = "2.1"
humantime-serde = "1.1.1"
hyper0 = { package = "hyper", version = "0.14" }
hyper = "1.4"
hyper-util = "0.1"
tokio-tungstenite = "0.21.0"
hyper = "0.14"
tokio-tungstenite = "0.20.0"
indexmap = "2"
indoc = "2"
ipnet = "2.9.0"
@@ -119,10 +116,9 @@ notify = "6.0.0"
num_cpus = "1.15"
num-traits = "0.2.15"
once_cell = "1.13"
opentelemetry = "0.24"
opentelemetry_sdk = "0.24"
opentelemetry-otlp = { version = "0.17", default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
opentelemetry-semantic-conventions = "0.16"
opentelemetry = "0.20.0"
opentelemetry-otlp = { version = "0.13.0", default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
opentelemetry-semantic-conventions = "0.12.0"
parking_lot = "0.12"
parquet = { version = "53", default-features = false, features = ["zstd"] }
parquet_derive = "53"
@@ -135,7 +131,7 @@ rand = "0.8"
redis = { version = "0.25.2", features = ["tokio-rustls-comp", "keep-alive"] }
regex = "1.10.2"
reqwest = { version = "0.12", default-features = false, features = ["rustls-tls"] }
reqwest-tracing = { version = "0.5", features = ["opentelemetry_0_24"] }
reqwest-tracing = { version = "0.5", features = ["opentelemetry_0_20"] }
reqwest-middleware = "0.3.0"
reqwest-retry = "0.5"
routerify = "3"
@@ -181,8 +177,8 @@ toml_edit = "0.22"
tonic = {version = "0.9", features = ["tls", "tls-roots"]}
tower-service = "0.3.2"
tracing = "0.1"
tracing-error = "0.2"
tracing-opentelemetry = "0.25"
tracing-error = "0.2.0"
tracing-opentelemetry = "0.21.0"
tracing-subscriber = { version = "0.3", default-features = false, features = ["smallvec", "fmt", "tracing-log", "std", "env-filter", "json"] }
try-lock = "0.2.5"
twox-hash = { version = "1.6.3", default-features = false }

View File

@@ -58,7 +58,7 @@ curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
1. Install XCode and dependencies
```
xcode-select --install
brew install protobuf openssl flex bison icu4c pkg-config m4
brew install protobuf openssl flex bison icu4c pkg-config
# add openssl to PATH, required for ed25519 keys generation in neon_local
echo 'export PATH="$(brew --prefix openssl)/bin:$PATH"' >> ~/.zshrc

View File

@@ -12,25 +12,10 @@ ARG DEBIAN_FLAVOR=bullseye-slim
#########################################################################################
FROM debian:$DEBIAN_FLAVOR AS build-deps
ARG DEBIAN_FLAVOR
RUN case $DEBIAN_FLAVOR in \
# Version-specific installs for Bullseye (PG14-PG16):
# The h3_pg extension needs a cmake 3.20+, but Debian bullseye has 3.18.
# Install newer version (3.25) from backports.
bullseye*) \
echo "deb http://deb.debian.org/debian bullseye-backports main" > /etc/apt/sources.list.d/bullseye-backports.list; \
VERSION_INSTALLS="cmake/bullseye-backports cmake-data/bullseye-backports"; \
;; \
# Version-specific installs for Bookworm (PG17):
bookworm*) \
VERSION_INSTALLS="cmake"; \
;; \
esac && \
apt update && \
apt install --no-install-recommends -y git autoconf automake libtool build-essential bison flex libreadline-dev \
zlib1g-dev libxml2-dev libcurl4-openssl-dev libossp-uuid-dev wget ca-certificates pkg-config libssl-dev \
libicu-dev libxslt1-dev liblz4-dev libzstd-dev zstd \
$VERSION_INSTALLS
RUN apt update && \
apt install -y git autoconf automake libtool build-essential bison flex libreadline-dev \
zlib1g-dev libxml2-dev libcurl4-openssl-dev libossp-uuid-dev wget pkg-config libssl-dev \
libicu-dev libxslt1-dev liblz4-dev libzstd-dev zstd
#########################################################################################
#
@@ -104,7 +89,7 @@ FROM build-deps AS postgis-build
ARG PG_VERSION
COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
RUN apt update && \
apt install --no-install-recommends -y gdal-bin libboost-dev libboost-thread-dev libboost-filesystem-dev \
apt install -y cmake gdal-bin libboost-dev libboost-thread-dev libboost-filesystem-dev \
libboost-system-dev libboost-iostreams-dev libboost-program-options-dev libboost-timer-dev \
libcgal-dev libgdal-dev libgmp-dev libmpfr-dev libopenscenegraph-dev libprotobuf-c-dev \
protobuf-c-compiler xsltproc
@@ -182,7 +167,7 @@ RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
apt update && \
apt install --no-install-recommends -y ninja-build python3-dev libncurses5 binutils clang
apt install -y ninja-build python3-dev libncurses5 binutils clang
RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
@@ -215,6 +200,27 @@ FROM build-deps AS h3-pg-build
ARG PG_VERSION
COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
case "$(uname -m)" in \
"x86_64") \
export CMAKE_CHECKSUM=739d372726cb23129d57a539ce1432453448816e345e1545f6127296926b6754 \
;; \
"aarch64") \
export CMAKE_CHECKSUM=281b42627c9a1beed03e29706574d04c6c53fae4994472e90985ef018dd29c02 \
;; \
*) \
echo "Unsupported architecture '$(uname -m)'. Supported are x86_64 and aarch64" && exit 1 \
;; \
esac && \
wget https://github.com/Kitware/CMake/releases/download/v3.24.2/cmake-3.24.2-linux-$(uname -m).sh \
-q -O /tmp/cmake-install.sh \
&& echo "${CMAKE_CHECKSUM} /tmp/cmake-install.sh" | sha256sum --check \
&& chmod u+x /tmp/cmake-install.sh \
&& /tmp/cmake-install.sh --skip-license --prefix=/usr/local/ \
&& rm /tmp/cmake-install.sh
RUN case "${PG_VERSION}" in "v17") \
mkdir -p /h3/usr/ && \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
@@ -500,6 +506,8 @@ RUN case "${PG_VERSION}" in "v17") \
export TIMESCALEDB_CHECKSUM=584a351c7775f0e067eaa0e7277ea88cab9077cc4c455cbbf09a5d9723dce95d \
;; \
esac && \
apt-get update && \
apt-get install -y cmake && \
wget https://github.com/timescale/timescaledb/archive/refs/tags/${TIMESCALEDB_VERSION}.tar.gz -O timescaledb.tar.gz && \
echo "${TIMESCALEDB_CHECKSUM} timescaledb.tar.gz" | sha256sum --check && \
mkdir timescaledb-src && cd timescaledb-src && tar xzf ../timescaledb.tar.gz --strip-components=1 -C . && \
@@ -587,7 +595,8 @@ RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
apt-get update && \
apt-get install --no-install-recommends -y \
apt-get install -y \
cmake \
libboost-iostreams1.74-dev \
libboost-regex1.74-dev \
libboost-serialization1.74-dev \
@@ -752,7 +761,7 @@ ARG PG_VERSION
COPY --from=pg-build /usr/local/pgsql/ /usr/local/pgsql/
RUN apt-get update && \
apt-get install --no-install-recommends -y curl libclang-dev && \
apt-get install -y curl libclang-dev cmake && \
useradd -ms /bin/bash nonroot -b /home
ENV HOME=/home/nonroot
@@ -862,25 +871,6 @@ RUN case "${PG_VERSION}" in "v17") \
cargo pgrx install --release && \
echo "trusted = true" >> /usr/local/pgsql/share/extension/ulid.control
#########################################################################################
#
# Layer "pg-session-jwt-build"
# Compile "pg_session_jwt" extension
#
#########################################################################################
FROM rust-extensions-build AS pg-session-jwt-build
ARG PG_VERSION
RUN case "${PG_VERSION}" in "v17") \
echo "pg_session_jwt does not yet have a release that supports pg17" && exit 0;; \
esac && \
wget https://github.com/neondatabase/pg_session_jwt/archive/ff0a72440e8ff584dab24b3f9b7c00c56c660b8e.tar.gz -O pg_session_jwt.tar.gz && \
echo "1fbb2b5a339263bcf6daa847fad8bccbc0b451cea6a62e6d3bf232b0087f05cb pg_session_jwt.tar.gz" | sha256sum --check && \
mkdir pg_session_jwt-src && cd pg_session_jwt-src && tar xzf ../pg_session_jwt.tar.gz --strip-components=1 -C . && \
sed -i 's/pgrx = "=0.11.3"/pgrx = { version = "=0.11.3", features = [ "unsafe-postgres" ] }/g' Cargo.toml && \
cargo pgrx install --release
#########################################################################################
#
# Layer "wal2json-build"
@@ -977,7 +967,6 @@ COPY --from=timescaledb-pg-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-hint-plan-pg-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-cron-pg-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-pgx-ulid-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-session-jwt-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=rdkit-pg-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-uuidv7-pg-build /usr/local/pgsql/ /usr/local/pgsql/
COPY --from=pg-roaringbitmap-pg-build /usr/local/pgsql/ /usr/local/pgsql/
@@ -1055,12 +1044,9 @@ FROM debian:$DEBIAN_FLAVOR AS pgbouncer
ARG DEBIAN_FLAVOR
RUN set -e \
&& apt-get update \
&& apt-get install --no-install-recommends -y \
&& apt-get install -y \
build-essential \
git \
ca-certificates \
autoconf \
automake \
libevent-dev \
libtool \
pkg-config
@@ -1075,20 +1061,6 @@ RUN set -e \
&& make -j $(nproc) dist_man_MANS= \
&& make install dist_man_MANS=
#########################################################################################
#
# Compile the Neon-specific `local_proxy` binary
#
#########################################################################################
FROM $REPOSITORY/$IMAGE:$TAG AS local_proxy
ARG BUILD_TAG
ENV BUILD_TAG=$BUILD_TAG
USER nonroot
# Copy entire project to get Cargo.* files with proper dependencies for the whole project
COPY --chown=nonroot . .
RUN mold -run cargo build --locked --profile release-line-debug-size-lto --bin local_proxy
#########################################################################################
#
# Layers "postgres-exporter" and "sql-exporter"
@@ -1182,6 +1154,11 @@ RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
cd /ext-src/pgvector-src && patch -p1 <../pgvector.patch
# cmake is required for the h3 test
RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
apt-get update && apt-get install -y cmake
RUN case "${PG_VERSION}" in "v17") \
echo "v17 extensions are not supported yet. Quit" && exit 0;; \
esac && \
@@ -1208,6 +1185,7 @@ ENV PGDATABASE=postgres
#########################################################################################
FROM debian:$DEBIAN_FLAVOR
ARG DEBIAN_FLAVOR
ENV DEBIAN_FLAVOR=$DEBIAN_FLAVOR
# Add user postgres
RUN mkdir /var/db && useradd -m -d /var/db/postgres postgres && \
echo "postgres:test_console_pass" | chpasswd && \
@@ -1227,10 +1205,6 @@ COPY --from=compute-tools --chown=postgres /home/nonroot/target/release-line-deb
COPY --from=pgbouncer /usr/local/pgbouncer/bin/pgbouncer /usr/local/bin/pgbouncer
COPY --chmod=0666 --chown=postgres compute/etc/pgbouncer.ini /etc/pgbouncer.ini
# local_proxy and its config
COPY --from=local_proxy --chown=postgres /home/nonroot/target/release-line-debug-size-lto/local_proxy /usr/local/bin/local_proxy
RUN mkdir -p /etc/local_proxy && chown postgres:postgres /etc/local_proxy
# Metrics exporter binaries and configuration files
COPY --from=postgres-exporter /bin/postgres_exporter /bin/postgres_exporter
COPY --from=sql-exporter /bin/sql_exporter /bin/sql_exporter
@@ -1284,7 +1258,7 @@ RUN apt update && \
libxml2 \
libxslt1.1 \
libzstd1 \
libcurl4 \
libcurl4-openssl-dev \
locales \
procps \
ca-certificates \

View File

@@ -94,68 +94,6 @@ metrics:
query: |
select sum(pg_database_size(datname)) as total from pg_database;
- metric_name: getpage_wait_seconds_count
type: counter
help: 'Number of getpage requests'
values: [getpage_wait_seconds_count]
query_ref: neon_perf_counters
- metric_name: getpage_wait_seconds_sum
type: counter
help: 'Time spent in getpage requests'
values: [getpage_wait_seconds_sum]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_requests_total
type: counter
help: 'Number of getpage issued for prefetching'
values: [getpage_prefetch_requests_total]
query_ref: neon_perf_counters
- metric_name: getpage_sync_requests_total
type: counter
help: 'Number of synchronous getpage issued'
values: [getpage_sync_requests_total]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_misses_total
type: counter
help: 'Total number of readahead misses; consisting of either prefetches that don''t satisfy the LSN bounds once the prefetch got read by the backend, or cases where somehow no readahead was issued for the read'
values: [getpage_prefetch_misses_total]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_discards_total
type: counter
help: 'Number of prefetch responses issued but not used'
values: [getpage_prefetch_discards_total]
query_ref: neon_perf_counters
- metric_name: pageserver_requests_sent_total
type: counter
help: 'Number of all requests sent to the pageserver (not just GetPage requests)'
values: [pageserver_requests_sent_total]
query_ref: neon_perf_counters
- metric_name: pageserver_disconnects_total
type: counter
help: 'Number of times that the connection to the pageserver was lost'
values: [pageserver_disconnects_total]
query_ref: neon_perf_counters
- metric_name: pageserver_send_flushes_total
type: counter
help: 'Number of flushes to the pageserver connection'
values: [pageserver_send_flushes_total]
query_ref: neon_perf_counters
- metric_name: getpage_wait_seconds_bucket
type: counter
help: 'Histogram buckets of getpage request latency'
key_labels:
- bucket_le
values: [value]
query_ref: getpage_wait_seconds_buckets
# DEPRECATED
- metric_name: lfc_approximate_working_set_size
type: gauge
@@ -306,26 +244,3 @@ metrics:
SELECT slot_name,
CASE WHEN wal_status = 'lost' THEN 1 ELSE 0 END AS wal_is_lost
FROM pg_replication_slots;
queries:
- query_name: neon_perf_counters
query: |
WITH c AS (
SELECT pg_catalog.jsonb_object_agg(metric, value) jb FROM neon.neon_perf_counters
)
SELECT d.*
FROM pg_catalog.jsonb_to_record((select jb from c)) as d(
getpage_wait_seconds_count numeric,
getpage_wait_seconds_sum numeric,
getpage_prefetch_requests_total numeric,
getpage_sync_requests_total numeric,
getpage_prefetch_misses_total numeric,
getpage_prefetch_discards_total numeric,
pageserver_requests_sent_total numeric,
pageserver_disconnects_total numeric,
pageserver_send_flushes_total numeric
);
- query_name: getpage_wait_seconds_buckets
query: |
SELECT bucket_le, value FROM neon.neon_perf_counters WHERE metric = 'getpage_wait_seconds_bucket';

View File

@@ -19,10 +19,6 @@ commands:
user: postgres
sysvInitAction: respawn
shell: '/usr/local/bin/pgbouncer /etc/pgbouncer.ini'
- name: local_proxy
user: postgres
sysvInitAction: respawn
shell: '/usr/local/bin/local_proxy --config-path /etc/local_proxy/config.json --pid-path /etc/local_proxy/pid --http 0.0.0.0:10432'
- name: postgres-exporter
user: nobody
sysvInitAction: respawn

View File

@@ -11,18 +11,16 @@ testing = []
[dependencies]
anyhow.workspace = true
camino.workspace = true
chrono.workspace = true
cfg-if.workspace = true
clap.workspace = true
flate2.workspace = true
futures.workspace = true
hyper0 = { workspace = true, features = ["full"] }
hyper = { workspace = true, features = ["full"] }
nix.workspace = true
notify.workspace = true
num_cpus.workspace = true
opentelemetry.workspace = true
opentelemetry_sdk.workspace = true
postgres.workspace = true
regex.workspace = true
serde_json.workspace = true

View File

@@ -218,7 +218,7 @@ fn startup_context_from_env() -> Option<opentelemetry::ContextGuard> {
}
if !startup_tracing_carrier.is_empty() {
use opentelemetry::propagation::TextMapPropagator;
use opentelemetry_sdk::propagation::TraceContextPropagator;
use opentelemetry::sdk::propagation::TraceContextPropagator;
let guard = TraceContextPropagator::new()
.extract(&startup_tracing_carrier)
.attach();

View File

@@ -34,7 +34,6 @@ use nix::sys::signal::{kill, Signal};
use remote_storage::{DownloadError, RemotePath};
use crate::checker::create_availability_check_data;
use crate::local_proxy;
use crate::logger::inlinify;
use crate::pg_helpers::*;
use crate::spec::*;
@@ -887,11 +886,6 @@ impl ComputeNode {
// 'Close' connection
drop(client);
if let Some(ref local_proxy) = spec.local_proxy_config {
info!("configuring local_proxy");
local_proxy::configure(local_proxy).context("apply_config local_proxy")?;
}
// Run migrations separately to not hold up cold starts
thread::spawn(move || {
let mut connstr = connstr.clone();
@@ -942,19 +936,6 @@ impl ComputeNode {
});
}
if let Some(ref local_proxy) = spec.local_proxy_config {
info!("configuring local_proxy");
// Spawn a thread to do the configuration,
// so that we don't block the main thread that starts Postgres.
let local_proxy = local_proxy.clone();
let _handle = Some(thread::spawn(move || {
if let Err(err) = local_proxy::configure(&local_proxy) {
error!("error while configuring local_proxy: {err:?}");
}
}));
}
// Write new config
let pgdata_path = Path::new(&self.pgdata);
let postgresql_conf_path = pgdata_path.join("postgresql.conf");
@@ -1042,19 +1023,6 @@ impl ComputeNode {
});
}
if let Some(local_proxy) = &pspec.spec.local_proxy_config {
info!("configuring local_proxy");
// Spawn a thread to do the configuration,
// so that we don't block the main thread that starts Postgres.
let local_proxy = local_proxy.clone();
let _handle = thread::spawn(move || {
if let Err(err) = local_proxy::configure(&local_proxy) {
error!("error while configuring local_proxy: {err:?}");
}
});
}
info!(
"start_compute spec.remote_extensions {:?}",
pspec.spec.remote_extensions
@@ -1092,26 +1060,19 @@ impl ComputeNode {
let pg_process = self.start_postgres(pspec.storage_auth_token.clone())?;
let config_time = Utc::now();
if pspec.spec.mode == ComputeMode::Primary {
if !pspec.spec.skip_pg_catalog_updates {
let pgdata_path = Path::new(&self.pgdata);
// temporarily reset max_cluster_size in config
// to avoid the possibility of hitting the limit, while we are applying config:
// creating new extensions, roles, etc...
config::with_compute_ctl_tmp_override(
pgdata_path,
"neon.max_cluster_size=-1",
|| {
self.pg_reload_conf()?;
self.apply_config(&compute_state)?;
Ok(())
},
)?;
if pspec.spec.mode == ComputeMode::Primary && !pspec.spec.skip_pg_catalog_updates {
let pgdata_path = Path::new(&self.pgdata);
// temporarily reset max_cluster_size in config
// to avoid the possibility of hitting the limit, while we are applying config:
// creating new extensions, roles, etc...
config::with_compute_ctl_tmp_override(pgdata_path, "neon.max_cluster_size=-1", || {
self.pg_reload_conf()?;
}
self.post_apply_config()?;
self.apply_config(&compute_state)?;
Ok(())
})?;
self.pg_reload_conf()?;
}
let startup_end_time = Utc::now();

View File

@@ -2,9 +2,6 @@
//! configuration.
#![deny(unsafe_code)]
#![deny(clippy::undocumented_unsafe_blocks)]
extern crate hyper0 as hyper;
pub mod checker;
pub mod config;
pub mod configurator;
@@ -15,7 +12,6 @@ pub mod catalog;
pub mod compute;
pub mod disk_quota;
pub mod extension_server;
pub mod local_proxy;
pub mod lsn_lease;
mod migration;
pub mod monitor;

View File

@@ -1,56 +0,0 @@
//! Local Proxy is a feature of our BaaS Neon Authorize project.
//!
//! Local Proxy validates JWTs and manages the pg_session_jwt extension.
//! It also maintains a connection pool to postgres.
use anyhow::{Context, Result};
use camino::Utf8Path;
use compute_api::spec::LocalProxySpec;
use nix::sys::signal::Signal;
use utils::pid_file::{self, PidFileRead};
pub fn configure(local_proxy: &LocalProxySpec) -> Result<()> {
write_local_proxy_conf("/etc/local_proxy/config.json".as_ref(), local_proxy)?;
notify_local_proxy("/etc/local_proxy/pid".as_ref())?;
Ok(())
}
/// Create or completely rewrite configuration file specified by `path`
fn write_local_proxy_conf(path: &Utf8Path, local_proxy: &LocalProxySpec) -> Result<()> {
let config =
serde_json::to_string_pretty(local_proxy).context("serializing LocalProxySpec to json")?;
std::fs::write(path, config).with_context(|| format!("writing {path}"))?;
Ok(())
}
/// Notify local proxy about a new config file.
fn notify_local_proxy(path: &Utf8Path) -> Result<()> {
match pid_file::read(path)? {
// if the file doesn't exist, or isn't locked, local_proxy isn't running
// and will naturally pick up our config later
PidFileRead::NotExist | PidFileRead::NotHeldByAnyProcess(_) => {}
PidFileRead::LockedByOtherProcess(pid) => {
// From the pid_file docs:
//
// > 1. The other process might exit at any time, turning the given PID stale.
// > 2. There is a small window in which `claim_for_current_process` has already
// > locked the file but not yet updates its contents. [`read`] will return
// > this variant here, but with the old file contents, i.e., a stale PID.
// >
// > The kernel is free to recycle PID once it has been `wait(2)`ed upon by
// > its creator. Thus, acting upon a stale PID, e.g., by issuing a `kill`
// > system call on it, bears the risk of killing an unrelated process.
// > This is an inherent limitation of using pidfiles.
// > The only race-free solution is to have a supervisor-process with a lifetime
// > that exceeds that of all of its child-processes (e.g., `runit`, `supervisord`).
//
// This is an ok risk as we only send a SIGHUP which likely won't actually
// kill the process, only reload config.
nix::sys::signal::kill(pid, Signal::SIGHUP).context("sending signal to local_proxy")?;
}
}
Ok(())
}

View File

@@ -1,3 +1,4 @@
use tracing_opentelemetry::OpenTelemetryLayer;
use tracing_subscriber::layer::SubscriberExt;
use tracing_subscriber::prelude::*;
@@ -22,7 +23,8 @@ pub fn init_tracing_and_logging(default_log_level: &str) -> anyhow::Result<()> {
.with_writer(std::io::stderr);
// Initialize OpenTelemetry
let otlp_layer = tracing_utils::init_tracing_without_runtime("compute_ctl");
let otlp_layer =
tracing_utils::init_tracing_without_runtime("compute_ctl").map(OpenTelemetryLayer::new);
// Put it all together
tracing_subscriber::registry()

View File

@@ -14,7 +14,7 @@ humantime.workspace = true
nix.workspace = true
once_cell.workspace = true
humantime-serde.workspace = true
hyper0.workspace = true
hyper.workspace = true
regex.workspace = true
reqwest = { workspace = true, features = ["blocking", "json"] }
scopeguard.workspace = true

View File

@@ -599,7 +599,6 @@ impl Endpoint {
remote_extensions,
pgbouncer_settings: None,
shard_stripe_size: Some(shard_stripe_size),
local_proxy_config: None,
};
let spec_path = self.endpoint_path().join("spec.json");
std::fs::write(spec_path, serde_json::to_string_pretty(&spec)?)?;

View File

@@ -168,9 +168,6 @@ pub struct NeonStorageControllerConf {
#[serde(with = "humantime_serde")]
pub heartbeat_interval: Duration,
#[serde(with = "humantime_serde")]
pub long_reconcile_threshold: Option<Duration>,
}
impl NeonStorageControllerConf {
@@ -193,7 +190,6 @@ impl Default for NeonStorageControllerConf {
split_threshold: None,
max_secondary_lag_bytes: None,
heartbeat_interval: Self::DEFAULT_HEARTBEAT_INTERVAL,
long_reconcile_threshold: None,
}
}
}

View File

@@ -3,7 +3,7 @@ use crate::{
local_env::{LocalEnv, NeonStorageControllerConf},
};
use camino::{Utf8Path, Utf8PathBuf};
use hyper0::Uri;
use hyper::Uri;
use nix::unistd::Pid;
use pageserver_api::{
controller_api::{
@@ -517,13 +517,6 @@ impl StorageController {
args.push(format!("--max-secondary-lag-bytes={lag}"))
}
if let Some(threshold) = self.config.long_reconcile_threshold {
args.push(format!(
"--long-reconcile-threshold={}",
humantime::Duration::from(threshold)
))
}
args.push(format!(
"--neon-local-repo-dir={}",
self.env.base_data_dir.display()

View File

@@ -5,7 +5,7 @@
Currently we build two main images:
- [neondatabase/neon](https://hub.docker.com/repository/docker/neondatabase/neon) — image with pre-built `pageserver`, `safekeeper` and `proxy` binaries and all the required runtime dependencies. Built from [/Dockerfile](/Dockerfile).
- [neondatabase/compute-node-v16](https://hub.docker.com/repository/docker/neondatabase/compute-node-v16) — compute node image with pre-built Postgres binaries from [neondatabase/postgres](https://github.com/neondatabase/postgres). Similar images exist for v15 and v14. Built from [/compute-node/Dockerfile](/compute/Dockerfile.compute-node).
- [neondatabase/compute-node-v16](https://hub.docker.com/repository/docker/neondatabase/compute-node-v16) — compute node image with pre-built Postgres binaries from [neondatabase/postgres](https://github.com/neondatabase/postgres). Similar images exist for v15 and v14.
And additional intermediate image:

View File

@@ -1,112 +0,0 @@
# AUX file v2
## Summary
This is a retrospective RFC describing a new storage strategy for AUX files.
## Motivation
The original aux file storage strategy stores everything in a single `AUX_FILES_KEY`.
Every time the compute node streams a `neon-file` record to the pageserver, it will
update the aux file hash map, and then write the serialized hash map into the key.
This creates serious space bloat. There was a fix to log delta records (i.e., update
a key in the hash map) to the aux file key. In this way, the pageserver only stores
the deltas at each of the LSNs. However, this improved v1 storage strategy still
requires us to store everything in an aux file cache in memory, because we cannot
fetch a single key (or file) from the compound `AUX_FILES_KEY`.
### Prior art
For storing large amount of small files, we can use a key-value store where the key
is the filename and the value is the file content.
## Requirements
- No space bloat, fixed space amplification.
- No write bloat, fixed write amplification.
## Impacted Components
pageserver
## Sparse Keyspace
In pageserver, we had assumed the keyspaces are always contiguous. For example, if the keyspace 0x0000-0xFFFF
exists in the pageserver, every single key in the key range would exist in the storage. Based on the prior
assumption, there are code that traverses the keyspace by iterating every single key.
```rust
loop {
// do something
key = key.next();
}
```
If a keyspace is very large, for example, containing `2^64` keys, this loop will take infinite time to run.
Therefore, we introduce the concept of sparse keyspace in this RFC. For a sparse keyspace, not every key would
exist in the key range. Developers should not attempt to iterate every single key in the keyspace. Instead,
they should fetch all the layer files in the key range, and then do a merge of them.
In aux file v2, we store aux files within the sparse keyspace of the prefix `AUX_KEY_PREFIX`.
## AUX v2 Keyspace and Key Mapping
Pageserver uses fixed-size keys. The key is 128b. In order to store files of arbitrary filenames into the
keyspace, we assign a predetermined prefix based on the directory storing the aux file, and use the FNV hash
of the filename for the rest bits of the key. The encoding scheme is defined in `encode_aux_file_key`.
For example, `pg_logical/mappings/test1` will be encoded as:
```
62 0000 01 01 7F8B83D94F7081693471ABF91C
^ aux prefix
^ assigned prefix of pg_logical/
^ assigned prefix of mappings/
^ 13B FNV hash of test1
^ not used due to key representation
```
The prefixes of the directories should be assigned every time we add a new type of aux file into the storage within `aux_file.rs`. For all directories without an assigned prefix, it will be put into the `0xFFFF` keyspace.
Note that inside pageserver, there are two representations of the keys: the 18B full key representation
and the 16B compact key representation. For the 18B representation, some fields have restricted ranges
of values. Therefore, the aux keys only use the 16B compact portion of the full key.
It is possible that two files get mapped to the same key due to hash collision. Therefore, the value of
each of the aux key is an array that contains all filenames and file content that should be stored in
this key.
We use `Value::Image` to store the aux keys. Therefore, page reconstruction works in the same way as before,
and we do not need addition code to support reconstructing the value. We simply get the latest image from
the storage.
## Inbound Logical Replication Key Mapping
For inbound logical replication, Postgres needs the `replorigin_checkpoint` file to store the data.
This file not directly stored in the pageserver using the aux v2 mechanism. It is constructed during
generating the basebackup by scanning the `REPL_ORIGIN_KEY_PREFIX` keyspace.
## Sparse Keyspace Read Path
There are two places we need to read the aux files from the pageserver:
* On the write path, when the compute node adds an aux file to the pageserver, we will retrieve the key from the storage, append the file to the hashed key, and write it back. The current `get` API already supports that.
* We use the vectored get API to retrieve all aux files during generating the basebackup. Because we need to scan a sparse keyspace, we slightly modified the vectored get path. The vectorized API will attempt to retrieve every single key within the requested key range, and therefore, we modified it in a way that keys within `NON_INHERITED_SPARSE_RANGE` will not trigger missing key error.
## Compaction and Image Layer Generation
With the add of sparse keyspaces, we also modified the compaction code to accommodate the fact that sparse keyspaces do not have every single key stored in the storage.
* L0 compaction: we modified the hole computation code so that it can handle sparse keyspaces when computing holes.
* Image layer creation: instead of calling `key.next()` and getting/reconstructing images for every single key, we use the vectored get API to scan all keys in the keyspace at a given LSN. Image layers are only created if there are too many delta layers between the latest LSN and the last image layer we generated for sparse keyspaces. The created image layer always cover the full aux key range for now, and could be optimized later.
## Migration
We decided not to make the new aux storage strategy (v1) compatible with the original one (v1). One feasible way of doing a seamless migration is to store new data in aux v2 while old data in aux v1, but this complicates file deletions. We want all users to start with a clean state with no aux files in the storage, and therefore, we need to do manual migrations for users using aux v1 by using the [migration script](https://github.com/neondatabase/aux_v2_migration).
During the period of migration, we store the aux policy in the `index_part.json` file. When a tenant is attached
with no policy set, the pageserver will scan the aux file keyspaces to identify the current aux policy being used (v1 or v2).
If a timeline has aux v1 files stored, it will use aux file policy v1 unless we do a manual migration for them. Otherwise, the default aux file policy for new timelines is aux v2. Users enrolled in logical replication before we set aux v2 as default use aux v1 policy. Users who tried setting up inbound replication (which was not supported at that time) may also create some file entries in aux v1 store, even if they did not enroll in the logical replication testing program.
The code for aux v2 migration is in https://github.com/neondatabase/aux_v2_migration. The toolkit scans all projects with logical replication enabled. For all these projects, it put the computes into maintenance mode (suspend all of then), call the migration API to switch the aux file policy on the pageserver (which drops all replication states), and restart all the computes.

View File

@@ -106,10 +106,6 @@ pub struct ComputeSpec {
// Stripe size for pageserver sharding, in pages
#[serde(default)]
pub shard_stripe_size: Option<usize>,
/// Local Proxy configuration used for JWT authentication
#[serde(default)]
pub local_proxy_config: Option<LocalProxySpec>,
}
/// Feature flag to signal `compute_ctl` to enable certain experimental functionality.
@@ -282,13 +278,11 @@ pub struct GenericOption {
/// declare a `trait` on it.
pub type GenericOptions = Option<Vec<GenericOption>>;
/// Configured the local_proxy application with the relevant JWKS and roles it should
/// Configured the local-proxy application with the relevant JWKS and roles it should
/// use for authorizing connect requests using JWT.
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct LocalProxySpec {
#[serde(default)]
#[serde(skip_serializing_if = "Option::is_none")]
pub jwks: Option<Vec<JwksSettings>>,
pub jwks: Vec<JwksSettings>,
}
#[derive(Clone, Debug, Deserialize, Serialize)]

View File

@@ -104,7 +104,7 @@ pub struct ConfigToml {
pub image_compression: ImageCompressionAlgorithm,
pub ephemeral_bytes_per_memory_kb: usize,
pub l0_flush: Option<crate::models::L0FlushConfig>,
pub virtual_file_direct_io: crate::models::virtual_file::DirectIoMode,
pub virtual_file_io_mode: Option<crate::models::virtual_file::IoMode>,
pub io_buffer_alignment: usize,
}
@@ -296,14 +296,7 @@ pub mod defaults {
pub const DEFAULT_INGEST_BATCH_SIZE: u64 = 100;
/// Soft limit for the maximum size of a vectored read.
///
/// This is determined by the largest NeonWalRecord that can exist (minus dbdir and reldir keys
/// which are bounded by the blob io limits only). As of this writing, that is a `NeonWalRecord::ClogSetCommitted` record,
/// with 32k xids. That's the max number of XIDS on a single CLOG page. The size of such a record
/// is `sizeof(Transactionid) * 32768 + (some fixed overhead from 'timestamp`, the Vec length and whatever extra serde serialization adds)`.
/// That is, slightly above 128 kB.
pub const DEFAULT_MAX_VECTORED_READ_BYTES: usize = 130 * 1024; // 130 KiB
pub const DEFAULT_MAX_VECTORED_READ_BYTES: usize = 128 * 1024; // 128 KiB
pub const DEFAULT_IMAGE_COMPRESSION: ImageCompressionAlgorithm =
ImageCompressionAlgorithm::Zstd { level: Some(1) };
@@ -388,7 +381,7 @@ impl Default for ConfigToml {
image_compression: (DEFAULT_IMAGE_COMPRESSION),
ephemeral_bytes_per_memory_kb: (DEFAULT_EPHEMERAL_BYTES_PER_MEMORY_KB),
l0_flush: None,
virtual_file_direct_io: crate::models::virtual_file::DirectIoMode::default(),
virtual_file_io_mode: None,
io_buffer_alignment: DEFAULT_IO_BUFFER_ALIGNMENT,

View File

@@ -748,16 +748,6 @@ impl Key {
self.field1 == 0x00 && self.field4 != 0 && self.field6 != 0xffffffff
}
#[inline(always)]
pub fn is_rel_dir_key(&self) -> bool {
self.field1 == 0x00
&& self.field2 != 0
&& self.field3 != 0
&& self.field4 == 0
&& self.field5 == 0
&& self.field6 == 1
}
/// Guaranteed to return `Ok()` if [`Self::is_rel_block_key`] returns `true` for `key`.
#[inline(always)]
pub fn to_rel_block(self) -> anyhow::Result<(RelTag, BlockNumber)> {

View File

@@ -972,8 +972,6 @@ pub struct TopTenantShardsResponse {
}
pub mod virtual_file {
use std::path::PathBuf;
#[derive(
Copy,
Clone,
@@ -994,50 +992,49 @@ pub mod virtual_file {
}
/// Direct IO modes for a pageserver.
#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize, Default)]
#[serde(tag = "mode", rename_all = "kebab-case", deny_unknown_fields)]
pub enum DirectIoMode {
/// Direct IO disabled (uses usual buffered IO).
#[default]
Disabled,
/// Direct IO disabled (performs checks and perf simulations).
Evaluate {
/// Alignment check level
alignment_check: DirectIoAlignmentCheckLevel,
/// Latency padded for performance simulation.
latency_padding: DirectIoLatencyPadding,
},
/// Direct IO enabled.
Enabled {
/// Actions to perform on alignment error.
on_alignment_error: DirectIoOnAlignmentErrorAction,
},
#[derive(
Copy,
Clone,
PartialEq,
Eq,
Hash,
strum_macros::EnumString,
strum_macros::Display,
serde_with::DeserializeFromStr,
serde_with::SerializeDisplay,
Debug,
)]
#[strum(serialize_all = "kebab-case")]
#[repr(u8)]
pub enum IoMode {
/// Uses buffered IO.
Buffered,
/// Uses direct IO, error out if the operation fails.
#[cfg(target_os = "linux")]
Direct,
}
#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize, Default)]
#[serde(rename_all = "kebab-case")]
pub enum DirectIoAlignmentCheckLevel {
#[default]
Error,
Log,
None,
impl IoMode {
pub const fn preferred() -> Self {
if cfg!(target_os = "linux") {
Self::Direct
} else {
Self::Buffered
}
}
}
#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize, Default)]
#[serde(rename_all = "kebab-case")]
pub enum DirectIoOnAlignmentErrorAction {
Error,
#[default]
FallbackToBuffered,
}
impl TryFrom<u8> for IoMode {
type Error = u8;
#[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize, serde::Serialize, Default)]
#[serde(tag = "type", rename_all = "kebab-case")]
pub enum DirectIoLatencyPadding {
/// Pad virtual file operations with IO to a fake file.
FakeFileRW { path: PathBuf },
#[default]
None,
fn try_from(value: u8) -> Result<Self, Self::Error> {
Ok(match value {
v if v == (IoMode::Buffered as u8) => IoMode::Buffered,
#[cfg(target_os = "linux")]
v if v == (IoMode::Direct as u8) => IoMode::Direct,
x => return Err(x),
})
}
}
}

View File

@@ -16,7 +16,7 @@ aws-sdk-s3.workspace = true
bytes.workspace = true
camino = { workspace = true, features = ["serde1"] }
humantime-serde.workspace = true
hyper0 = { workspace = true, features = ["stream"] }
hyper = { workspace = true, features = ["stream"] }
futures.workspace = true
serde.workspace = true
serde_json.workspace = true

View File

@@ -14,7 +14,7 @@ use std::time::SystemTime;
use super::REMOTE_STORAGE_PREFIX_SEPARATOR;
use anyhow::Result;
use azure_core::request_options::{IfMatchCondition, MaxResults, Metadata, Range};
use azure_core::request_options::{MaxResults, Metadata, Range};
use azure_core::{Continuable, RetryOptions};
use azure_identity::DefaultAzureCredential;
use azure_storage::StorageCredentials;
@@ -33,10 +33,10 @@ use tracing::debug;
use utils::backoff;
use crate::metrics::{start_measuring_requests, AttemptOutcome, RequestKind};
use crate::ListingObject;
use crate::{
config::AzureConfig, error::Cancelled, ConcurrencyLimiter, Download, DownloadError,
DownloadOpts, Listing, ListingMode, ListingObject, RemotePath, RemoteStorage, StorageMetadata,
TimeTravelError, TimeoutOrCancel,
config::AzureConfig, error::Cancelled, ConcurrencyLimiter, Download, DownloadError, Listing,
ListingMode, RemotePath, RemoteStorage, StorageMetadata, TimeTravelError, TimeoutOrCancel,
};
pub struct AzureBlobStorage {
@@ -259,7 +259,6 @@ fn to_download_error(error: azure_core::Error) -> DownloadError {
if let Some(http_err) = error.as_http_error() {
match http_err.status() {
StatusCode::NotFound => DownloadError::NotFound,
StatusCode::NotModified => DownloadError::Unmodified,
StatusCode::BadRequest => DownloadError::BadInput(anyhow::Error::new(error)),
_ => DownloadError::Other(anyhow::Error::new(error)),
}
@@ -485,16 +484,11 @@ impl RemoteStorage for AzureBlobStorage {
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
let blob_client = self.client.blob_client(self.relative_path_to_name(from));
let mut builder = blob_client.get();
if let Some(ref etag) = opts.etag {
builder = builder.if_match(IfMatchCondition::NotMatch(etag.to_string()))
}
let builder = blob_client.get();
self.download_for_builder(builder, cancel).await
}

View File

@@ -5,8 +5,6 @@ pub enum DownloadError {
BadInput(anyhow::Error),
/// The file was not found in the remote storage.
NotFound,
/// The caller provided an ETag, and the file was not modified.
Unmodified,
/// A cancellation token aborted the download, typically during
/// tenant detach or process shutdown.
Cancelled,
@@ -26,7 +24,6 @@ impl std::fmt::Display for DownloadError {
write!(f, "Failed to download a remote file due to user input: {e}")
}
DownloadError::NotFound => write!(f, "No file found for the remote object id given"),
DownloadError::Unmodified => write!(f, "File was not modified"),
DownloadError::Cancelled => write!(f, "Cancelled, shutting down"),
DownloadError::Timeout => write!(f, "timeout"),
DownloadError::Other(e) => write!(f, "Failed to download a remote file: {e:?}"),
@@ -41,7 +38,7 @@ impl DownloadError {
pub fn is_permanent(&self) -> bool {
use DownloadError::*;
match self {
BadInput(_) | NotFound | Unmodified | Cancelled => true,
BadInput(_) | NotFound | Cancelled => true,
Timeout | Other(_) => false,
}
}

View File

@@ -161,14 +161,6 @@ pub struct Listing {
pub keys: Vec<ListingObject>,
}
/// Options for downloads. The default value is a plain GET.
#[derive(Default)]
pub struct DownloadOpts {
/// If given, returns [`DownloadError::Unmodified`] if the object still has
/// the same ETag (using If-None-Match).
pub etag: Option<Etag>,
}
/// Storage (potentially remote) API to manage its state.
/// This storage tries to be unaware of any layered repository context,
/// providing basic CRUD operations for storage files.
@@ -253,7 +245,6 @@ pub trait RemoteStorage: Send + Sync + 'static {
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError>;
@@ -410,18 +401,16 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
}
}
/// See [`RemoteStorage::download`]
pub async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
match self {
Self::LocalFs(s) => s.download(from, opts, cancel).await,
Self::AwsS3(s) => s.download(from, opts, cancel).await,
Self::AzureBlob(s) => s.download(from, opts, cancel).await,
Self::Unreliable(s) => s.download(from, opts, cancel).await,
Self::LocalFs(s) => s.download(from, cancel).await,
Self::AwsS3(s) => s.download(from, cancel).await,
Self::AzureBlob(s) => s.download(from, cancel).await,
Self::Unreliable(s) => s.download(from, cancel).await,
}
}
@@ -583,7 +572,7 @@ impl GenericRemoteStorage {
) -> Result<Download, DownloadError> {
match byte_range {
Some((start, end)) => self.download_byte_range(from, start, end, cancel).await,
None => self.download(from, &DownloadOpts::default(), cancel).await,
None => self.download(from, cancel).await,
}
}

View File

@@ -23,8 +23,8 @@ use tokio_util::{io::ReaderStream, sync::CancellationToken};
use utils::crashsafe::path_with_suffix_extension;
use crate::{
Download, DownloadError, DownloadOpts, Listing, ListingMode, ListingObject, RemotePath,
TimeTravelError, TimeoutOrCancel, REMOTE_STORAGE_PREFIX_SEPARATOR,
Download, DownloadError, Listing, ListingMode, ListingObject, RemotePath, TimeTravelError,
TimeoutOrCancel, REMOTE_STORAGE_PREFIX_SEPARATOR,
};
use super::{RemoteStorage, StorageMetadata};
@@ -494,17 +494,11 @@ impl RemoteStorage for LocalFs {
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
let target_path = from.with_base(&self.storage_root);
let file_metadata = file_metadata(&target_path).await?;
let etag = mock_etag(&file_metadata);
if opts.etag.as_ref() == Some(&etag) {
return Err(DownloadError::Unmodified);
}
let source = ReaderStream::new(
fs::OpenOptions::new()
@@ -525,6 +519,7 @@ impl RemoteStorage for LocalFs {
let cancel_or_timeout = crate::support::cancel_or_timeout(self.timeout, cancel.clone());
let source = crate::support::DownloadStream::new(cancel_or_timeout, source);
let etag = mock_etag(&file_metadata);
Ok(Download {
metadata,
last_modified: file_metadata
@@ -697,7 +692,7 @@ mod fs_tests {
) -> anyhow::Result<String> {
let cancel = CancellationToken::new();
let download = storage
.download(remote_storage_path, &DownloadOpts::default(), &cancel)
.download(remote_storage_path, &cancel)
.await
.map_err(|e| anyhow::anyhow!("Download failed: {e}"))?;
ensure!(
@@ -778,8 +773,8 @@ mod fs_tests {
"We should upload and download the same contents"
);
let non_existing_path = RemotePath::new(Utf8Path::new("somewhere/else"))?;
match storage.download(&non_existing_path, &DownloadOpts::default(), &cancel).await {
let non_existing_path = "somewhere/else";
match storage.download(&RemotePath::new(Utf8Path::new(non_existing_path))?, &cancel).await {
Err(DownloadError::NotFound) => {} // Should get NotFound for non existing keys
other => panic!("Should get a NotFound error when downloading non-existing storage files, but got: {other:?}"),
}
@@ -1106,13 +1101,7 @@ mod fs_tests {
storage.upload(body, len, &path, None, &cancel).await?;
}
let read = aggregate(
storage
.download(&path, &DownloadOpts::default(), &cancel)
.await?
.download_stream,
)
.await?;
let read = aggregate(storage.download(&path, &cancel).await?.download_stream).await?;
assert_eq!(body, read);
let shorter = Bytes::from_static(b"shorter body");
@@ -1123,13 +1112,7 @@ mod fs_tests {
storage.upload(body, len, &path, None, &cancel).await?;
}
let read = aggregate(
storage
.download(&path, &DownloadOpts::default(), &cancel)
.await?
.download_stream,
)
.await?;
let read = aggregate(storage.download(&path, &cancel).await?.download_stream).await?;
assert_eq!(shorter, read);
Ok(())
}
@@ -1162,13 +1145,7 @@ mod fs_tests {
storage.upload(body, len, &path, None, &cancel).await?;
}
let read = aggregate(
storage
.download(&path, &DownloadOpts::default(), &cancel)
.await?
.download_stream,
)
.await?;
let read = aggregate(storage.download(&path, &cancel).await?.download_stream).await?;
assert_eq!(body, read);
Ok(())

View File

@@ -28,13 +28,12 @@ use aws_sdk_s3::{
Client,
};
use aws_smithy_async::rt::sleep::TokioSleep;
use http_types::StatusCode;
use aws_smithy_types::{body::SdkBody, DateTime};
use aws_smithy_types::{byte_stream::ByteStream, date_time::ConversionError};
use bytes::Bytes;
use futures::stream::Stream;
use hyper0::Body;
use hyper::Body;
use scopeguard::ScopeGuard;
use tokio_util::sync::CancellationToken;
use utils::backoff;
@@ -45,8 +44,8 @@ use crate::{
error::Cancelled,
metrics::{start_counting_cancelled_wait, start_measuring_requests},
support::PermitCarrying,
ConcurrencyLimiter, Download, DownloadError, DownloadOpts, Listing, ListingMode, ListingObject,
RemotePath, RemoteStorage, TimeTravelError, TimeoutOrCancel, MAX_KEYS_PER_DELETE,
ConcurrencyLimiter, Download, DownloadError, Listing, ListingMode, ListingObject, RemotePath,
RemoteStorage, TimeTravelError, TimeoutOrCancel, MAX_KEYS_PER_DELETE,
REMOTE_STORAGE_PREFIX_SEPARATOR,
};
@@ -68,7 +67,6 @@ pub struct S3Bucket {
struct GetObjectRequest {
bucket: String,
key: String,
etag: Option<String>,
range: Option<String>,
}
impl S3Bucket {
@@ -250,18 +248,13 @@ impl S3Bucket {
let started_at = start_measuring_requests(kind);
let mut builder = self
let get_object = self
.client
.get_object()
.bucket(request.bucket)
.key(request.key)
.set_range(request.range);
if let Some(etag) = request.etag {
builder = builder.if_none_match(etag);
}
let get_object = builder.send();
.set_range(request.range)
.send();
let get_object = tokio::select! {
res = get_object => res,
@@ -284,20 +277,6 @@ impl S3Bucket {
);
return Err(DownloadError::NotFound);
}
Err(SdkError::ServiceError(e))
// aws_smithy_runtime_api::http::response::StatusCode isn't
// re-exported by any aws crates, so just check the numeric
// status against http_types::StatusCode instead of pulling it.
if e.raw().status().as_u16() == StatusCode::NotModified =>
{
// Count an unmodified file as a success.
crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
kind,
AttemptOutcome::Ok,
started_at,
);
return Err(DownloadError::Unmodified);
}
Err(e) => {
crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
kind,
@@ -794,7 +773,6 @@ impl RemoteStorage for S3Bucket {
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
// if prefix is not none then download file `prefix/from`
@@ -803,7 +781,6 @@ impl RemoteStorage for S3Bucket {
GetObjectRequest {
bucket: self.bucket_name.clone(),
key: self.relative_path_to_s3_object(from),
etag: opts.etag.as_ref().map(|e| e.to_string()),
range: None,
},
cancel,
@@ -830,7 +807,6 @@ impl RemoteStorage for S3Bucket {
GetObjectRequest {
bucket: self.bucket_name.clone(),
key: self.relative_path_to_s3_object(from),
etag: None,
range,
},
cancel,

View File

@@ -12,8 +12,8 @@ use std::{collections::hash_map::Entry, sync::Arc};
use tokio_util::sync::CancellationToken;
use crate::{
Download, DownloadError, DownloadOpts, GenericRemoteStorage, Listing, ListingMode, RemotePath,
RemoteStorage, StorageMetadata, TimeTravelError,
Download, DownloadError, GenericRemoteStorage, Listing, ListingMode, RemotePath, RemoteStorage,
StorageMetadata, TimeTravelError,
};
pub struct UnreliableWrapper {
@@ -167,12 +167,11 @@ impl RemoteStorage for UnreliableWrapper {
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
self.attempt(RemoteOp::Download(from.clone()))
.map_err(DownloadError::Other)?;
self.inner.download(from, opts, cancel).await
self.inner.download(from, cancel).await
}
async fn download_byte_range(

View File

@@ -1,7 +1,8 @@
use anyhow::Context;
use camino::Utf8Path;
use futures::StreamExt;
use remote_storage::{DownloadError, DownloadOpts, ListingMode, ListingObject, RemotePath};
use remote_storage::ListingMode;
use remote_storage::RemotePath;
use std::sync::Arc;
use std::{collections::HashSet, num::NonZeroU32};
use test_context::test_context;
@@ -283,10 +284,7 @@ async fn upload_download_works(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<
ctx.client.upload(data, len, &path, None, &cancel).await?;
// Normal download request
let dl = ctx
.client
.download(&path, &DownloadOpts::default(), &cancel)
.await?;
let dl = ctx.client.download(&path, &cancel).await?;
let buf = download_to_vec(dl).await?;
assert_eq!(&buf, &orig);
@@ -339,54 +337,6 @@ async fn upload_download_works(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<
Ok(())
}
/// Tests that conditional downloads work properly, by returning
/// DownloadError::Unmodified when the object ETag matches the given ETag.
#[test_context(MaybeEnabledStorage)]
#[tokio::test]
async fn download_conditional(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<()> {
let MaybeEnabledStorage::Enabled(ctx) = ctx else {
return Ok(());
};
let cancel = CancellationToken::new();
// Create a file.
let path = RemotePath::new(Utf8Path::new(format!("{}/file", ctx.base_prefix).as_str()))?;
let data = bytes::Bytes::from_static("foo".as_bytes());
let (stream, len) = wrap_stream(data);
ctx.client.upload(stream, len, &path, None, &cancel).await?;
// Download it to obtain its etag.
let mut opts = DownloadOpts::default();
let download = ctx.client.download(&path, &opts, &cancel).await?;
// Download with the etag yields DownloadError::Unmodified.
opts.etag = Some(download.etag);
let result = ctx.client.download(&path, &opts, &cancel).await;
assert!(
matches!(result, Err(DownloadError::Unmodified)),
"expected DownloadError::Unmodified, got {result:?}"
);
// Replace the file contents.
let data = bytes::Bytes::from_static("bar".as_bytes());
let (stream, len) = wrap_stream(data);
ctx.client.upload(stream, len, &path, None, &cancel).await?;
// A download with the old etag should yield the new file.
let download = ctx.client.download(&path, &opts, &cancel).await?;
assert_ne!(download.etag, opts.etag.unwrap(), "ETag did not change");
// A download with the new etag should yield Unmodified again.
opts.etag = Some(download.etag);
let result = ctx.client.download(&path, &opts, &cancel).await;
assert!(
matches!(result, Err(DownloadError::Unmodified)),
"expected DownloadError::Unmodified, got {result:?}"
);
Ok(())
}
#[test_context(MaybeEnabledStorage)]
#[tokio::test]
async fn copy_works(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<()> {
@@ -414,10 +364,7 @@ async fn copy_works(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<()> {
// Normal download request
ctx.client.copy_object(&path, &path_dest, &cancel).await?;
let dl = ctx
.client
.download(&path_dest, &DownloadOpts::default(), &cancel)
.await?;
let dl = ctx.client.download(&path_dest, &cancel).await?;
let buf = download_to_vec(dl).await?;
assert_eq!(&buf, &orig);
@@ -429,56 +376,3 @@ async fn copy_works(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<()> {
Ok(())
}
/// Tests that head_object works properly.
#[test_context(MaybeEnabledStorage)]
#[tokio::test]
async fn head_object(ctx: &mut MaybeEnabledStorage) -> anyhow::Result<()> {
let MaybeEnabledStorage::Enabled(ctx) = ctx else {
return Ok(());
};
let cancel = CancellationToken::new();
let path = RemotePath::new(Utf8Path::new(format!("{}/file", ctx.base_prefix).as_str()))?;
// Errors on missing file.
let result = ctx.client.head_object(&path, &cancel).await;
assert!(
matches!(result, Err(DownloadError::NotFound)),
"expected NotFound, got {result:?}"
);
// Create the file.
let data = bytes::Bytes::from_static("foo".as_bytes());
let (stream, len) = wrap_stream(data);
ctx.client.upload(stream, len, &path, None, &cancel).await?;
// Fetch the head metadata.
let object = ctx.client.head_object(&path, &cancel).await?;
assert_eq!(
object,
ListingObject {
key: path.clone(),
last_modified: object.last_modified, // ignore
size: 3
}
);
// Wait for a couple of seconds, and then update the file to check the last
// modified timestamp.
tokio::time::sleep(std::time::Duration::from_secs(2)).await;
let data = bytes::Bytes::from_static("bar".as_bytes());
let (stream, len) = wrap_stream(data);
ctx.client.upload(stream, len, &path, None, &cancel).await?;
let new = ctx.client.head_object(&path, &cancel).await?;
assert!(
!new.last_modified
.duration_since(object.last_modified)?
.is_zero(),
"last_modified did not advance"
);
Ok(())
}

View File

@@ -12,8 +12,8 @@ use anyhow::Context;
use camino::Utf8Path;
use futures_util::StreamExt;
use remote_storage::{
DownloadError, DownloadOpts, GenericRemoteStorage, ListingMode, RemotePath,
RemoteStorageConfig, RemoteStorageKind, S3Config,
DownloadError, GenericRemoteStorage, ListingMode, RemotePath, RemoteStorageConfig,
RemoteStorageKind, S3Config,
};
use test_context::test_context;
use test_context::AsyncTestContext;
@@ -121,8 +121,7 @@ async fn s3_time_travel_recovery_works(ctx: &mut MaybeEnabledStorage) -> anyhow:
// A little check to ensure that our clock is not too far off from the S3 clock
{
let opts = DownloadOpts::default();
let dl = retry(|| ctx.client.download(&path2, &opts, &cancel)).await?;
let dl = retry(|| ctx.client.download(&path2, &cancel)).await?;
let last_modified = dl.last_modified;
let half_wt = WAIT_TIME.mul_f32(0.5);
let t0_hwt = t0 + half_wt;
@@ -160,12 +159,7 @@ async fn s3_time_travel_recovery_works(ctx: &mut MaybeEnabledStorage) -> anyhow:
let t2_files_recovered = list_files(&ctx.client, &cancel).await?;
println!("after recovery to t2: {t2_files_recovered:?}");
assert_eq!(t2_files, t2_files_recovered);
let path2_recovered_t2 = download_to_vec(
ctx.client
.download(&path2, &DownloadOpts::default(), &cancel)
.await?,
)
.await?;
let path2_recovered_t2 = download_to_vec(ctx.client.download(&path2, &cancel).await?).await?;
assert_eq!(path2_recovered_t2, new_data.as_bytes());
// after recovery to t1: path1 is back, path2 has the old content
@@ -176,12 +170,7 @@ async fn s3_time_travel_recovery_works(ctx: &mut MaybeEnabledStorage) -> anyhow:
let t1_files_recovered = list_files(&ctx.client, &cancel).await?;
println!("after recovery to t1: {t1_files_recovered:?}");
assert_eq!(t1_files, t1_files_recovered);
let path2_recovered_t1 = download_to_vec(
ctx.client
.download(&path2, &DownloadOpts::default(), &cancel)
.await?,
)
.await?;
let path2_recovered_t1 = download_to_vec(ctx.client.download(&path2, &cancel).await?).await?;
assert_eq!(path2_recovered_t1, old_data.as_bytes());
// after recovery to t0: everything is gone except for path1
@@ -427,7 +416,7 @@ async fn download_is_timeouted(ctx: &mut MaybeEnabledStorage) {
let started_at = std::time::Instant::now();
let mut stream = ctx
.client
.download(&path, &DownloadOpts::default(), &cancel)
.download(&path, &cancel)
.await
.expect("download succeeds")
.download_stream;
@@ -502,7 +491,7 @@ async fn download_is_cancelled(ctx: &mut MaybeEnabledStorage) {
{
let stream = ctx
.client
.download(&path, &DownloadOpts::default(), &cancel)
.download(&path, &cancel)
.await
.expect("download succeeds")
.download_stream;

View File

@@ -5,15 +5,13 @@ edition.workspace = true
license.workspace = true
[dependencies]
hyper0.workspace = true
opentelemetry = { workspace = true, features = ["trace"] }
opentelemetry_sdk = { workspace = true, features = ["rt-tokio"] }
opentelemetry-otlp = { workspace = true, default-features = false, features = ["http-proto", "trace", "http", "reqwest-client"] }
hyper.workspace = true
opentelemetry = { workspace = true, features=["rt-tokio"] }
opentelemetry-otlp = { workspace = true, default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
opentelemetry-semantic-conventions.workspace = true
tokio = { workspace = true, features = ["rt", "rt-multi-thread"] }
tracing.workspace = true
tracing-opentelemetry.workspace = true
tracing-subscriber.workspace = true
[dev-dependencies]
tracing-subscriber.workspace = true # For examples in docs

View File

@@ -1,7 +1,7 @@
//! Tracing wrapper for Hyper HTTP server
use hyper0::HeaderMap;
use hyper0::{Body, Request, Response};
use hyper::HeaderMap;
use hyper::{Body, Request, Response};
use std::future::Future;
use tracing::Instrument;
use tracing_opentelemetry::OpenTelemetrySpanExt;

View File

@@ -10,6 +10,7 @@
//!
//! ```rust,no_run
//! use tracing_subscriber::prelude::*;
//! use tracing_opentelemetry::OpenTelemetryLayer;
//!
//! #[tokio::main]
//! async fn main() {
@@ -21,7 +22,7 @@
//! .with_writer(std::io::stderr);
//!
//! // Initialize OpenTelemetry. Exports tracing spans as OpenTelemetry traces
//! let otlp_layer = tracing_utils::init_tracing("my_application").await;
//! let otlp_layer = tracing_utils::init_tracing("my_application").await.map(OpenTelemetryLayer::new);
//!
//! // Put it all together
//! tracing_subscriber::registry()
@@ -34,14 +35,14 @@
#![deny(unsafe_code)]
#![deny(clippy::undocumented_unsafe_blocks)]
pub mod http;
use opentelemetry::trace::TracerProvider;
use opentelemetry::sdk::Resource;
use opentelemetry::KeyValue;
use opentelemetry_sdk::Resource;
use tracing::Subscriber;
use tracing_subscriber::registry::LookupSpan;
use tracing_subscriber::Layer;
use opentelemetry_otlp::WithExportConfig;
use opentelemetry_otlp::{OTEL_EXPORTER_OTLP_ENDPOINT, OTEL_EXPORTER_OTLP_TRACES_ENDPOINT};
pub use tracing_opentelemetry::OpenTelemetryLayer;
pub mod http;
/// Set up OpenTelemetry exporter, using configuration from environment variables.
///
@@ -70,10 +71,7 @@ use tracing_subscriber::Layer;
///
/// This doesn't block, but is marked as 'async' to hint that this must be called in
/// asynchronous execution context.
pub async fn init_tracing<S>(service_name: &str) -> Option<impl Layer<S>>
where
S: Subscriber + for<'span> LookupSpan<'span>,
{
pub async fn init_tracing(service_name: &str) -> Option<opentelemetry::sdk::trace::Tracer> {
if std::env::var("OTEL_SDK_DISABLED") == Ok("true".to_string()) {
return None;
};
@@ -82,10 +80,9 @@ where
/// Like `init_tracing`, but creates a separate tokio Runtime for the tracing
/// tasks.
pub fn init_tracing_without_runtime<S>(service_name: &str) -> Option<impl Layer<S>>
where
S: Subscriber + for<'span> LookupSpan<'span>,
{
pub fn init_tracing_without_runtime(
service_name: &str,
) -> Option<opentelemetry::sdk::trace::Tracer> {
if std::env::var("OTEL_SDK_DISABLED") == Ok("true".to_string()) {
return None;
};
@@ -116,36 +113,54 @@ where
Some(init_tracing_internal(service_name.to_string()))
}
fn init_tracing_internal<S>(service_name: String) -> impl Layer<S>
where
S: Subscriber + for<'span> LookupSpan<'span>,
{
// Sets up exporter from the OTEL_EXPORTER_* environment variables.
let exporter = opentelemetry_otlp::new_exporter().http();
fn init_tracing_internal(service_name: String) -> opentelemetry::sdk::trace::Tracer {
// Set up exporter from the OTEL_EXPORTER_* environment variables
let mut exporter = opentelemetry_otlp::new_exporter().http().with_env();
// TODO: opentelemetry::global::set_error_handler() with custom handler that
// bypasses default tracing layers, but logs regular looking log
// messages.
// XXX opentelemetry-otlp v0.18.0 has a bug in how it uses the
// OTEL_EXPORTER_OTLP_ENDPOINT env variable. According to the
// OpenTelemetry spec at
// <https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/protocol/exporter.md#endpoint-urls-for-otlphttp>,
// the full exporter URL is formed by appending "/v1/traces" to the value
// of OTEL_EXPORTER_OTLP_ENDPOINT. However, opentelemetry-otlp only does
// that with the grpc-tonic exporter. Other exporters, like the HTTP
// exporter, use the URL from OTEL_EXPORTER_OTLP_ENDPOINT as is, without
// appending "/v1/traces".
//
// See https://github.com/open-telemetry/opentelemetry-rust/pull/950
//
// Work around that by checking OTEL_EXPORTER_OTLP_ENDPOINT, and setting
// the endpoint url with the "/v1/traces" path ourselves. If the bug is
// fixed in a later version, we can remove this code. But if we don't
// remember to remove this, it won't do any harm either, as the crate will
// just ignore the OTEL_EXPORTER_OTLP_ENDPOINT setting when the endpoint
// is set directly with `with_endpoint`.
if std::env::var(OTEL_EXPORTER_OTLP_TRACES_ENDPOINT).is_err() {
if let Ok(mut endpoint) = std::env::var(OTEL_EXPORTER_OTLP_ENDPOINT) {
if !endpoint.ends_with('/') {
endpoint.push('/');
}
endpoint.push_str("v1/traces");
exporter = exporter.with_endpoint(endpoint);
}
}
// Propagate trace information in the standard W3C TraceContext format.
opentelemetry::global::set_text_map_propagator(
opentelemetry_sdk::propagation::TraceContextPropagator::new(),
opentelemetry::sdk::propagation::TraceContextPropagator::new(),
);
let tracer = opentelemetry_otlp::new_pipeline()
opentelemetry_otlp::new_pipeline()
.tracing()
.with_exporter(exporter)
.with_trace_config(opentelemetry_sdk::trace::Config::default().with_resource(
Resource::new(vec![KeyValue::new(
.with_trace_config(
opentelemetry::sdk::trace::config().with_resource(Resource::new(vec![KeyValue::new(
opentelemetry_semantic_conventions::resource::SERVICE_NAME,
service_name,
)]),
))
.install_batch(opentelemetry_sdk::runtime::Tokio)
)])),
)
.install_batch(opentelemetry::runtime::Tokio)
.expect("could not initialize opentelemetry exporter")
.tracer("global");
tracing_opentelemetry::layer().with_tracer(tracer)
}
// Shutdown trace pipeline gracefully, so that it has a chance to send any

View File

@@ -22,7 +22,7 @@ chrono.workspace = true
git-version.workspace = true
hex = { workspace = true, features = ["serde"] }
humantime.workspace = true
hyper0 = { workspace = true, features = ["full"] }
hyper = { workspace = true, features = ["full"] }
fail.workspace = true
futures = { workspace = true}
jsonwebtoken.workspace = true

View File

@@ -2,8 +2,6 @@
//! between other crates in this repository.
#![deny(clippy::undocumented_unsafe_blocks)]
extern crate hyper0 as hyper;
pub mod backoff;
/// `Lsn` type implements common tasks on Log Sequence Numbers

View File

@@ -7,13 +7,11 @@ use axum::{
extract::{ws::WebSocket, State, WebSocketUpgrade},
response::Response,
};
use axum::{routing::get, Router};
use axum::{routing::get, Router, Server};
use clap::Parser;
use futures::Future;
use std::net::SocketAddr;
use std::{fmt::Debug, time::Duration};
use sysinfo::{RefreshKind, System, SystemExt};
use tokio::net::TcpListener;
use tokio::{sync::broadcast, task::JoinHandle};
use tokio_util::sync::CancellationToken;
use tracing::{error, info};
@@ -134,14 +132,14 @@ pub async fn start(args: &'static Args, token: CancellationToken) -> anyhow::Res
args,
});
let addr_str = args.addr();
let addr: SocketAddr = addr_str.parse().expect("parsing address should not fail");
let listener = TcpListener::bind(&addr)
.await
let addr = args.addr();
let bound = Server::try_bind(&addr.parse().expect("parsing address should not fail"))
.with_context(|| format!("failed to bind to {addr}"))?;
info!(addr_str, "server bound");
axum::serve(listener, app.into_make_service())
info!(addr, "server bound");
bound
.serve(app.into_make_service())
.await
.context("server exited")?;

View File

@@ -30,7 +30,7 @@ futures.workspace = true
hex.workspace = true
humantime.workspace = true
humantime-serde.workspace = true
hyper0.workspace = true
hyper.workspace = true
itertools.workspace = true
md5.workspace = true
nix.workspace = true

View File

@@ -164,12 +164,10 @@ fn criterion_benchmark(c: &mut Criterion) {
let conf: &'static PageServerConf = Box::leak(Box::new(
pageserver::config::PageServerConf::dummy_conf(temp_dir.path().to_path_buf()),
));
virtual_file::init(
16384,
virtual_file::io_engine_for_bench(),
pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
);
page_cache::init(conf.page_cache_size);
let align = pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
virtual_file::init(16384, virtual_file::io_engine_for_bench(), align);
page_cache::init(conf.page_cache_size, align);
{
let mut group = c.benchmark_group("ingest-small-values");

View File

@@ -550,6 +550,19 @@ impl Client {
.map_err(Error::ReceiveBody)
}
/// Configs io mode at runtime.
pub async fn put_io_mode(
&self,
mode: &pageserver_api::models::virtual_file::IoMode,
) -> Result<()> {
let uri = format!("{}/v1/io_mode", self.mgmt_api_endpoint);
self.request(Method::PUT, uri, mode)
.await?
.json()
.await
.map_err(Error::ReceiveBody)
}
pub async fn get_utilization(&self) -> Result<PageserverUtilization> {
let uri = format!("{}/v1/utilization", self.mgmt_api_endpoint);
self.get(uri)

View File

@@ -151,13 +151,10 @@ pub(crate) async fn main(cmd: &AnalyzeLayerMapCmd) -> Result<()> {
let max_holes = cmd.max_holes.unwrap_or(DEFAULT_MAX_HOLES);
let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
let align = pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
// Initialize virtual_file (file desriptor cache) and page cache which are needed to access layer persistent B-Tree.
pageserver::virtual_file::init(
10,
virtual_file::api::IoEngineKind::StdFs,
pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
);
pageserver::page_cache::init(100);
pageserver::virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, align);
pageserver::page_cache::init(100, align);
let mut total_delta_layers = 0usize;
let mut total_image_layers = 0usize;

View File

@@ -59,8 +59,9 @@ pub(crate) enum LayerCmd {
async fn read_delta_file(path: impl AsRef<Path>, ctx: &RequestContext) -> Result<()> {
let path = Utf8Path::from_path(path.as_ref()).expect("non-Unicode path");
virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, 1);
page_cache::init(100);
let align = pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, align);
page_cache::init(100, align);
let file = VirtualFile::open(path, ctx).await?;
let file_id = page_cache::next_file_id();
let block_reader = FileBlockReader::new(&file, file_id);
@@ -190,12 +191,10 @@ pub(crate) async fn main(cmd: &LayerCmd) -> Result<()> {
new_tenant_id,
new_timeline_id,
} => {
pageserver::virtual_file::init(
10,
virtual_file::api::IoEngineKind::StdFs,
pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT,
);
pageserver::page_cache::init(100);
let align = pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
pageserver::virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, align);
pageserver::page_cache::init(100, align);
let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);

View File

@@ -205,12 +205,9 @@ fn read_pg_control_file(control_file_path: &Utf8Path) -> anyhow::Result<()> {
async fn print_layerfile(path: &Utf8Path) -> anyhow::Result<()> {
// Basic initialization of things that don't change after startup
virtual_file::init(
10,
virtual_file::api::IoEngineKind::StdFs,
DEFAULT_IO_BUFFER_ALIGNMENT,
);
page_cache::init(100);
let align = DEFAULT_IO_BUFFER_ALIGNMENT;
virtual_file::init(10, virtual_file::api::IoEngineKind::StdFs, align);
page_cache::init(100, align);
let ctx = RequestContext::new(TaskKind::DebugTool, DownloadBehavior::Error);
dump_layerfile_from_path(path, true, &ctx).await
}

View File

@@ -63,6 +63,10 @@ pub(crate) struct Args {
#[clap(long)]
set_io_alignment: Option<usize>,
/// Before starting the benchmark, live-reconfigure the pageserver to use specified io mode (buffered vs. direct).
#[clap(long)]
set_io_mode: Option<pageserver_api::models::virtual_file::IoMode>,
targets: Option<Vec<TenantTimelineId>>,
}
@@ -133,6 +137,10 @@ async fn main_impl(
mgmt_api_client.put_io_alignment(align).await?;
}
if let Some(mode) = &args.set_io_mode {
mgmt_api_client.put_io_mode(mode).await?;
}
// discover targets
let timelines: Vec<TenantTimelineId> = crate::util::cli::targets::discover(
&mgmt_api_client,

View File

@@ -125,7 +125,7 @@ fn main() -> anyhow::Result<()> {
// after setting up logging, log the effective IO engine choice and read path implementations
info!(?conf.virtual_file_io_engine, "starting with virtual_file IO engine");
info!(?conf.virtual_file_direct_io, "starting with virtual_file Direct IO settings");
info!(?conf.virtual_file_io_mode, "starting with virtual_file Direct IO settings");
info!(?conf.io_buffer_alignment, "starting with setting for IO buffer alignment");
// The tenants directory contains all the pageserver local disk state.
@@ -173,7 +173,7 @@ fn main() -> anyhow::Result<()> {
conf.virtual_file_io_engine,
conf.io_buffer_alignment,
);
page_cache::init(conf.page_cache_size);
page_cache::init(conf.page_cache_size, conf.io_buffer_alignment);
start_pageserver(launch_ts, conf).context("Failed to start pageserver")?;
@@ -575,7 +575,7 @@ fn start_pageserver(
.build()
.map_err(|err| anyhow!(err))?;
let service = utils::http::RouterService::new(router).unwrap();
let server = hyper0::Server::from_tcp(http_listener)?
let server = hyper::Server::from_tcp(http_listener)?
.serve(service)
.with_graceful_shutdown({
let cancel = cancel.clone();

View File

@@ -174,7 +174,7 @@ pub struct PageServerConf {
pub l0_flush: crate::l0_flush::L0FlushConfig,
/// Direct IO settings
pub virtual_file_direct_io: virtual_file::DirectIoMode,
pub virtual_file_io_mode: virtual_file::IoMode,
pub io_buffer_alignment: usize,
}
@@ -325,7 +325,7 @@ impl PageServerConf {
image_compression,
ephemeral_bytes_per_memory_kb,
l0_flush,
virtual_file_direct_io,
virtual_file_io_mode,
concurrent_tenant_warmup,
concurrent_tenant_size_logical_size_queries,
virtual_file_io_engine,
@@ -368,7 +368,6 @@ impl PageServerConf {
max_vectored_read_bytes,
image_compression,
ephemeral_bytes_per_memory_kb,
virtual_file_direct_io,
io_buffer_alignment,
// ------------------------------------------------------------
@@ -408,6 +407,7 @@ impl PageServerConf {
l0_flush: l0_flush
.map(crate::l0_flush::L0FlushConfig::from)
.unwrap_or_default(),
virtual_file_io_mode: virtual_file_io_mode.unwrap_or(virtual_file::IoMode::preferred()),
};
// ------------------------------------------------------------

View File

@@ -17,6 +17,7 @@ use hyper::header;
use hyper::StatusCode;
use hyper::{Body, Request, Response, Uri};
use metrics::launch_timestamp::LaunchTimestamp;
use pageserver_api::models::virtual_file::IoMode;
use pageserver_api::models::AuxFilePolicy;
use pageserver_api::models::DownloadRemoteLayersTaskSpawnRequest;
use pageserver_api::models::IngestAuxFilesRequest;
@@ -56,7 +57,6 @@ use utils::http::endpoint::request_span;
use utils::http::request::must_parse_query_param;
use utils::http::request::{get_request_param, must_get_query_param, parse_query_param};
use crate::config::PageServerConf;
use crate::context::{DownloadBehavior, RequestContext};
use crate::deletion_queue::DeletionQueueClient;
use crate::pgdatadir_mapping::LsnForTimestamp;
@@ -81,6 +81,7 @@ use crate::tenant::timeline::CompactionError;
use crate::tenant::timeline::Timeline;
use crate::tenant::GetTimelineError;
use crate::tenant::{LogicalSizeCalculationCause, PageReconstructError};
use crate::{config::PageServerConf, tenant::mgr};
use crate::{disk_usage_eviction_task, tenant};
use pageserver_api::models::{
StatusResponse, TenantConfigRequest, TenantInfo, TimelineCreateRequest, TimelineGcRequest,
@@ -1719,13 +1720,8 @@ async fn timeline_gc_handler(
let gc_req: TimelineGcRequest = json_request(&mut request).await?;
let state = get_state(&request);
let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
let gc_result = state
.tenant_manager
.immediate_gc(tenant_shard_id, timeline_id, gc_req, cancel, &ctx)
.await?;
let gc_result = mgr::immediate_gc(tenant_shard_id, timeline_id, gc_req, cancel, &ctx).await?;
json_response(StatusCode::OK, gc_result)
}
@@ -1742,10 +1738,6 @@ async fn timeline_compact_handler(
let state = get_state(&request);
let mut flags = EnumSet::empty();
if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
flags |= CompactFlags::ForceL0Compaction;
}
if Some(true) == parse_query_param::<_, bool>(&request, "force_repartition")? {
flags |= CompactFlags::ForceRepartition;
}
@@ -1792,9 +1784,6 @@ async fn timeline_checkpoint_handler(
let state = get_state(&request);
let mut flags = EnumSet::empty();
if Some(true) == parse_query_param::<_, bool>(&request, "force_l0_compaction")? {
flags |= CompactFlags::ForceL0Compaction;
}
if Some(true) == parse_query_param::<_, bool>(&request, "force_repartition")? {
flags |= CompactFlags::ForceRepartition;
}
@@ -2393,6 +2382,16 @@ async fn put_io_alignment_handler(
json_response(StatusCode::OK, ())
}
async fn put_io_mode_handler(
mut r: Request<Body>,
_cancel: CancellationToken,
) -> Result<Response<Body>, ApiError> {
check_permission(&r, None)?;
let mode: IoMode = json_request(&mut r).await?;
crate::virtual_file::set_io_mode(mode);
json_response(StatusCode::OK, ())
}
/// Polled by control plane.
///
/// See [`crate::utilization`].
@@ -3083,6 +3082,7 @@ pub fn make_router(
.put("/v1/io_alignment", |r| {
api_handler(r, put_io_alignment_handler)
})
.put("/v1/io_mode", |r| api_handler(r, put_io_mode_handler))
.put(
"/v1/tenant/:tenant_shard_id/timeline/:timeline_id/force_aux_policy_switch",
|r| api_handler(r, force_aux_policy_switch_handler),

View File

@@ -13,8 +13,6 @@ pub mod http;
pub mod import_datadir;
pub mod l0_flush;
extern crate hyper0 as hyper;
use futures::{stream::FuturesUnordered, StreamExt};
pub use pageserver_api::keyspace;
use tokio_util::sync::CancellationToken;

View File

@@ -82,6 +82,7 @@ use once_cell::sync::OnceCell;
use crate::{
context::RequestContext,
metrics::{page_cache_eviction_metrics, PageCacheSizeMetrics},
virtual_file::{self, dio::IoBufferMut},
};
static PAGE_CACHE: OnceCell<PageCache> = OnceCell::new();
@@ -90,8 +91,8 @@ const TEST_PAGE_CACHE_SIZE: usize = 50;
///
/// Initialize the page cache. This must be called once at page server startup.
///
pub fn init(size: usize) {
if PAGE_CACHE.set(PageCache::new(size)).is_err() {
pub fn init(size: usize, align: usize) {
if PAGE_CACHE.set(PageCache::new(size, align)).is_err() {
panic!("page cache already initialized");
}
}
@@ -106,7 +107,12 @@ pub fn get() -> &'static PageCache {
// page cache is usable in unit tests.
//
if cfg!(test) {
PAGE_CACHE.get_or_init(|| PageCache::new(TEST_PAGE_CACHE_SIZE))
PAGE_CACHE.get_or_init(|| {
PageCache::new(
TEST_PAGE_CACHE_SIZE,
virtual_file::get_io_buffer_alignment(),
)
})
} else {
PAGE_CACHE.get().expect("page cache not initialized")
}
@@ -637,13 +643,11 @@ impl PageCache {
/// Initialize a new page cache
///
/// This should be called only once at page server startup.
fn new(num_pages: usize) -> Self {
fn new(num_pages: usize, align: usize) -> Self {
assert!(num_pages > 0, "page cache size must be > 0");
// We could use Vec::leak here, but that potentially also leaks
// uninitialized reserved capacity. With into_boxed_slice and Box::leak
// this is avoided.
let page_buffer = Box::leak(vec![0u8; num_pages * PAGE_SZ].into_boxed_slice());
let page_buffer =
IoBufferMut::with_capacity_aligned_zeroed(num_pages * PAGE_SZ, align).leak();
let size_metrics = &crate::metrics::PAGE_CACHE_SIZE;
size_metrics.max_bytes.set_page_sz(num_pages);

View File

@@ -97,7 +97,6 @@ use crate::tenant::remote_timeline_client::MaybeDeletedIndexPart;
use crate::tenant::remote_timeline_client::INITDB_PATH;
use crate::tenant::storage_layer::DeltaLayer;
use crate::tenant::storage_layer::ImageLayer;
use crate::walingest::WalLagCooldown;
use crate::walredo;
use crate::InitializationOrder;
use std::collections::hash_map::Entry;
@@ -320,9 +319,6 @@ pub struct Tenant {
/// background warmup.
pub(crate) activate_now_sem: tokio::sync::Semaphore,
/// Time it took for the tenant to activate. Zero if not active yet.
attach_wal_lag_cooldown: Arc<std::sync::OnceLock<WalLagCooldown>>,
// Cancellation token fires when we have entered shutdown(). This is a parent of
// Timelines' cancellation token.
pub(crate) cancel: CancellationToken,
@@ -1004,15 +1000,11 @@ impl Tenant {
// Remote preload is complete.
drop(remote_load_completion);
// We will time the duration of the attach phase unless this is a creation (attach will do no work)
let attach_start = std::time::Instant::now();
let attached = {
let _attach_timer = Some(TENANT.attach.start_timer());
tenant_clone.attach(preload, &ctx).await
};
let attach_duration = attach_start.elapsed();
_ = tenant_clone.attach_wal_lag_cooldown.set(WalLagCooldown::new(attach_start, attach_duration));
match attached {
Ok(()) => {
@@ -2762,7 +2754,6 @@ impl Tenant {
pg_version,
state,
last_aux_file_policy,
self.attach_wal_lag_cooldown.clone(),
self.cancel.child_token(),
);
@@ -2869,7 +2860,6 @@ impl Tenant {
Some(Duration::from_secs(3600 * 24)),
)),
activate_now_sem: tokio::sync::Semaphore::new(0),
attach_wal_lag_cooldown: Arc::new(std::sync::OnceLock::new()),
cancel: CancellationToken::default(),
gate: Gate::default(),
timeline_get_throttle: Arc::new(throttle::Throttle::new(

View File

@@ -8,6 +8,7 @@
//! We cannot use global or default config instead, because wrong settings
//! may lead to a data loss.
//!
use anyhow::bail;
pub(crate) use pageserver_api::config::TenantConfigToml as TenantConf;
use pageserver_api::models::AuxFilePolicy;
use pageserver_api::models::CompactionAlgorithmSettings;
@@ -440,6 +441,29 @@ impl TryFrom<&'_ models::TenantConfig> for TenantConfOpt {
}
}
impl TryFrom<toml_edit::Item> for TenantConfOpt {
type Error = anyhow::Error;
fn try_from(item: toml_edit::Item) -> Result<Self, Self::Error> {
match item {
toml_edit::Item::Value(value) => {
let d = value.into_deserializer();
return serde_path_to_error::deserialize(d)
.map_err(|e| anyhow::anyhow!("{}: {}", e.path(), e.inner().message()));
}
toml_edit::Item::Table(table) => {
let deserializer =
toml_edit::de::Deserializer::from(toml_edit::DocumentMut::from(table));
return serde_path_to_error::deserialize(deserializer)
.map_err(|e| anyhow::anyhow!("{}: {}", e.path(), e.inner().message()));
}
_ => {
bail!("expected non-inline table but found {item}")
}
}
}
}
/// This is a conversion from our internal tenant config object to the one used
/// in external APIs.
impl From<TenantConfOpt> for models::TenantConfig {

View File

@@ -84,7 +84,7 @@ impl Drop for EphemeralFile {
fn drop(&mut self) {
// unlink the file
// we are clear to do this, because we have entered a gate
let path = &self.buffered_writer.as_inner().as_inner().path;
let path = self.buffered_writer.as_inner().as_inner().path();
let res = std::fs::remove_file(path);
if let Err(e) = res {
if e.kind() != std::io::ErrorKind::NotFound {
@@ -356,7 +356,7 @@ mod tests {
}
let file_contents =
std::fs::read(&file.buffered_writer.as_inner().as_inner().path).unwrap();
std::fs::read(file.buffered_writer.as_inner().as_inner().path()).unwrap();
assert_eq!(file_contents, &content[0..cap]);
let buffer_contents = file.buffered_writer.inspect_buffer();
@@ -392,7 +392,7 @@ mod tests {
.buffered_writer
.as_inner()
.as_inner()
.path
.path()
.metadata()
.unwrap();
assert_eq!(

View File

@@ -2197,82 +2197,6 @@ impl TenantManager {
Ok((wanted_bytes, shard_count as u32))
}
#[instrument(skip_all, fields(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id))]
pub(crate) async fn immediate_gc(
&self,
tenant_shard_id: TenantShardId,
timeline_id: TimelineId,
gc_req: TimelineGcRequest,
cancel: CancellationToken,
ctx: &RequestContext,
) -> Result<GcResult, ApiError> {
let tenant = {
let guard = self.tenants.read().unwrap();
guard
.get(&tenant_shard_id)
.cloned()
.with_context(|| format!("tenant {tenant_shard_id}"))
.map_err(|e| ApiError::NotFound(e.into()))?
};
let gc_horizon = gc_req.gc_horizon.unwrap_or_else(|| tenant.get_gc_horizon());
// Use tenant's pitr setting
let pitr = tenant.get_pitr_interval();
tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
// Run in task_mgr to avoid race with tenant_detach operation
let ctx: RequestContext =
ctx.detached_child(TaskKind::GarbageCollector, DownloadBehavior::Download);
let _gate_guard = tenant.gate.enter().map_err(|_| ApiError::ShuttingDown)?;
fail::fail_point!("immediate_gc_task_pre");
#[allow(unused_mut)]
let mut result = tenant
.gc_iteration(Some(timeline_id), gc_horizon, pitr, &cancel, &ctx)
.await;
// FIXME: `gc_iteration` can return an error for multiple reasons; we should handle it
// better once the types support it.
#[cfg(feature = "testing")]
{
// we need to synchronize with drop completion for python tests without polling for
// log messages
if let Ok(result) = result.as_mut() {
let mut js = tokio::task::JoinSet::new();
for layer in std::mem::take(&mut result.doomed_layers) {
js.spawn(layer.wait_drop());
}
tracing::info!(
total = js.len(),
"starting to wait for the gc'd layers to be dropped"
);
while let Some(res) = js.join_next().await {
res.expect("wait_drop should not panic");
}
}
let timeline = tenant.get_timeline(timeline_id, false).ok();
let rtc = timeline.as_ref().map(|x| &x.remote_client);
if let Some(rtc) = rtc {
// layer drops schedule actions on remote timeline client to actually do the
// deletions; don't care about the shutdown error, just exit fast
drop(rtc.wait_completion().await);
}
}
result.map_err(|e| match e {
GcError::TenantCancelled | GcError::TimelineCancelled => ApiError::ShuttingDown,
GcError::TimelineNotFound => {
ApiError::NotFound(anyhow::anyhow!("Timeline not found").into())
}
other => ApiError::InternalServerError(anyhow::anyhow!(other)),
})
}
}
#[derive(Debug, thiserror::Error)]
@@ -2417,7 +2341,7 @@ enum TenantSlotDropError {
/// Errors that can happen any time we are walking the tenant map to try and acquire
/// the TenantSlot for a particular tenant.
#[derive(Debug, thiserror::Error)]
pub(crate) enum TenantMapError {
pub enum TenantMapError {
// Tried to read while initializing
#[error("tenant map is still initializing")]
StillInitializing,
@@ -2447,7 +2371,7 @@ pub(crate) enum TenantMapError {
/// The `old_value` may be dropped before the SlotGuard is dropped, by calling
/// `drop_old_value`. It is an error to call this without shutting down
/// the conents of `old_value`.
pub(crate) struct SlotGuard {
pub struct SlotGuard {
tenant_shard_id: TenantShardId,
old_value: Option<TenantSlot>,
upserted: bool,
@@ -2840,6 +2764,81 @@ use {
utils::http::error::ApiError,
};
#[instrument(skip_all, fields(tenant_id=%tenant_shard_id.tenant_id, shard_id=%tenant_shard_id.shard_slug(), %timeline_id))]
pub(crate) async fn immediate_gc(
tenant_shard_id: TenantShardId,
timeline_id: TimelineId,
gc_req: TimelineGcRequest,
cancel: CancellationToken,
ctx: &RequestContext,
) -> Result<GcResult, ApiError> {
let tenant = {
let guard = TENANTS.read().unwrap();
guard
.get(&tenant_shard_id)
.cloned()
.with_context(|| format!("tenant {tenant_shard_id}"))
.map_err(|e| ApiError::NotFound(e.into()))?
};
let gc_horizon = gc_req.gc_horizon.unwrap_or_else(|| tenant.get_gc_horizon());
// Use tenant's pitr setting
let pitr = tenant.get_pitr_interval();
tenant.wait_to_become_active(ACTIVE_TENANT_TIMEOUT).await?;
// Run in task_mgr to avoid race with tenant_detach operation
let ctx: RequestContext =
ctx.detached_child(TaskKind::GarbageCollector, DownloadBehavior::Download);
let _gate_guard = tenant.gate.enter().map_err(|_| ApiError::ShuttingDown)?;
fail::fail_point!("immediate_gc_task_pre");
#[allow(unused_mut)]
let mut result = tenant
.gc_iteration(Some(timeline_id), gc_horizon, pitr, &cancel, &ctx)
.await;
// FIXME: `gc_iteration` can return an error for multiple reasons; we should handle it
// better once the types support it.
#[cfg(feature = "testing")]
{
// we need to synchronize with drop completion for python tests without polling for
// log messages
if let Ok(result) = result.as_mut() {
let mut js = tokio::task::JoinSet::new();
for layer in std::mem::take(&mut result.doomed_layers) {
js.spawn(layer.wait_drop());
}
tracing::info!(
total = js.len(),
"starting to wait for the gc'd layers to be dropped"
);
while let Some(res) = js.join_next().await {
res.expect("wait_drop should not panic");
}
}
let timeline = tenant.get_timeline(timeline_id, false).ok();
let rtc = timeline.as_ref().map(|x| &x.remote_client);
if let Some(rtc) = rtc {
// layer drops schedule actions on remote timeline client to actually do the
// deletions; don't care about the shutdown error, just exit fast
drop(rtc.wait_completion().await);
}
}
result.map_err(|e| match e {
GcError::TenantCancelled | GcError::TimelineCancelled => ApiError::ShuttingDown,
GcError::TimelineNotFound => {
ApiError::NotFound(anyhow::anyhow!("Timeline not found").into())
}
other => ApiError::InternalServerError(anyhow::anyhow!(other)),
})
}
#[cfg(test)]
mod tests {
use std::collections::BTreeMap;

View File

@@ -27,7 +27,7 @@ use crate::tenant::Generation;
use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
use crate::virtual_file::{on_fatal_io_error, MaybeFatalIo, VirtualFile};
use crate::TEMP_FILE_SUFFIX;
use remote_storage::{DownloadError, DownloadOpts, GenericRemoteStorage, ListingMode, RemotePath};
use remote_storage::{DownloadError, GenericRemoteStorage, ListingMode, RemotePath};
use utils::crashsafe::path_with_suffix_extension;
use utils::id::{TenantId, TimelineId};
use utils::pausable_failpoint;
@@ -153,9 +153,7 @@ async fn download_object<'a>(
.with_context(|| format!("create a destination file for layer '{dst_path}'"))
.map_err(DownloadError::Other)?;
let download = storage
.download(src_path, &DownloadOpts::default(), cancel)
.await?;
let download = storage.download(src_path, cancel).await?;
pausable_failpoint!("before-downloading-layer-stream-pausable");
@@ -206,9 +204,7 @@ async fn download_object<'a>(
.with_context(|| format!("create a destination file for layer '{dst_path}'"))
.map_err(DownloadError::Other)?;
let mut download = storage
.download(src_path, &DownloadOpts::default(), cancel)
.await?;
let mut download = storage.download(src_path, cancel).await?;
pausable_failpoint!("before-downloading-layer-stream-pausable");
@@ -348,9 +344,7 @@ async fn do_download_index_part(
let index_part_bytes = download_retry_forever(
|| async {
let download = storage
.download(&remote_path, &DownloadOpts::default(), cancel)
.await?;
let download = storage.download(&remote_path, cancel).await?;
let mut bytes = Vec::new();
@@ -532,15 +526,10 @@ pub(crate) async fn download_initdb_tar_zst(
.with_context(|| format!("tempfile creation {temp_path}"))
.map_err(DownloadError::Other)?;
let download = match storage
.download(&remote_path, &DownloadOpts::default(), cancel)
.await
{
let download = match storage.download(&remote_path, cancel).await {
Ok(dl) => dl,
Err(DownloadError::NotFound) => {
storage
.download(&remote_preserved_path, &DownloadOpts::default(), cancel)
.await?
storage.download(&remote_preserved_path, cancel).await?
}
Err(other) => Err(other)?,
};

View File

@@ -49,7 +49,7 @@ use futures::Future;
use metrics::UIntGauge;
use pageserver_api::models::SecondaryProgress;
use pageserver_api::shard::TenantShardId;
use remote_storage::{DownloadError, DownloadOpts, Etag, GenericRemoteStorage};
use remote_storage::{DownloadError, Etag, GenericRemoteStorage};
use tokio_util::sync::CancellationToken;
use tracing::{info_span, instrument, warn, Instrument};
@@ -944,34 +944,36 @@ impl<'a> TenantDownloader<'a> {
) -> Result<HeatMapDownload, UpdateError> {
debug_assert_current_span_has_tenant_id();
let tenant_shard_id = self.secondary_state.get_tenant_shard_id();
// TODO: pull up etag check into the request, to do a conditional GET rather than
// issuing a GET and then maybe ignoring the response body
// (https://github.com/neondatabase/neon/issues/6199)
tracing::debug!("Downloading heatmap for secondary tenant",);
let heatmap_path = remote_heatmap_path(tenant_shard_id);
let cancel = &self.secondary_state.cancel;
let opts = DownloadOpts {
etag: prev_etag.cloned(),
};
backoff::retry(
|| async {
let download = match self
let download = self
.remote_storage
.download(&heatmap_path, &opts, cancel)
.download(&heatmap_path, cancel)
.await
{
Ok(download) => download,
Err(DownloadError::Unmodified) => return Ok(HeatMapDownload::Unmodified),
Err(err) => return Err(err.into()),
};
.map_err(UpdateError::from)?;
let mut heatmap_bytes = Vec::new();
let mut body = tokio_util::io::StreamReader::new(download.download_stream);
let _size = tokio::io::copy_buf(&mut body, &mut heatmap_bytes).await?;
Ok(HeatMapDownload::Modified(HeatMapModified {
etag: download.etag,
last_modified: download.last_modified,
bytes: heatmap_bytes,
}))
SECONDARY_MODE.download_heatmap.inc();
if Some(&download.etag) == prev_etag {
Ok(HeatMapDownload::Unmodified)
} else {
let mut heatmap_bytes = Vec::new();
let mut body = tokio_util::io::StreamReader::new(download.download_stream);
let _size = tokio::io::copy_buf(&mut body, &mut heatmap_bytes).await?;
Ok(HeatMapDownload::Modified(HeatMapModified {
etag: download.etag,
last_modified: download.last_modified,
bytes: heatmap_bytes,
}))
}
},
|e| matches!(e, UpdateError::NoData | UpdateError::Cancelled),
FAILED_DOWNLOAD_WARN_THRESHOLD,
@@ -982,7 +984,6 @@ impl<'a> TenantDownloader<'a> {
.await
.ok_or_else(|| UpdateError::Cancelled)
.and_then(|x| x)
.inspect(|_| SECONDARY_MODE.download_heatmap.inc())
}
/// Download heatmap layers that are not present on local disk, or update their

View File

@@ -43,17 +43,16 @@ use crate::tenant::vectored_blob_io::{
VectoredReadPlanner,
};
use crate::tenant::PageReconstructError;
use crate::virtual_file::dio::IoBufferMut;
use crate::virtual_file::owned_buffers_io::io_buf_ext::{FullSlice, IoBufExt};
use crate::virtual_file::{self, MaybeFatalIo, VirtualFile};
use crate::{walrecord, TEMP_FILE_SUFFIX};
use crate::{DELTA_FILE_MAGIC, STORAGE_FORMAT_VERSION};
use anyhow::{anyhow, bail, ensure, Context, Result};
use bytes::BytesMut;
use camino::{Utf8Path, Utf8PathBuf};
use futures::StreamExt;
use itertools::Itertools;
use pageserver_api::config::MaxVectoredReadBytes;
use pageserver_api::key::DBDIR_KEY;
use pageserver_api::keyspace::KeySpace;
use pageserver_api::models::ImageCompressionAlgorithm;
use pageserver_api::shard::TenantShardId;
@@ -573,7 +572,7 @@ impl DeltaLayerWriterInner {
ensure!(
metadata.len() <= S3_UPLOAD_LIMIT,
"Created delta layer file at {} of size {} above limit {S3_UPLOAD_LIMIT}!",
file.path,
file.path(),
metadata.len()
);
@@ -791,7 +790,7 @@ impl DeltaLayerInner {
max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
ctx: &RequestContext,
) -> anyhow::Result<Self> {
let file = VirtualFile::open(path, ctx)
let file = VirtualFile::open_v2(path, ctx)
.await
.context("open layer file")?;
@@ -964,25 +963,14 @@ impl DeltaLayerInner {
.blobs_at
.as_slice()
.iter()
.filter_map(|(_, blob_meta)| {
if blob_meta.key.is_rel_dir_key() || blob_meta.key == DBDIR_KEY {
// The size of values for these keys is unbounded and can
// grow very large in pathological cases.
None
} else {
Some(format!("{}@{}", blob_meta.key, blob_meta.lsn))
}
})
.map(|(_, blob_meta)| format!("{}@{}", blob_meta.key, blob_meta.lsn))
.join(", ");
if !offenders.is_empty() {
tracing::warn!(
"Oversized vectored read ({} > {}) for keys {}",
largest_read_size,
read_size_soft_max,
offenders
);
}
tracing::warn!(
"Oversized vectored read ({} > {}) for keys {}",
largest_read_size,
read_size_soft_max,
offenders
);
}
largest_read_size
@@ -1003,7 +991,8 @@ impl DeltaLayerInner {
.0
.into();
let buf_size = Self::get_min_read_buffer_size(&reads, max_vectored_read_bytes);
let mut buf = Some(BytesMut::with_capacity(buf_size));
let align = virtual_file::get_io_buffer_alignment();
let mut buf = Some(IoBufferMut::with_capacity_aligned(buf_size, align));
// Note that reads are processed in reverse order (from highest key+lsn).
// This is the order that `ReconstructState` requires such that it can
@@ -1022,7 +1011,7 @@ impl DeltaLayerInner {
blob_meta.key,
PageReconstructError::Other(anyhow!(
"Failed to read blobs from virtual file {}: {}",
self.file.path,
self.file.path(),
kind
)),
);
@@ -1030,7 +1019,7 @@ impl DeltaLayerInner {
// We have "lost" the buffer since the lower level IO api
// doesn't return the buffer on error. Allocate a new one.
buf = Some(BytesMut::with_capacity(buf_size));
buf = Some(IoBufferMut::with_capacity_aligned(buf_size, align));
continue;
}
@@ -1048,7 +1037,7 @@ impl DeltaLayerInner {
meta.meta.key,
PageReconstructError::Other(anyhow!(e).context(format!(
"Failed to decompress blob from virtual file {}",
self.file.path,
self.file.path(),
))),
);
@@ -1066,7 +1055,7 @@ impl DeltaLayerInner {
meta.meta.key,
PageReconstructError::Other(anyhow!(e).context(format!(
"Failed to deserialize blob from virtual file {}",
self.file.path,
self.file.path(),
))),
);
@@ -1198,14 +1187,14 @@ impl DeltaLayerInner {
let mut prev: Option<(Key, Lsn, BlobRef)> = None;
let mut read_builder: Option<ChunkedVectoredReadBuilder> = None;
let align = virtual_file::get_io_buffer_alignment();
let max_read_size = self
.max_vectored_read_bytes
.map(|x| x.0.get())
.unwrap_or(8192);
let mut buffer = Some(BytesMut::with_capacity(max_read_size));
let align = virtual_file::get_io_buffer_alignment();
let mut buffer = Some(IoBufferMut::with_capacity_aligned(max_read_size, align));
// FIXME: buffering of DeltaLayerWriter
let mut per_blob_copy = Vec::new();
@@ -1564,12 +1553,12 @@ impl<'a> DeltaLayerIterator<'a> {
let vectored_blob_reader = VectoredBlobReader::new(&self.delta_layer.file);
let mut next_batch = std::collections::VecDeque::new();
let buf_size = plan.size();
let buf = BytesMut::with_capacity(buf_size);
let align = virtual_file::get_io_buffer_alignment();
let buf = IoBufferMut::with_capacity_aligned(buf_size, align);
let blobs_buf = vectored_blob_reader
.read_blobs(&plan, buf, self.ctx)
.await?;
let frozen_buf = blobs_buf.buf.freeze();
let view = BufView::new_bytes(frozen_buf);
let view = BufView::new_slice(&blobs_buf.buf);
for meta in blobs_buf.blobs.iter() {
let blob_read = meta.read(&view).await?;
let value = Value::des(&blob_read)?;
@@ -1944,7 +1933,9 @@ pub(crate) mod test {
&vectored_reads,
constants::MAX_VECTORED_READ_BYTES,
);
let mut buf = Some(BytesMut::with_capacity(buf_size));
let align = virtual_file::get_io_buffer_alignment();
let mut buf = Some(IoBufferMut::with_capacity_aligned(buf_size, align));
for read in vectored_reads {
let blobs_buf = vectored_blob_reader

View File

@@ -40,16 +40,16 @@ use crate::tenant::vectored_blob_io::{
VectoredReadPlanner,
};
use crate::tenant::PageReconstructError;
use crate::virtual_file::dio::IoBufferMut;
use crate::virtual_file::owned_buffers_io::io_buf_ext::IoBufExt;
use crate::virtual_file::{self, MaybeFatalIo, VirtualFile};
use crate::{IMAGE_FILE_MAGIC, STORAGE_FORMAT_VERSION, TEMP_FILE_SUFFIX};
use anyhow::{anyhow, bail, ensure, Context, Result};
use bytes::{Bytes, BytesMut};
use bytes::Bytes;
use camino::{Utf8Path, Utf8PathBuf};
use hex;
use itertools::Itertools;
use pageserver_api::config::MaxVectoredReadBytes;
use pageserver_api::key::DBDIR_KEY;
use pageserver_api::keyspace::KeySpace;
use pageserver_api::shard::{ShardIdentity, TenantShardId};
use rand::{distributions::Alphanumeric, Rng};
@@ -389,7 +389,7 @@ impl ImageLayerInner {
max_vectored_read_bytes: Option<MaxVectoredReadBytes>,
ctx: &RequestContext,
) -> anyhow::Result<Self> {
let file = VirtualFile::open(path, ctx)
let file = VirtualFile::open_v2(path, ctx)
.await
.context("open layer file")?;
let file_id = page_cache::next_file_id();
@@ -543,14 +543,15 @@ impl ImageLayerInner {
.await?;
let vectored_blob_reader = VectoredBlobReader::new(&self.file);
let align = virtual_file::get_io_buffer_alignment();
let mut key_count = 0;
for read in plan.into_iter() {
let buf_size = read.size();
let buf = BytesMut::with_capacity(buf_size);
let buf = IoBufferMut::with_capacity_aligned(buf_size, align);
let blobs_buf = vectored_blob_reader.read_blobs(&read, buf, ctx).await?;
let frozen_buf = blobs_buf.buf.freeze();
let view = BufView::new_bytes(frozen_buf);
let view = BufView::new_slice(&blobs_buf.buf);
for meta in blobs_buf.blobs.iter() {
let img_buf = meta.read(&view).await?;
@@ -588,34 +589,23 @@ impl ImageLayerInner {
.blobs_at
.as_slice()
.iter()
.filter_map(|(_, blob_meta)| {
if blob_meta.key.is_rel_dir_key() || blob_meta.key == DBDIR_KEY {
// The size of values for these keys is unbounded and can
// grow very large in pathological cases.
None
} else {
Some(format!("{}@{}", blob_meta.key, blob_meta.lsn))
}
})
.map(|(_, blob_meta)| format!("{}@{}", blob_meta.key, blob_meta.lsn))
.join(", ");
if !offenders.is_empty() {
tracing::warn!(
"Oversized vectored read ({} > {}) for keys {}",
buf_size,
max_vectored_read_bytes,
offenders
);
}
tracing::warn!(
"Oversized vectored read ({} > {}) for keys {}",
buf_size,
max_vectored_read_bytes,
offenders
);
}
let buf = BytesMut::with_capacity(buf_size);
let align = virtual_file::get_io_buffer_alignment();
let buf = IoBufferMut::with_capacity_aligned(buf_size, align);
let res = vectored_blob_reader.read_blobs(&read, buf, ctx).await;
match res {
Ok(blobs_buf) => {
let frozen_buf = blobs_buf.buf.freeze();
let view = BufView::new_bytes(frozen_buf);
let view = BufView::new_slice(&blobs_buf.buf);
for meta in blobs_buf.blobs.iter() {
let img_buf = meta.read(&view).await;
@@ -626,7 +616,7 @@ impl ImageLayerInner {
meta.meta.key,
PageReconstructError::Other(anyhow!(e).context(format!(
"Failed to decompress blob from virtual file {}",
self.file.path,
self.file.path(),
))),
);
@@ -647,7 +637,7 @@ impl ImageLayerInner {
blob_meta.key,
PageReconstructError::from(anyhow!(
"Failed to read blobs from virtual file {}: {}",
self.file.path,
self.file.path(),
kind
)),
);
@@ -1051,12 +1041,12 @@ impl<'a> ImageLayerIterator<'a> {
let vectored_blob_reader = VectoredBlobReader::new(&self.image_layer.file);
let mut next_batch = std::collections::VecDeque::new();
let buf_size = plan.size();
let buf = BytesMut::with_capacity(buf_size);
let align = virtual_file::get_io_buffer_alignment();
let buf = IoBufferMut::with_capacity_aligned(buf_size, align);
let blobs_buf = vectored_blob_reader
.read_blobs(&plan, buf, self.ctx)
.await?;
let frozen_buf = blobs_buf.buf.freeze();
let view = BufView::new_bytes(frozen_buf);
let view = BufView::new_slice(&blobs_buf.buf);
for meta in blobs_buf.blobs.iter() {
let img_buf = meta.read(&view).await?;
next_batch.push_back((

View File

@@ -442,13 +442,11 @@ impl Layer {
// Visibility was modified to Visible: maybe log about this
match ctx.task_kind() {
TaskKind::CalculateSyntheticSize
| TaskKind::OndemandLogicalSizeCalculation
| TaskKind::GarbageCollector
| TaskKind::MgmtRequest => {
// This situation is expected in code paths do binary searches of the LSN space to resolve
// an LSN to a timestamp, which happens during GC, during GC cutoff calculations in synthetic size,
// and on-demand for certain HTTP API requests. On-demand logical size calculation is also included
// because it is run as a sub-task of synthetic size.
// and on-demand for certain HTTP API requests.
}
_ => {
// In all other contexts, it is unusual to do I/O involving layers which are not visible at
@@ -458,8 +456,8 @@ impl Layer {
// This case is legal in brief time windows: for example an in-flight getpage request can hold on to a layer object
// which was covered by a concurrent compaction.
tracing::info!(
layer=%self,
"became visible as a result of access",
"Layer {} became visible as a result of access",
self.0.desc.key()
);
}
}
@@ -688,9 +686,7 @@ impl Drop for LayerInner {
// and we could be delaying shutdown for nothing.
}
let timeline = self.timeline.upgrade();
if let Some(timeline) = timeline.as_ref() {
if let Some(timeline) = self.timeline.upgrade() {
// Only need to decrement metrics if the timeline still exists: otherwise
// it will have already de-registered these metrics via TimelineMetrics::shutdown
if self.desc.is_delta() {
@@ -721,6 +717,7 @@ impl Drop for LayerInner {
let path = std::mem::take(&mut self.path);
let file_name = self.layer_desc().layer_name();
let file_size = self.layer_desc().file_size;
let timeline = self.timeline.clone();
let meta = self.metadata();
let status = self.status.take();
@@ -730,7 +727,7 @@ impl Drop for LayerInner {
// carry this until we are finished for [`Layer::wait_drop`] support
let _status = status;
let Some(timeline) = timeline else {
let Some(timeline) = timeline.upgrade() else {
// no need to nag that timeline is gone: under normal situation on
// task_mgr::remove_tenant_from_memory the timeline is gone before we get dropped.
LAYER_IMPL_METRICS.inc_deletes_failed(DeleteFailed::TimelineGone);

View File

@@ -48,6 +48,7 @@ use utils::{
sync::gate::{Gate, GateGuard},
};
use std::pin::pin;
use std::sync::atomic::Ordering as AtomicOrdering;
use std::sync::{Arc, Mutex, RwLock, Weak};
use std::time::{Duration, Instant, SystemTime};
@@ -61,7 +62,6 @@ use std::{
collections::btree_map::Entry,
ops::{Deref, Range},
};
use std::{pin::pin, sync::OnceLock};
use crate::{
aux_file::AuxFileSizeEstimator,
@@ -71,7 +71,6 @@ use crate::{
metadata::TimelineMetadata,
storage_layer::{inmemory_layer::IndexEntry, PersistentLayerDesc},
},
walingest::WalLagCooldown,
walredo,
};
use crate::{
@@ -430,8 +429,6 @@ pub struct Timeline {
pub(crate) l0_flush_global_state: L0FlushGlobalState,
pub(crate) handles: handle::PerTimelineState<crate::page_service::TenantManagerTypes>,
pub(crate) attach_wal_lag_cooldown: Arc<OnceLock<WalLagCooldown>>,
}
pub struct WalReceiverInfo {
@@ -740,7 +737,6 @@ pub enum GetLogicalSizePriority {
pub(crate) enum CompactFlags {
ForceRepartition,
ForceImageLayerCreation,
ForceL0Compaction,
EnhancedGcBottomMostCompaction,
DryRun,
}
@@ -2134,7 +2130,6 @@ impl Timeline {
pg_version: u32,
state: TimelineState,
aux_file_policy: Option<AuxFilePolicy>,
attach_wal_lag_cooldown: Arc<OnceLock<WalLagCooldown>>,
cancel: CancellationToken,
) -> Arc<Self> {
let disk_consistent_lsn = metadata.disk_consistent_lsn();
@@ -2276,8 +2271,6 @@ impl Timeline {
l0_flush_global_state: resources.l0_flush_global_state,
handles: Default::default(),
attach_wal_lag_cooldown,
};
if aux_file_policy == Some(AuxFilePolicy::V1) {

View File

@@ -11,7 +11,6 @@ pub(crate) struct RangeAnalysis {
has_image: bool,
num_of_deltas_above_image: usize,
total_num_of_deltas: usize,
num_of_l0: usize,
}
impl Timeline {
@@ -21,10 +20,8 @@ impl Timeline {
let mut delta_ranges = Vec::new();
let mut image_ranges = Vec::new();
let num_of_l0;
let all_layer_files = {
let guard = self.layers.read().await;
num_of_l0 = guard.layer_map().unwrap().level0_deltas().len();
guard.all_persistent_layers()
};
let lsn = self.get_last_record_lsn();
@@ -85,7 +82,6 @@ impl Timeline {
has_image: image_layer.is_some(),
num_of_deltas_above_image: maybe_delta_layers.len(),
total_num_of_deltas: pitr_delta_layers.len(),
num_of_l0,
});
}

View File

@@ -353,13 +353,7 @@ impl Timeline {
// 2. Compact
let timer = self.metrics.compact_time_histo.start_timer();
let fully_compacted = self
.compact_level0(
target_file_size,
flags.contains(CompactFlags::ForceL0Compaction),
ctx,
)
.await?;
let fully_compacted = self.compact_level0(target_file_size, ctx).await?;
timer.stop_and_record();
let mut partitioning = dense_partitioning;
@@ -664,7 +658,6 @@ impl Timeline {
async fn compact_level0(
self: &Arc<Self>,
target_file_size: u64,
force_compaction_ignore_threshold: bool,
ctx: &RequestContext,
) -> Result<bool, CompactionError> {
let CompactLevel0Phase1Result {
@@ -686,15 +679,9 @@ impl Timeline {
let now = tokio::time::Instant::now();
stats.read_lock_acquisition_micros =
DurationRecorder::Recorded(RecordedDuration(now - begin), now);
self.compact_level0_phase1(
phase1_layers_locked,
stats,
target_file_size,
force_compaction_ignore_threshold,
&ctx,
)
.instrument(phase1_span)
.await?
self.compact_level0_phase1(phase1_layers_locked, stats, target_file_size, &ctx)
.instrument(phase1_span)
.await?
};
if new_layers.is_empty() && deltas_to_compact.is_empty() {
@@ -713,7 +700,6 @@ impl Timeline {
guard: tokio::sync::RwLockReadGuard<'a, LayerManager>,
mut stats: CompactLevel0Phase1StatsBuilder,
target_file_size: u64,
force_compaction_ignore_threshold: bool,
ctx: &RequestContext,
) -> Result<CompactLevel0Phase1Result, CompactionError> {
stats.read_lock_held_spawn_blocking_startup_micros =
@@ -725,26 +711,11 @@ impl Timeline {
// Only compact if enough layers have accumulated.
let threshold = self.get_compaction_threshold();
if level0_deltas.is_empty() || level0_deltas.len() < threshold {
if force_compaction_ignore_threshold {
if !level0_deltas.is_empty() {
info!(
level0_deltas = level0_deltas.len(),
threshold, "too few deltas to compact, but forcing compaction"
);
} else {
info!(
level0_deltas = level0_deltas.len(),
threshold, "too few deltas to compact, cannot force compaction"
);
return Ok(CompactLevel0Phase1Result::default());
}
} else {
debug!(
level0_deltas = level0_deltas.len(),
threshold, "too few deltas to compact"
);
return Ok(CompactLevel0Phase1Result::default());
}
debug!(
level0_deltas = level0_deltas.len(),
threshold, "too few deltas to compact"
);
return Ok(CompactLevel0Phase1Result::default());
}
let mut level0_deltas = level0_deltas

View File

@@ -18,7 +18,7 @@
use std::collections::BTreeMap;
use std::ops::Deref;
use bytes::{Bytes, BytesMut};
use bytes::Bytes;
use pageserver_api::key::Key;
use tokio::io::AsyncWriteExt;
use tokio_epoll_uring::BoundedBuf;
@@ -27,6 +27,7 @@ use utils::vec_map::VecMap;
use crate::context::RequestContext;
use crate::tenant::blob_io::{BYTE_UNCOMPRESSED, BYTE_ZSTD, LEN_COMPRESSION_BIT_MASK};
use crate::virtual_file::dio::IoBufferMut;
use crate::virtual_file::{self, VirtualFile};
/// Metadata bundled with the start and end offset of a blob.
@@ -158,7 +159,7 @@ impl std::fmt::Display for VectoredBlob {
/// Return type of [`VectoredBlobReader::read_blobs`]
pub struct VectoredBlobsBuf {
/// Buffer for all blobs in this read
pub buf: BytesMut,
pub buf: IoBufferMut,
/// Offsets into the buffer and metadata for all blobs in this read
pub blobs: Vec<VectoredBlob>,
}
@@ -460,7 +461,7 @@ impl<'a> VectoredBlobReader<'a> {
pub async fn read_blobs(
&self,
read: &VectoredRead,
buf: BytesMut,
buf: IoBufferMut,
ctx: &RequestContext,
) -> Result<VectoredBlobsBuf, std::io::Error> {
assert!(read.size() > 0);
@@ -945,7 +946,8 @@ mod tests {
// Multiply by two (compressed data might need more space), and add a few bytes for the header
let reserved_bytes = blobs.iter().map(|bl| bl.len()).max().unwrap() * 2 + 16;
let mut buf = BytesMut::with_capacity(reserved_bytes);
let align = virtual_file::get_io_buffer_alignment();
let mut buf = IoBufferMut::with_capacity_aligned(reserved_bytes, align);
let align = virtual_file::get_io_buffer_alignment();
let vectored_blob_reader = VectoredBlobReader::new(&file);

View File

@@ -23,10 +23,12 @@ use pageserver_api::config::defaults::DEFAULT_IO_BUFFER_ALIGNMENT;
use pageserver_api::shard::TenantShardId;
use std::fs::File;
use std::io::{Error, ErrorKind, Seek, SeekFrom};
#[cfg(target_os = "linux")]
use std::os::unix::fs::OpenOptionsExt;
use tokio_epoll_uring::{BoundedBuf, IoBuf, IoBufMut, Slice};
use std::os::fd::{AsRawFd, FromRawFd, IntoRawFd, OwnedFd, RawFd};
use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
use std::sync::atomic::{AtomicBool, AtomicU8, AtomicUsize, Ordering};
use tokio::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
use tokio::time::Instant;
@@ -38,10 +40,11 @@ pub use io_engine::FeatureTestResult as IoEngineFeatureTestResult;
mod metadata;
mod open_options;
use self::owned_buffers_io::write::OwnedAsyncWriter;
pub(crate) use api::DirectIoMode;
pub(crate) use api::IoMode;
pub(crate) use io_engine::IoEngineKind;
pub(crate) use metadata::Metadata;
pub(crate) use open_options::*;
pub(crate) mod dio;
pub(crate) mod owned_buffers_io {
//! Abstractions for IO with owned buffers.
@@ -53,6 +56,7 @@ pub(crate) mod owned_buffers_io {
//! but for the time being we're proving out the primitives in the neon.git repo
//! for faster iteration.
pub(crate) mod io_buf_aligned;
pub(crate) mod io_buf_ext;
pub(crate) mod slice;
pub(crate) mod write;
@@ -61,6 +65,176 @@ pub(crate) mod owned_buffers_io {
}
}
#[derive(Debug)]
pub enum VirtualFile {
Buffered(VirtualFileInner),
Direct(VirtualFileInner),
}
impl VirtualFile {
fn inner(&self) -> &VirtualFileInner {
match self {
Self::Buffered(file) => file,
Self::Direct(file) => file,
}
}
fn inner_mut(&mut self) -> &mut VirtualFileInner {
match self {
Self::Buffered(file) => file,
Self::Direct(file) => file,
}
}
fn into_inner(self) -> VirtualFileInner {
match self {
Self::Buffered(file) => file,
Self::Direct(file) => file,
}
}
/// Open a file in read-only mode. Like File::open.
pub async fn open<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<Self, std::io::Error> {
let file = VirtualFileInner::open(path, ctx).await?;
Ok(Self::Buffered(file))
}
/// Open a file in read-only mode. Like File::open.
///
/// `O_DIRECT` will be enabled base on `virtual_file_io_mode`.
pub async fn open_v2<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<Self, std::io::Error> {
Self::open_with_options_v2(path.as_ref(), OpenOptions::new().read(true), ctx).await
}
pub async fn create<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<Self, std::io::Error> {
let file = VirtualFileInner::create(path, ctx).await?;
Ok(Self::Buffered(file))
}
pub async fn create_v2<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<Self, std::io::Error> {
VirtualFile::open_with_options_v2(
path.as_ref(),
OpenOptions::new().write(true).create(true).truncate(true),
ctx,
)
.await
}
pub async fn open_with_options<P: AsRef<Utf8Path>>(
path: P,
open_options: &OpenOptions,
ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
) -> Result<Self, std::io::Error> {
let file = VirtualFileInner::open_with_options(path, open_options, ctx).await?;
Ok(Self::Buffered(file))
}
pub async fn open_with_options_v2<P: AsRef<Utf8Path>>(
path: P,
open_options: &mut OpenOptions, // Uses `&mut` here to add `O_DIRECT`.
ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
) -> Result<Self, std::io::Error> {
let file = match get_io_mode() {
IoMode::Buffered => {
let file = VirtualFileInner::open_with_options(path, open_options, ctx).await?;
Self::Buffered(file)
}
#[cfg(target_os = "linux")]
IoMode::Direct => {
let file = VirtualFileInner::open_with_options(
path,
open_options.custom_flags(nix::libc::O_DIRECT),
ctx,
)
.await?;
Self::Direct(file)
}
};
Ok(file)
}
pub fn path(&self) -> &Utf8Path {
self.inner().path.as_path()
}
pub async fn crashsafe_overwrite<B: BoundedBuf<Buf = Buf> + Send, Buf: IoBuf + Send>(
final_path: Utf8PathBuf,
tmp_path: Utf8PathBuf,
content: B,
) -> std::io::Result<()> {
VirtualFileInner::crashsafe_overwrite(final_path, tmp_path, content).await
}
pub async fn sync_all(&self) -> Result<(), Error> {
self.inner().sync_all().await
}
pub async fn sync_data(&self) -> Result<(), Error> {
self.inner().sync_data().await
}
pub async fn metadata(&self) -> Result<Metadata, Error> {
self.inner().metadata().await
}
pub fn remove(self) {
self.into_inner().remove();
}
pub async fn seek(&mut self, pos: SeekFrom) -> Result<u64, Error> {
self.inner_mut().seek(pos).await
}
pub async fn read_exact_at<Buf>(
&self,
slice: Slice<Buf>,
offset: u64,
ctx: &RequestContext,
) -> Result<Slice<Buf>, Error>
where
Buf: IoBufMut + Send,
{
self.inner().read_exact_at(slice, offset, ctx).await
}
pub async fn read_exact_at_page(
&self,
page: PageWriteGuard<'static>,
offset: u64,
ctx: &RequestContext,
) -> Result<PageWriteGuard<'static>, Error> {
self.inner().read_exact_at_page(page, offset, ctx).await
}
pub async fn write_all_at<Buf: IoBuf + Send>(
&self,
buf: FullSlice<Buf>,
offset: u64,
ctx: &RequestContext,
) -> (FullSlice<Buf>, Result<(), Error>) {
self.inner().write_all_at(buf, offset, ctx).await
}
pub async fn write_all<Buf: IoBuf + Send>(
&mut self,
buf: FullSlice<Buf>,
ctx: &RequestContext,
) -> (FullSlice<Buf>, Result<usize, Error>) {
self.inner_mut().write_all(buf, ctx).await
}
}
///
/// A virtual file descriptor. You can use this just like std::fs::File, but internally
/// the underlying file is closed if the system is low on file descriptors,
@@ -77,7 +251,7 @@ pub(crate) mod owned_buffers_io {
/// 'tag' field is used to detect whether the handle still is valid or not.
///
#[derive(Debug)]
pub struct VirtualFile {
pub struct VirtualFileInner {
/// Lazy handle to the global file descriptor cache. The slot that this points to
/// might contain our File, or it may be empty, or it may contain a File that
/// belongs to a different VirtualFile.
@@ -350,12 +524,12 @@ macro_rules! with_file {
}};
}
impl VirtualFile {
impl VirtualFileInner {
/// Open a file in read-only mode. Like File::open.
pub async fn open<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<VirtualFile, std::io::Error> {
) -> Result<VirtualFileInner, std::io::Error> {
Self::open_with_options(path.as_ref(), OpenOptions::new().read(true), ctx).await
}
@@ -364,7 +538,7 @@ impl VirtualFile {
pub async fn create<P: AsRef<Utf8Path>>(
path: P,
ctx: &RequestContext,
) -> Result<VirtualFile, std::io::Error> {
) -> Result<VirtualFileInner, std::io::Error> {
Self::open_with_options(
path.as_ref(),
OpenOptions::new().write(true).create(true).truncate(true),
@@ -382,7 +556,7 @@ impl VirtualFile {
path: P,
open_options: &OpenOptions,
_ctx: &RequestContext, /* TODO: carry a pointer to the metrics in the RequestContext instead of the parsing https://github.com/neondatabase/neon/issues/6107 */
) -> Result<VirtualFile, std::io::Error> {
) -> Result<VirtualFileInner, std::io::Error> {
let path_ref = path.as_ref();
let path_str = path_ref.to_string();
let parts = path_str.split('/').collect::<Vec<&str>>();
@@ -413,7 +587,7 @@ impl VirtualFile {
open_options.open(path_ref.as_std_path()).await?
});
// Strip all options other than read and write.
// Strip all options other than read and write (O_DIRECT).
//
// It would perhaps be nicer to check just for the read and write flags
// explicitly, but OpenOptions doesn't contain any functions to read flags,
@@ -423,7 +597,7 @@ impl VirtualFile {
reopen_options.create_new(false);
reopen_options.truncate(false);
let vfile = VirtualFile {
let vfile = VirtualFileInner {
handle: RwLock::new(handle),
pos: 0,
path: path_ref.to_path_buf(),
@@ -1034,6 +1208,21 @@ impl tokio_epoll_uring::IoFd for FileGuard {
#[cfg(test)]
impl VirtualFile {
pub(crate) async fn read_blk(
&self,
blknum: u32,
ctx: &RequestContext,
) -> Result<crate::tenant::block_io::BlockLease<'_>, std::io::Error> {
self.inner().read_blk(blknum, ctx).await
}
async fn read_to_end(&mut self, buf: &mut Vec<u8>, ctx: &RequestContext) -> Result<(), Error> {
self.inner_mut().read_to_end(buf, ctx).await
}
}
#[cfg(test)]
impl VirtualFileInner {
pub(crate) async fn read_blk(
&self,
blknum: u32,
@@ -1067,7 +1256,7 @@ impl VirtualFile {
}
}
impl Drop for VirtualFile {
impl Drop for VirtualFileInner {
/// If a VirtualFile is dropped, close the underlying file if it was open.
fn drop(&mut self) {
let handle = self.handle.get_mut();
@@ -1216,6 +1405,15 @@ pub(crate) fn get_io_buffer_alignment() -> usize {
}
}
static IO_MODE: AtomicU8 = AtomicU8::new(IoMode::preferred() as u8);
pub(crate) fn set_io_mode(mode: IoMode) {
IO_MODE.store(mode as u8, std::sync::atomic::Ordering::Relaxed);
}
pub(crate) fn get_io_mode() -> IoMode {
IoMode::try_from(IO_MODE.load(Ordering::Relaxed)).unwrap()
}
#[cfg(test)]
mod tests {
use crate::context::DownloadBehavior;
@@ -1524,7 +1722,7 @@ mod tests {
// Open the file many times.
let mut files = Vec::new();
for _ in 0..VIRTUAL_FILES {
let f = VirtualFile::open_with_options(
let f = VirtualFileInner::open_with_options(
&test_file_path,
OpenOptions::new().read(true),
&ctx,
@@ -1576,7 +1774,7 @@ mod tests {
let path = testdir.join("myfile");
let tmp_path = testdir.join("myfile.tmp");
VirtualFile::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"foo".to_vec())
VirtualFileInner::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"foo".to_vec())
.await
.unwrap();
let mut file = MaybeVirtualFile::from(VirtualFile::open(&path, &ctx).await.unwrap());
@@ -1585,7 +1783,7 @@ mod tests {
assert!(!tmp_path.exists());
drop(file);
VirtualFile::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"bar".to_vec())
VirtualFileInner::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"bar".to_vec())
.await
.unwrap();
let mut file = MaybeVirtualFile::from(VirtualFile::open(&path, &ctx).await.unwrap());
@@ -1608,7 +1806,7 @@ mod tests {
std::fs::write(&tmp_path, "some preexisting junk that should be removed").unwrap();
assert!(tmp_path.exists());
VirtualFile::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"foo".to_vec())
VirtualFileInner::crashsafe_overwrite(path.clone(), tmp_path.clone(), b"foo".to_vec())
.await
.unwrap();

View File

@@ -0,0 +1,410 @@
#![allow(unused)]
use core::slice;
use std::{
alloc::{self, Layout},
cmp,
mem::{ManuallyDrop, MaybeUninit},
ops::{Deref, DerefMut},
ptr::{addr_of_mut, NonNull},
};
use bytes::buf::UninitSlice;
struct IoBufferPtr(*mut u8);
// SAFETY: We gurantees no one besides `IoBufferPtr` itself has the raw pointer.
unsafe impl Send for IoBufferPtr {}
/// An aligned buffer type used for I/O.
pub struct IoBufferMut {
ptr: IoBufferPtr,
capacity: usize,
len: usize,
align: usize,
}
impl IoBufferMut {
/// Constructs a new, empty `IoBufferMut` with at least the specified capacity and alignment.
///
/// The buffer will be able to hold at most `capacity` elements and will never resize.
///
///
/// # Panics
///
/// Panics if the new capacity exceeds `isize::MAX` _bytes_, or if the following alignment requirement is not met:
/// * `align` must not be zero,
///
/// * `align` must be a power of two,
///
/// * `capacity`, when rounded up to the nearest multiple of `align`,
/// must not overflow isize (i.e., the rounded value must be
/// less than or equal to `isize::MAX`).
pub fn with_capacity_aligned(capacity: usize, align: usize) -> Self {
let layout = Layout::from_size_align(capacity, align).expect("Invalid layout");
// SAFETY: Making an allocation with a sized and aligned layout. The memory is manually freed with the same layout.
let ptr = unsafe {
let ptr = alloc::alloc(layout);
if ptr.is_null() {
alloc::handle_alloc_error(layout);
}
IoBufferPtr(ptr)
};
IoBufferMut {
ptr,
capacity,
len: 0,
align,
}
}
/// Constructs a new `IoBufferMut` with at least the specified capacity and alignment, filled with zeros.
pub fn with_capacity_aligned_zeroed(capacity: usize, align: usize) -> Self {
use bytes::BufMut;
let mut buf = Self::with_capacity_aligned(capacity, align);
buf.put_bytes(0, capacity);
buf.len = capacity;
buf
}
/// Returns the total number of bytes the buffer can hold.
#[inline]
pub fn capacity(&self) -> usize {
self.capacity
}
/// Returns the alignment of the buffer.
#[inline]
pub fn align(&self) -> usize {
self.align
}
/// Returns the number of bytes in the buffer, also referred to as its 'length'.
#[inline]
pub fn len(&self) -> usize {
self.len
}
/// Force the length of the buffer to `new_len`.
#[inline]
unsafe fn set_len(&mut self, new_len: usize) {
debug_assert!(new_len <= self.capacity());
self.len = new_len;
}
#[inline]
fn as_ptr(&self) -> *const u8 {
self.ptr.0
}
#[inline]
fn as_mut_ptr(&mut self) -> *mut u8 {
self.ptr.0
}
/// Extracts a slice containing the entire buffer.
///
/// Equivalent to `&s[..]`.
#[inline]
fn as_slice(&self) -> &[u8] {
// SAFETY: The pointer is valid and `len` bytes are initialized.
unsafe { slice::from_raw_parts(self.as_ptr(), self.len) }
}
/// Extracts a mutable slice of the entire buffer.
///
/// Equivalent to `&mut s[..]`.
fn as_mut_slice(&mut self) -> &mut [u8] {
// SAFETY: The pointer is valid and `len` bytes are initialized.
unsafe { slice::from_raw_parts_mut(self.as_mut_ptr(), self.len) }
}
/// Drops the all the contents of the buffer, setting its length to `0`.
#[inline]
pub fn clear(&mut self) {
self.len = 0;
}
/// Reserves capacity for at least `additional` more bytes to be inserted
/// in the given `IoBufferMut`. The collection may reserve more space to
/// speculatively avoid frequent reallocations. After calling `reserve`,
/// capacity will be greater than or equal to `self.len() + additional`.
/// Does nothing if capacity is already sufficient.
///
/// # Panics
///
/// Panics if the new capacity exceeds `isize::MAX` _bytes_.
pub fn reserve(&mut self, additional: usize) {
if additional > self.capacity() - self.len() {
self.reserve_inner(additional);
}
}
fn reserve_inner(&mut self, additional: usize) {
let Some(required_cap) = self.len().checked_add(additional) else {
capacity_overflow()
};
let old_capacity = self.capacity();
let align = self.align();
// This guarantees exponential growth. The doubling cannot overflow
// because `cap <= isize::MAX` and the type of `cap` is `usize`.
let cap = cmp::max(old_capacity * 2, required_cap);
if !is_valid_alloc(cap) {
capacity_overflow()
}
let new_layout = Layout::from_size_align(cap, self.align()).expect("Invalid layout");
let old_ptr = self.as_mut_ptr();
// SAFETY: old allocation was allocated with std::alloc::alloc with the same layout,
// and we panics on null pointer.
let (ptr, cap) = unsafe {
let old_layout = Layout::from_size_align_unchecked(old_capacity, align);
let ptr = alloc::realloc(old_ptr, old_layout, new_layout.size());
if ptr.is_null() {
alloc::handle_alloc_error(new_layout);
}
(IoBufferPtr(ptr), cap)
};
self.ptr = ptr;
self.capacity = cap;
}
/// Consumes and leaks the `IoBufferMut`, returning a mutable reference to the contents, &'a mut [u8].
pub fn leak<'a>(self) -> &'a mut [u8] {
let mut buf = ManuallyDrop::new(self);
// SAFETY: leaking the buffer as intended.
unsafe { slice::from_raw_parts_mut(buf.as_mut_ptr(), buf.len) }
}
}
fn capacity_overflow() -> ! {
panic!("capacity overflow")
}
// We need to guarantee the following:
// * We don't ever allocate `> isize::MAX` byte-size objects.
// * We don't overflow `usize::MAX` and actually allocate too little.
//
// On 64-bit we just need to check for overflow since trying to allocate
// `> isize::MAX` bytes will surely fail. On 32-bit and 16-bit we need to add
// an extra guard for this in case we're running on a platform which can use
// all 4GB in user-space, e.g., PAE or x32.
#[inline]
fn is_valid_alloc(alloc_size: usize) -> bool {
!(usize::BITS < 64 && alloc_size > isize::MAX as usize)
}
impl Drop for IoBufferMut {
fn drop(&mut self) {
// SAFETY: memory was allocated with std::alloc::alloc with the same layout.
unsafe {
alloc::dealloc(
self.as_mut_ptr(),
Layout::from_size_align_unchecked(self.capacity, self.align),
)
}
}
}
impl Deref for IoBufferMut {
type Target = [u8];
fn deref(&self) -> &Self::Target {
self.as_slice()
}
}
impl DerefMut for IoBufferMut {
fn deref_mut(&mut self) -> &mut Self::Target {
self.as_mut_slice()
}
}
/// SAFETY: When advancing the internal cursor, the caller needs to make sure the bytes advcanced past have been initialized.
unsafe impl bytes::BufMut for IoBufferMut {
#[inline]
fn remaining_mut(&self) -> usize {
// Although a `Vec` can have at most isize::MAX bytes, we never want to grow `IoBufferMut`.
// Thus, it can have at most `self.capacity` bytes.
self.capacity() - self.len()
}
// SAFETY: Caller needs to make sure the bytes being advanced past have been initialized.
#[inline]
unsafe fn advance_mut(&mut self, cnt: usize) {
let len: usize = self.len();
let remaining = self.remaining_mut();
if remaining < cnt {
panic_advance(cnt, remaining);
}
// Addition will not overflow since the sum is at most the capacity.
self.set_len(len + cnt);
}
#[inline]
fn chunk_mut(&mut self) -> &mut bytes::buf::UninitSlice {
let cap = self.capacity();
let len = self.len();
// SAFETY: Since `self.ptr` is valid for `cap` bytes, `self.ptr.add(len)` must be
// valid for `cap - len` bytes. The subtraction will not underflow since
// `len <= cap`.
unsafe { UninitSlice::from_raw_parts_mut(self.as_mut_ptr().add(len), cap - len) }
}
}
/// Panic with a nice error message.
#[cold]
fn panic_advance(idx: usize, len: usize) -> ! {
panic!(
"advance out of bounds: the len is {} but advancing by {}",
len, idx
);
}
/// Safety: [`IoBufferMut`] has exclusive ownership of the io buffer,
/// and the location remains stable even if [`Self`] is moved.
unsafe impl tokio_epoll_uring::IoBuf for IoBufferMut {
fn stable_ptr(&self) -> *const u8 {
self.as_ptr()
}
fn bytes_init(&self) -> usize {
self.len()
}
fn bytes_total(&self) -> usize {
self.capacity()
}
}
// SAFETY: See above.
unsafe impl tokio_epoll_uring::IoBufMut for IoBufferMut {
fn stable_mut_ptr(&mut self) -> *mut u8 {
self.as_mut_ptr()
}
unsafe fn set_init(&mut self, init_len: usize) {
if self.len() < init_len {
self.set_len(init_len);
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_with_capacity_aligned() {
const ALIGN: usize = 4 * 1024;
let v = IoBufferMut::with_capacity_aligned(ALIGN * 4, ALIGN);
assert_eq!(v.len(), 0);
assert_eq!(v.capacity(), ALIGN * 4);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
let v = IoBufferMut::with_capacity_aligned(ALIGN / 2, ALIGN);
assert_eq!(v.len(), 0);
assert_eq!(v.capacity(), ALIGN / 2);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
}
#[test]
fn test_with_capacity_aligned_zeroed() {
const ALIGN: usize = 4 * 1024;
let v = IoBufferMut::with_capacity_aligned_zeroed(ALIGN, ALIGN);
assert_eq!(v.len(), ALIGN);
assert_eq!(v.capacity(), ALIGN);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
assert_eq!(&v[..], &[0; ALIGN])
}
#[test]
fn test_reserve() {
use bytes::BufMut;
const ALIGN: usize = 4 * 1024;
let mut v = IoBufferMut::with_capacity_aligned(ALIGN, ALIGN);
let capacity = v.capacity();
v.reserve(capacity);
assert_eq!(v.capacity(), capacity);
let data = [b'a'; ALIGN];
v.put(&data[..]);
v.reserve(capacity);
assert!(v.capacity() >= capacity * 2);
assert_eq!(&v[..], &data[..]);
let capacity = v.capacity();
v.clear();
v.reserve(capacity);
assert_eq!(capacity, v.capacity());
}
#[test]
fn test_bytes_put() {
use bytes::BufMut;
const ALIGN: usize = 4 * 1024;
let mut v = IoBufferMut::with_capacity_aligned(ALIGN * 4, ALIGN);
let x = [b'a'; ALIGN];
for _ in 0..2 {
for _ in 0..4 {
v.put(&x[..]);
}
assert_eq!(v.len(), ALIGN * 4);
assert_eq!(v.capacity(), ALIGN * 4);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
v.clear()
}
assert_eq!(v.len(), 0);
assert_eq!(v.capacity(), ALIGN * 4);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
}
#[test]
#[should_panic]
fn test_bytes_put_panic() {
use bytes::BufMut;
const ALIGN: usize = 4 * 1024;
let mut v = IoBufferMut::with_capacity_aligned(ALIGN * 4, ALIGN);
let x = [b'a'; ALIGN];
for _ in 0..5 {
v.put_slice(&x[..]);
}
}
#[test]
fn test_io_buf_put_slice() {
use tokio_epoll_uring::BoundedBufMut;
const ALIGN: usize = 4 * 1024;
let mut v = IoBufferMut::with_capacity_aligned(ALIGN, ALIGN);
let x = [b'a'; ALIGN];
for _ in 0..2 {
v.put_slice(&x[..]);
assert_eq!(v.len(), ALIGN);
assert_eq!(v.capacity(), ALIGN);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
v.clear()
}
assert_eq!(v.len(), 0);
assert_eq!(v.capacity(), ALIGN);
assert_eq!(v.align(), ALIGN);
assert_eq!(v.as_ptr().align_offset(ALIGN), 0);
}
}

View File

@@ -0,0 +1,9 @@
#![allow(unused)]
use tokio_epoll_uring::IoBufMut;
use crate::virtual_file::dio::IoBufferMut;
pub(crate) trait IoBufAlignedMut: IoBufMut {}
impl IoBufAlignedMut for IoBufferMut {}

View File

@@ -1,5 +1,6 @@
//! See [`FullSlice`].
use crate::virtual_file::dio::IoBufferMut;
use bytes::{Bytes, BytesMut};
use std::ops::{Deref, Range};
use tokio_epoll_uring::{BoundedBuf, IoBuf, Slice};
@@ -76,3 +77,4 @@ macro_rules! impl_io_buf_ext {
impl_io_buf_ext!(Bytes);
impl_io_buf_ext!(BytesMut);
impl_io_buf_ext!(Vec<u8>);
impl_io_buf_ext!(IoBufferMut);

View File

@@ -21,10 +21,7 @@
//! redo Postgres process, but some records it can handle directly with
//! bespoken Rust code.
use std::sync::Arc;
use std::sync::OnceLock;
use std::time::Duration;
use std::time::Instant;
use std::time::SystemTime;
use pageserver_api::shard::ShardIdentity;
@@ -72,29 +69,7 @@ impl CheckPoint {
}
}
/// Temporary limitation of WAL lag warnings after attach
///
/// After tenant attach, we want to limit WAL lag warnings because
/// we don't look at the WAL until the attach is complete, which
/// might take a while.
pub struct WalLagCooldown {
/// Until when should this limitation apply at all
active_until: std::time::Instant,
/// The maximum lag to suppress. Lags above this limit get reported anyways.
max_lag: Duration,
}
impl WalLagCooldown {
pub fn new(attach_start: Instant, attach_duration: Duration) -> Self {
Self {
active_until: attach_start + attach_duration * 3 + Duration::from_secs(120),
max_lag: attach_duration * 2 + Duration::from_secs(60),
}
}
}
pub struct WalIngest {
attach_wal_lag_cooldown: Arc<OnceLock<WalLagCooldown>>,
shard: ShardIdentity,
checkpoint: CheckPoint,
checkpoint_modified: bool,
@@ -128,7 +103,6 @@ impl WalIngest {
shard: *timeline.get_shard_identity(),
checkpoint,
checkpoint_modified: false,
attach_wal_lag_cooldown: timeline.attach_wal_lag_cooldown.clone(),
warn_ingest_lag: WarnIngestLag {
lag_msg_ratelimit: RateLimit::new(std::time::Duration::from_secs(10)),
future_lsn_msg_ratelimit: RateLimit::new(std::time::Duration::from_secs(10)),
@@ -1455,13 +1429,6 @@ impl WalIngest {
Ok(lag) => {
if lag > conf.wait_lsn_timeout {
rate_limits.lag_msg_ratelimit.call2(|rate_limit_stats| {
if let Some(cooldown) = self.attach_wal_lag_cooldown.get() {
if std::time::Instant::now() < cooldown.active_until && lag <= cooldown.max_lag {
return;
}
} else {
// Still loading? We shouldn't be here
}
let lag = humantime::format_duration(lag);
warn!(%rate_limit_stats, %lag, "ingesting record with timestamp lagging more than wait_lsn_timeout");
})

View File

@@ -42,7 +42,6 @@
#include "hll.h"
#include "bitmap.h"
#include "neon.h"
#define CriticalAssert(cond) do if (!(cond)) elog(PANIC, "Assertion %s failed at %s:%d: ", #cond, __FILE__, __LINE__); while (0)
@@ -174,9 +173,7 @@ lfc_disable(char const *op)
* If the reason of error is ENOSPC, then truncation of file may
* help to reclaim some space
*/
pgstat_report_wait_start(WAIT_EVENT_NEON_LFC_TRUNCATE);
int rc = ftruncate(lfc_desc, 0);
pgstat_report_wait_end();
if (rc < 0)
elog(WARNING, "Failed to truncate local file cache %s: %m", lfc_path);
@@ -772,10 +769,8 @@ lfc_readv_select(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
if (iteration_hits != 0)
{
pgstat_report_wait_start(WAIT_EVENT_NEON_LFC_READ);
rc = preadv(lfc_desc, iov, blocks_in_chunk,
((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
pgstat_report_wait_end();
if (rc != (BLCKSZ * blocks_in_chunk))
{
@@ -949,11 +944,8 @@ lfc_writev(NRelFileInfo rinfo, ForkNumber forkNum, BlockNumber blkno,
lfc_ctl->writes += blocks_in_chunk;
LWLockRelease(lfc_lock);
pgstat_report_wait_start(WAIT_EVENT_NEON_LFC_WRITE);
rc = pwritev(lfc_desc, iov, blocks_in_chunk,
((off_t) entry_offset * BLOCKS_PER_CHUNK + chunk_offs) * BLCKSZ);
pgstat_report_wait_end();
if (rc != BLCKSZ * blocks_in_chunk)
{
lfc_disable("write");

View File

@@ -490,7 +490,7 @@ pageserver_connect(shardno_t shard_no, int elevel)
WL_EXIT_ON_PM_DEATH | WL_LATCH_SET | WL_SOCKET_READABLE,
PQsocket(shard->conn),
0,
WAIT_EVENT_NEON_PS_STARTING);
PG_WAIT_EXTENSION);
elog(DEBUG5, "PGRES_POLLING_READING=>%d", rc);
if (rc & WL_LATCH_SET)
{
@@ -512,7 +512,7 @@ pageserver_connect(shardno_t shard_no, int elevel)
WL_EXIT_ON_PM_DEATH | WL_LATCH_SET | WL_SOCKET_WRITEABLE,
PQsocket(shard->conn),
0,
WAIT_EVENT_NEON_PS_STARTING);
PG_WAIT_EXTENSION);
elog(DEBUG5, "PGRES_POLLING_WRITING=>%d", rc);
if (rc & WL_LATCH_SET)
{
@@ -608,8 +608,7 @@ pageserver_connect(shardno_t shard_no, int elevel)
WaitEvent event;
/* Sleep until there's something to do */
(void) WaitEventSetWait(shard->wes_read, -1L, &event, 1,
WAIT_EVENT_NEON_PS_CONFIGURING);
(void) WaitEventSetWait(shard->wes_read, -1L, &event, 1, PG_WAIT_EXTENSION);
ResetLatch(MyLatch);
CHECK_FOR_INTERRUPTS();
@@ -657,8 +656,7 @@ static int
call_PQgetCopyData(shardno_t shard_no, char **buffer)
{
int ret;
PageServer *shard = &page_servers[shard_no];
PGconn *pageserver_conn = shard->conn;
PGconn *pageserver_conn = page_servers[shard_no].conn;
retry:
ret = PQgetCopyData(pageserver_conn, buffer, 1 /* async */ );
@@ -668,8 +666,7 @@ retry:
WaitEvent event;
/* Sleep until there's something to do */
(void) WaitEventSetWait(shard->wes_read, -1L, &event, 1,
WAIT_EVENT_NEON_PS_READ);
(void) WaitEventSetWait(page_servers[shard_no].wes_read, -1L, &event, 1, PG_WAIT_EXTENSION);
ResetLatch(MyLatch);
CHECK_FOR_INTERRUPTS();
@@ -940,7 +937,7 @@ PagestoreShmemInit(void)
LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
pagestore_shared = ShmemInitStruct("libpagestore shared state",
sizeof(PagestoreShmemState),
PagestoreShmemSize(),
&found);
if (!found)
{

View File

@@ -41,9 +41,6 @@
#include "pagestore_client.h"
#include "control_plane_connector.h"
#include "walsender_hooks.h"
#if PG_MAJORVERSION_NUM >= 16
#include "storage/ipc.h"
#endif
PG_MODULE_MAGIC;
void _PG_init(void);
@@ -52,23 +49,6 @@ static int logical_replication_max_snap_files = 300;
static int running_xacts_overflow_policy;
#if PG_MAJORVERSION_NUM >= 16
static shmem_startup_hook_type prev_shmem_startup_hook;
static void neon_shmem_startup_hook(void);
#endif
#if PG_MAJORVERSION_NUM >= 17
uint32 WAIT_EVENT_NEON_LFC_MAINTENANCE;
uint32 WAIT_EVENT_NEON_LFC_READ;
uint32 WAIT_EVENT_NEON_LFC_TRUNCATE;
uint32 WAIT_EVENT_NEON_LFC_WRITE;
uint32 WAIT_EVENT_NEON_PS_STARTING;
uint32 WAIT_EVENT_NEON_PS_CONFIGURING;
uint32 WAIT_EVENT_NEON_PS_SEND;
uint32 WAIT_EVENT_NEON_PS_READ;
uint32 WAIT_EVENT_NEON_WAL_DL;
#endif
enum RunningXactsOverflowPolicies {
OP_IGNORE,
OP_SKIP,
@@ -655,9 +635,6 @@ _PG_init(void)
*/
#if PG_VERSION_NUM >= 160000
load_file("$libdir/neon_rmgr", false);
prev_shmem_startup_hook = shmem_startup_hook;
shmem_startup_hook = neon_shmem_startup_hook;
#endif
pg_init_libpagestore();
@@ -744,25 +721,3 @@ backpressure_throttling_time(PG_FUNCTION_ARGS)
{
PG_RETURN_UINT64(BackpressureThrottlingTime());
}
#if PG_MAJORVERSION_NUM >= 16
static void
neon_shmem_startup_hook(void)
{
/* Initialize */
if (prev_shmem_startup_hook)
prev_shmem_startup_hook();
#if PG_PG_MAJORVERSION_NUM >= 17
WAIT_EVENT_NEON_LFC_MAINTENANCE = WaitEventExtensionNew("Neon/FileCache_Maintenance");
WAIT_EVENT_NEON_LFC_READ = WaitEventExtensionNew("Neon/FileCache_Read");
WAIT_EVENT_NEON_LFC_TRUNCATE = WaitEventExtensionNew("Neon/FileCache_Truncate");
WAIT_EVENT_NEON_LFC_WRITE = WaitEventExtensionNew("Neon/FileCache_Write");
WAIT_EVENT_NEON_PS_STARTING = WaitEventExtensionNew("Neon/PS_Starting");
WAIT_EVENT_NEON_PS_CONFIGURING = WaitEventExtensionNew("Neon/PS_Configuring");
WAIT_EVENT_NEON_PS_SEND = WaitEventExtensionNew("Neon/PS_SendIO");
WAIT_EVENT_NEON_PS_READ = WaitEventExtensionNew("Neon/PS_ReadIO");
WAIT_EVENT_NEON_WAL_DL = WaitEventExtensionNew("Neon/WAL_Download");
#endif
}
#endif

View File

@@ -1,6 +1,8 @@
# neon extension
comment = 'cloud storage for PostgreSQL'
default_version = '1.5'
# TODO: bump default version to 1.5, after we are certain that we don't
# need to rollback the compute image
default_version = '1.4'
module_pathname = '$libdir/neon'
relocatable = true
trusted = true

View File

@@ -12,7 +12,6 @@
#ifndef NEON_H
#define NEON_H
#include "access/xlogreader.h"
#include "utils/wait_event.h"
/* GUCs */
extern char *neon_auth_token;
@@ -23,28 +22,6 @@ extern char *wal_acceptors_list;
extern int wal_acceptor_reconnect_timeout;
extern int wal_acceptor_connection_timeout;
#if PG_MAJORVERSION_NUM >= 17
extern uint32 WAIT_EVENT_NEON_LFC_MAINTENANCE;
extern uint32 WAIT_EVENT_NEON_LFC_READ;
extern uint32 WAIT_EVENT_NEON_LFC_TRUNCATE;
extern uint32 WAIT_EVENT_NEON_LFC_WRITE;
extern uint32 WAIT_EVENT_NEON_PS_STARTING;
extern uint32 WAIT_EVENT_NEON_PS_CONFIGURING;
extern uint32 WAIT_EVENT_NEON_PS_SEND;
extern uint32 WAIT_EVENT_NEON_PS_READ;
extern uint32 WAIT_EVENT_NEON_WAL_DL;
#else
#define WAIT_EVENT_NEON_LFC_MAINTENANCE PG_WAIT_EXTENSION
#define WAIT_EVENT_NEON_LFC_READ WAIT_EVENT_BUFFILE_READ
#define WAIT_EVENT_NEON_LFC_TRUNCATE WAIT_EVENT_BUFFILE_TRUNCATE
#define WAIT_EVENT_NEON_LFC_WRITE WAIT_EVENT_BUFFILE_WRITE
#define WAIT_EVENT_NEON_PS_STARTING PG_WAIT_EXTENSION
#define WAIT_EVENT_NEON_PS_CONFIGURING PG_WAIT_EXTENSION
#define WAIT_EVENT_NEON_PS_SEND PG_WAIT_EXTENSION
#define WAIT_EVENT_NEON_PS_READ PG_WAIT_EXTENSION
#define WAIT_EVENT_NEON_WAL_DL WAIT_EVENT_WAL_READ
#endif
extern void pg_init_libpagestore(void);
extern void pg_init_walproposer(void);

View File

@@ -27,8 +27,7 @@ NeonPerfCountersShmemSize(void)
{
Size size = 0;
size = add_size(size, mul_size(NUM_NEON_PERF_COUNTER_SLOTS,
sizeof(neon_per_backend_counters)));
size = add_size(size, mul_size(MaxBackends, sizeof(neon_per_backend_counters)));
return size;
}
@@ -40,7 +39,7 @@ NeonPerfCountersShmemInit(void)
neon_per_backend_counters_shared =
ShmemInitStruct("Neon perf counters",
mul_size(NUM_NEON_PERF_COUNTER_SLOTS,
mul_size(MaxBackends,
sizeof(neon_per_backend_counters)),
&found);
Assert(found == IsUnderPostmaster);
@@ -138,7 +137,7 @@ neon_perf_counters_to_metrics(neon_per_backend_counters *counters)
metrics[i].is_bucket = false;
metrics[i].value = (double) counters->pageserver_requests_sent_total;
i++;
metrics[i].name = "pageserver_disconnects_total";
metrics[i].name = "pageserver_requests_disconnects_total";
metrics[i].is_bucket = false;
metrics[i].value = (double) counters->pageserver_disconnects_total;
i++;
@@ -193,7 +192,7 @@ neon_get_backend_perf_counters(PG_FUNCTION_ARGS)
/* We put all the tuples into a tuplestore in one go. */
InitMaterializedSRF(fcinfo, 0);
for (int procno = 0; procno < NUM_NEON_PERF_COUNTER_SLOTS; procno++)
for (int procno = 0; procno < MaxBackends; procno++)
{
PGPROC *proc = GetPGProcByNumber(procno);
int pid = proc->pid;
@@ -232,7 +231,7 @@ neon_get_perf_counters(PG_FUNCTION_ARGS)
InitMaterializedSRF(fcinfo, 0);
/* Aggregate the counters across all backends */
for (int procno = 0; procno < NUM_NEON_PERF_COUNTER_SLOTS; procno++)
for (int procno = 0; procno < MaxBackends; procno++)
{
neon_per_backend_counters *counters = &neon_per_backend_counters_shared[procno];

View File

@@ -96,14 +96,6 @@ typedef struct
/* Pointer to the shared memory array of neon_per_backend_counters structs */
extern neon_per_backend_counters *neon_per_backend_counters_shared;
/*
* Size of the perf counters array in shared memory. One slot for each backend
* and aux process. IOW one for each PGPROC slot, except for slots reserved
* for prepared transactions, because they're not real processes and cannot do
* I/O.
*/
#define NUM_NEON_PERF_COUNTER_SLOTS (MaxBackends + NUM_AUXILIARY_PROCS)
#if PG_VERSION_NUM >= 170000
#define MyNeonCounters (&neon_per_backend_counters_shared[MyProcNumber])
#else

View File

@@ -1773,20 +1773,6 @@ neon_init(void)
if (MyPState != NULL)
return;
/*
* Sanity check that theperf counters array is sized correctly. We got
* this wrong once, and the formula for max number of backends and aux
* processes might well change in the future, so better safe than sorry.
* This is a very cheap check so we do it even without assertions. On
* v14, this gets called before initializing MyProc, so we cannot perform
* the check here. That's OK, we don't expect the logic to change in old
* releases.
*/
#if PG_VERSION_NUM>=150000
if (MyNeonCounters >= &neon_per_backend_counters_shared[NUM_NEON_PERF_COUNTER_SLOTS])
elog(ERROR, "MyNeonCounters points past end of array");
#endif
prfs_size = offsetof(PrefetchState, prf_buffer) +
sizeof(PrefetchRequest) * readahead_buffer_size;

View File

@@ -213,7 +213,7 @@ WalProposerPoll(WalProposer *wp)
rc = wp->api.wait_event_set(wp, timeout, &sk, &events);
/* Exit loop if latch is set (we got new WAL) */
if (rc == 1 && (events & WL_LATCH_SET))
if ((rc == 1 && events & WL_LATCH_SET))
break;
/*

View File

@@ -422,9 +422,6 @@ backpressure_throttling_impl(void)
TimestampTz start,
stop;
bool retry = false;
char *new_status = NULL;
const char *old_status;
int len;
if (PointerIsValid(PrevProcessInterruptsCallback))
retry = PrevProcessInterruptsCallback();
@@ -445,24 +442,14 @@ backpressure_throttling_impl(void)
if (lag == 0)
return retry;
old_status = get_ps_display(&len);
new_status = (char *) palloc(len + 64 + 1);
memcpy(new_status, old_status, len);
snprintf(new_status + len, 64, "backpressure throttling: lag %lu", lag);
set_ps_display(new_status);
new_status[len] = '\0'; /* truncate off " backpressure ..." to later reset the ps */
/* Suspend writers until replicas catch up */
set_ps_display("backpressure throttling");
elog(DEBUG2, "backpressure throttling: lag %lu", lag);
start = GetCurrentTimestamp();
pg_usleep(BACK_PRESSURE_DELAY);
stop = GetCurrentTimestamp();
pg_atomic_add_fetch_u64(&walprop_shared->backpressureThrottlingTime, stop - start);
/* Reset ps display */
set_ps_display(new_status);
pfree(new_status);
return true;
}
@@ -1814,7 +1801,7 @@ walprop_pg_wait_event_set(WalProposer *wp, long timeout, Safekeeper **sk, uint32
* If wait is terminated by latch set (walsenders' latch is set on each
* wal flush). (no need for pm death check due to WL_EXIT_ON_PM_DEATH)
*/
if ((rc == 1 && (event.events & WL_LATCH_SET)) || late_cv_trigger)
if ((rc == 1 && event.events & WL_LATCH_SET) || late_cv_trigger)
{
/* Reset our latch */
ResetLatch(MyLatch);
@@ -1826,7 +1813,7 @@ walprop_pg_wait_event_set(WalProposer *wp, long timeout, Safekeeper **sk, uint32
* If the event contains something about the socket, it means we got an
* event from a safekeeper socket.
*/
if (rc == 1 && (event.events & WL_SOCKET_MASK))
if (rc == 1 && (event.events & (WL_SOCKET_MASK)))
{
*sk = (Safekeeper *) event.user_data;
*events = event.events;

View File

@@ -160,7 +160,7 @@ NeonWALPageRead(
WL_LATCH_SET | WL_EXIT_ON_PM_DEATH | reader_events,
sock,
timeout_ms,
WAIT_EVENT_NEON_WAL_DL);
WAIT_EVENT_WAL_SENDER_MAIN);
}
}
}
@@ -191,14 +191,13 @@ NeonOnDemandXLogReaderRoutines(XLogReaderRoutine *xlr)
if (!wal_reader)
{
XLogRecPtr basebackupLsn = GetRedoStartLsn();
XLogRecPtr epochStartLsn = pg_atomic_read_u64(&GetWalpropShmemState()->propEpochStartLsn);
/* should never happen */
if (basebackupLsn == 0)
if (epochStartLsn == 0)
{
elog(ERROR, "unable to start walsender when basebackupLsn is 0");
elog(ERROR, "Unable to start walsender when propEpochStartLsn is 0!");
}
wal_reader = NeonWALReaderAllocate(wal_segment_size, basebackupLsn, "[walsender] ");
wal_reader = NeonWALReaderAllocate(wal_segment_size, epochStartLsn, "[walsender] ");
}
xlr->page_read = NeonWALPageRead;
xlr->segment_open = NeonWALReadSegmentOpen;

154
poetry.lock generated
View File

@@ -2064,80 +2064,73 @@ test = ["enum34", "ipaddress", "mock", "pywin32", "wmi"]
[[package]]
name = "psycopg2-binary"
version = "2.9.9"
version = "2.9.6"
description = "psycopg2 - Python-PostgreSQL Database Adapter"
optional = false
python-versions = ">=3.7"
python-versions = ">=3.6"
files = [
{file = "psycopg2-binary-2.9.9.tar.gz", hash = "sha256:7f01846810177d829c7692f1f5ada8096762d9172af1b1a28d4ab5b77c923c1c"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:c2470da5418b76232f02a2fcd2229537bb2d5a7096674ce61859c3229f2eb202"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:c6af2a6d4b7ee9615cbb162b0738f6e1fd1f5c3eda7e5da17861eacf4c717ea7"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:75723c3c0fbbf34350b46a3199eb50638ab22a0228f93fb472ef4d9becc2382b"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:83791a65b51ad6ee6cf0845634859d69a038ea9b03d7b26e703f94c7e93dbcf9"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:0ef4854e82c09e84cc63084a9e4ccd6d9b154f1dbdd283efb92ecd0b5e2b8c84"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ed1184ab8f113e8d660ce49a56390ca181f2981066acc27cf637d5c1e10ce46e"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:d2997c458c690ec2bc6b0b7ecbafd02b029b7b4283078d3b32a852a7ce3ddd98"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:b58b4710c7f4161b5e9dcbe73bb7c62d65670a87df7bcce9e1faaad43e715245"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-musllinux_1_1_ppc64le.whl", hash = "sha256:0c009475ee389757e6e34611d75f6e4f05f0cf5ebb76c6037508318e1a1e0d7e"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:8dbf6d1bc73f1d04ec1734bae3b4fb0ee3cb2a493d35ede9badbeb901fb40f6f"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-win32.whl", hash = "sha256:3f78fd71c4f43a13d342be74ebbc0666fe1f555b8837eb113cb7416856c79682"},
{file = "psycopg2_binary-2.9.9-cp310-cp310-win_amd64.whl", hash = "sha256:876801744b0dee379e4e3c38b76fc89f88834bb15bf92ee07d94acd06ec890a0"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:ee825e70b1a209475622f7f7b776785bd68f34af6e7a46e2e42f27b659b5bc26"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:1ea665f8ce695bcc37a90ee52de7a7980be5161375d42a0b6c6abedbf0d81f0f"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:143072318f793f53819048fdfe30c321890af0c3ec7cb1dfc9cc87aa88241de2"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:c332c8d69fb64979ebf76613c66b985414927a40f8defa16cf1bc028b7b0a7b0"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f7fc5a5acafb7d6ccca13bfa8c90f8c51f13d8fb87d95656d3950f0158d3ce53"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:977646e05232579d2e7b9c59e21dbe5261f403a88417f6a6512e70d3f8a046be"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:b6356793b84728d9d50ead16ab43c187673831e9d4019013f1402c41b1db9b27"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:bc7bb56d04601d443f24094e9e31ae6deec9ccb23581f75343feebaf30423359"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-musllinux_1_1_ppc64le.whl", hash = "sha256:77853062a2c45be16fd6b8d6de2a99278ee1d985a7bd8b103e97e41c034006d2"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:78151aa3ec21dccd5cdef6c74c3e73386dcdfaf19bced944169697d7ac7482fc"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-win32.whl", hash = "sha256:dc4926288b2a3e9fd7b50dc6a1909a13bbdadfc67d93f3374d984e56f885579d"},
{file = "psycopg2_binary-2.9.9-cp311-cp311-win_amd64.whl", hash = "sha256:b76bedd166805480ab069612119ea636f5ab8f8771e640ae103e05a4aae3e417"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:8532fd6e6e2dc57bcb3bc90b079c60de896d2128c5d9d6f24a63875a95a088cf"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8f8544b092a29a6ddd72f3556a9fcf249ec412e10ad28be6a0c0d948924f2212"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:2d423c8d8a3c82d08fe8af900ad5b613ce3632a1249fd6a223941d0735fce493"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:2e5afae772c00980525f6d6ecf7cbca55676296b580c0e6abb407f15f3706996"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6e6f98446430fdf41bd36d4faa6cb409f5140c1c2cf58ce0bbdaf16af7d3f119"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:c77e3d1862452565875eb31bdb45ac62502feabbd53429fdc39a1cc341d681ba"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-musllinux_1_1_i686.whl", hash = "sha256:cb16c65dcb648d0a43a2521f2f0a2300f40639f6f8c1ecbc662141e4e3e1ee07"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-musllinux_1_1_ppc64le.whl", hash = "sha256:911dda9c487075abd54e644ccdf5e5c16773470a6a5d3826fda76699410066fb"},
{file = "psycopg2_binary-2.9.9-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:57fede879f08d23c85140a360c6a77709113efd1c993923c59fde17aa27599fe"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:2293b001e319ab0d869d660a704942c9e2cce19745262a8aba2115ef41a0a42a"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:03ef7df18daf2c4c07e2695e8cfd5ee7f748a1d54d802330985a78d2a5a6dca9"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:0a602ea5aff39bb9fac6308e9c9d82b9a35c2bf288e184a816002c9fae930b77"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8359bf4791968c5a78c56103702000105501adb557f3cf772b2c207284273984"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:275ff571376626195ab95a746e6a04c7df8ea34638b99fc11160de91f2fef503"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:f9b5571d33660d5009a8b3c25dc1db560206e2d2f89d3df1cb32d72c0d117d52"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:420f9bbf47a02616e8554e825208cb947969451978dceb77f95ad09c37791dae"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-musllinux_1_1_ppc64le.whl", hash = "sha256:4154ad09dac630a0f13f37b583eae260c6aa885d67dfbccb5b02c33f31a6d420"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:a148c5d507bb9b4f2030a2025c545fccb0e1ef317393eaba42e7eabd28eb6041"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-win32.whl", hash = "sha256:68fc1f1ba168724771e38bee37d940d2865cb0f562380a1fb1ffb428b75cb692"},
{file = "psycopg2_binary-2.9.9-cp37-cp37m-win_amd64.whl", hash = "sha256:281309265596e388ef483250db3640e5f414168c5a67e9c665cafce9492eda2f"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:60989127da422b74a04345096c10d416c2b41bd7bf2a380eb541059e4e999980"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:246b123cc54bb5361588acc54218c8c9fb73068bf227a4a531d8ed56fa3ca7d6"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:34eccd14566f8fe14b2b95bb13b11572f7c7d5c36da61caf414d23b91fcc5d94"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:18d0ef97766055fec15b5de2c06dd8e7654705ce3e5e5eed3b6651a1d2a9a152"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:d3f82c171b4ccd83bbaf35aa05e44e690113bd4f3b7b6cc54d2219b132f3ae55"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ead20f7913a9c1e894aebe47cccf9dc834e1618b7aa96155d2091a626e59c972"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:ca49a8119c6cbd77375ae303b0cfd8c11f011abbbd64601167ecca18a87e7cdd"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:323ba25b92454adb36fa425dc5cf6f8f19f78948cbad2e7bc6cdf7b0d7982e59"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-musllinux_1_1_ppc64le.whl", hash = "sha256:1236ed0952fbd919c100bc839eaa4a39ebc397ed1c08a97fc45fee2a595aa1b3"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:729177eaf0aefca0994ce4cffe96ad3c75e377c7b6f4efa59ebf003b6d398716"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-win32.whl", hash = "sha256:804d99b24ad523a1fe18cc707bf741670332f7c7412e9d49cb5eab67e886b9b5"},
{file = "psycopg2_binary-2.9.9-cp38-cp38-win_amd64.whl", hash = "sha256:a6cdcc3ede532f4a4b96000b6362099591ab4a3e913d70bcbac2b56c872446f7"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:72dffbd8b4194858d0941062a9766f8297e8868e1dd07a7b36212aaa90f49472"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:30dcc86377618a4c8f3b72418df92e77be4254d8f89f14b8e8f57d6d43603c0f"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:31a34c508c003a4347d389a9e6fcc2307cc2150eb516462a7a17512130de109e"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:15208be1c50b99203fe88d15695f22a5bed95ab3f84354c494bcb1d08557df67"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1873aade94b74715be2246321c8650cabf5a0d098a95bab81145ffffa4c13876"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:3a58c98a7e9c021f357348867f537017057c2ed7f77337fd914d0bedb35dace7"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:4686818798f9194d03c9129a4d9a702d9e113a89cb03bffe08c6cf799e053291"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:ebdc36bea43063116f0486869652cb2ed7032dbc59fbcb4445c4862b5c1ecf7f"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-musllinux_1_1_ppc64le.whl", hash = "sha256:ca08decd2697fdea0aea364b370b1249d47336aec935f87b8bbfd7da5b2ee9c1"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:ac05fb791acf5e1a3e39402641827780fe44d27e72567a000412c648a85ba860"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-win32.whl", hash = "sha256:9dba73be7305b399924709b91682299794887cbbd88e38226ed9f6712eabee90"},
{file = "psycopg2_binary-2.9.9-cp39-cp39-win_amd64.whl", hash = "sha256:f7ae5d65ccfbebdfa761585228eb4d0df3a8b15cfb53bd953e713e09fbb12957"},
{file = "psycopg2-binary-2.9.6.tar.gz", hash = "sha256:1f64dcfb8f6e0c014c7f55e51c9759f024f70ea572fbdef123f85318c297947c"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:d26e0342183c762de3276cca7a530d574d4e25121ca7d6e4a98e4f05cb8e4df7"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:c48d8f2db17f27d41fb0e2ecd703ea41984ee19362cbce52c097963b3a1b4365"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ffe9dc0a884a8848075e576c1de0290d85a533a9f6e9c4e564f19adf8f6e54a7"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:8a76e027f87753f9bd1ab5f7c9cb8c7628d1077ef927f5e2446477153a602f2c"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:6460c7a99fc939b849431f1e73e013d54aa54293f30f1109019c56a0b2b2ec2f"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ae102a98c547ee2288637af07393dd33f440c25e5cd79556b04e3fca13325e5f"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:9972aad21f965599ed0106f65334230ce826e5ae69fda7cbd688d24fa922415e"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:7a40c00dbe17c0af5bdd55aafd6ff6679f94a9be9513a4c7e071baf3d7d22a70"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-musllinux_1_1_ppc64le.whl", hash = "sha256:cacbdc5839bdff804dfebc058fe25684cae322987f7a38b0168bc1b2df703fb1"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:7f0438fa20fb6c7e202863e0d5ab02c246d35efb1d164e052f2f3bfe2b152bd0"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-win32.whl", hash = "sha256:b6c8288bb8a84b47e07013bb4850f50538aa913d487579e1921724631d02ea1b"},
{file = "psycopg2_binary-2.9.6-cp310-cp310-win_amd64.whl", hash = "sha256:61b047a0537bbc3afae10f134dc6393823882eb263088c271331602b672e52e9"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:964b4dfb7c1c1965ac4c1978b0f755cc4bd698e8aa2b7667c575fb5f04ebe06b"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:afe64e9b8ea66866a771996f6ff14447e8082ea26e675a295ad3bdbffdd72afb"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:15e2ee79e7cf29582ef770de7dab3d286431b01c3bb598f8e05e09601b890081"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:dfa74c903a3c1f0d9b1c7e7b53ed2d929a4910e272add6700c38f365a6002820"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:b83456c2d4979e08ff56180a76429263ea254c3f6552cd14ada95cff1dec9bb8"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0645376d399bfd64da57148694d78e1f431b1e1ee1054872a5713125681cf1be"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:e99e34c82309dd78959ba3c1590975b5d3c862d6f279f843d47d26ff89d7d7e1"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:4ea29fc3ad9d91162c52b578f211ff1c931d8a38e1f58e684c45aa470adf19e2"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-musllinux_1_1_ppc64le.whl", hash = "sha256:4ac30da8b4f57187dbf449294d23b808f8f53cad6b1fc3623fa8a6c11d176dd0"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:e78e6e2a00c223e164c417628572a90093c031ed724492c763721c2e0bc2a8df"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-win32.whl", hash = "sha256:1876843d8e31c89c399e31b97d4b9725a3575bb9c2af92038464231ec40f9edb"},
{file = "psycopg2_binary-2.9.6-cp311-cp311-win_amd64.whl", hash = "sha256:b4b24f75d16a89cc6b4cdff0eb6a910a966ecd476d1e73f7ce5985ff1328e9a6"},
{file = "psycopg2_binary-2.9.6-cp36-cp36m-win32.whl", hash = "sha256:498807b927ca2510baea1b05cc91d7da4718a0f53cb766c154c417a39f1820a0"},
{file = "psycopg2_binary-2.9.6-cp36-cp36m-win_amd64.whl", hash = "sha256:0d236c2825fa656a2d98bbb0e52370a2e852e5a0ec45fc4f402977313329174d"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:34b9ccdf210cbbb1303c7c4db2905fa0319391bd5904d32689e6dd5c963d2ea8"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:84d2222e61f313c4848ff05353653bf5f5cf6ce34df540e4274516880d9c3763"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:30637a20623e2a2eacc420059be11527f4458ef54352d870b8181a4c3020ae6b"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8122cfc7cae0da9a3077216528b8bb3629c43b25053284cc868744bfe71eb141"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:38601cbbfe600362c43714482f43b7c110b20cb0f8172422c616b09b85a750c5"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:c7e62ab8b332147a7593a385d4f368874d5fe4ad4e341770d4983442d89603e3"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:2ab652e729ff4ad76d400df2624d223d6e265ef81bb8aa17fbd63607878ecbee"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-musllinux_1_1_ppc64le.whl", hash = "sha256:c83a74b68270028dc8ee74d38ecfaf9c90eed23c8959fca95bd703d25b82c88e"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:d4e6036decf4b72d6425d5b29bbd3e8f0ff1059cda7ac7b96d6ac5ed34ffbacd"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-win32.whl", hash = "sha256:a8c28fd40a4226b4a84bdf2d2b5b37d2c7bd49486b5adcc200e8c7ec991dfa7e"},
{file = "psycopg2_binary-2.9.6-cp37-cp37m-win_amd64.whl", hash = "sha256:51537e3d299be0db9137b321dfb6a5022caaab275775680e0c3d281feefaca6b"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:cf4499e0a83b7b7edcb8dabecbd8501d0d3a5ef66457200f77bde3d210d5debb"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:7e13a5a2c01151f1208d5207e42f33ba86d561b7a89fca67c700b9486a06d0e2"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0e0f754d27fddcfd74006455b6e04e6705d6c31a612ec69ddc040a5468e44b4e"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:d57c3fd55d9058645d26ae37d76e61156a27722097229d32a9e73ed54819982a"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:71f14375d6f73b62800530b581aed3ada394039877818b2d5f7fc77e3bb6894d"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:441cc2f8869a4f0f4bb408475e5ae0ee1f3b55b33f350406150277f7f35384fc"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:65bee1e49fa6f9cf327ce0e01c4c10f39165ee76d35c846ade7cb0ec6683e303"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:af335bac6b666cc6aea16f11d486c3b794029d9df029967f9938a4bed59b6a19"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-musllinux_1_1_ppc64le.whl", hash = "sha256:cfec476887aa231b8548ece2e06d28edc87c1397ebd83922299af2e051cf2827"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:65c07febd1936d63bfde78948b76cd4c2a411572a44ac50719ead41947d0f26b"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-win32.whl", hash = "sha256:4dfb4be774c4436a4526d0c554af0cc2e02082c38303852a36f6456ece7b3503"},
{file = "psycopg2_binary-2.9.6-cp38-cp38-win_amd64.whl", hash = "sha256:02c6e3cf3439e213e4ee930308dc122d6fb4d4bea9aef4a12535fbd605d1a2fe"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:e9182eb20f41417ea1dd8e8f7888c4d7c6e805f8a7c98c1081778a3da2bee3e4"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:8a6979cf527e2603d349a91060f428bcb135aea2be3201dff794813256c274f1"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8338a271cb71d8da40b023a35d9c1e919eba6cbd8fa20a54b748a332c355d896"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:e3ed340d2b858d6e6fb5083f87c09996506af483227735de6964a6100b4e6a54"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f81e65376e52f03422e1fb475c9514185669943798ed019ac50410fb4c4df232"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bfb13af3c5dd3a9588000910178de17010ebcccd37b4f9794b00595e3a8ddad3"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:4c727b597c6444a16e9119386b59388f8a424223302d0c06c676ec8b4bc1f963"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:4d67fbdaf177da06374473ef6f7ed8cc0a9dc640b01abfe9e8a2ccb1b1402c1f"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-musllinux_1_1_ppc64le.whl", hash = "sha256:0892ef645c2fabb0c75ec32d79f4252542d0caec1d5d949630e7d242ca4681a3"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:02c0f3757a4300cf379eb49f543fb7ac527fb00144d39246ee40e1df684ab514"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-win32.whl", hash = "sha256:c3dba7dab16709a33a847e5cd756767271697041fbe3fe97c215b1fc1f5c9848"},
{file = "psycopg2_binary-2.9.6-cp39-cp39-win_amd64.whl", hash = "sha256:f6a88f384335bb27812293fdb11ac6aee2ca3f51d3c7820fe03de0a304ab6249"},
]
[[package]]
@@ -2584,6 +2577,7 @@ files = [
{file = "PyYAML-6.0.1-cp311-cp311-win_amd64.whl", hash = "sha256:bf07ee2fef7014951eeb99f56f39c9bb4af143d8aa3c21b1677805985307da34"},
{file = "PyYAML-6.0.1-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:855fb52b0dc35af121542a76b9a84f8d1cd886ea97c84703eaa6d88e37a2ad28"},
{file = "PyYAML-6.0.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:40df9b996c2b73138957fe23a16a4f0ba614f4c0efce1e9406a184b6d07fa3a9"},
{file = "PyYAML-6.0.1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a08c6f0fe150303c1c6b71ebcd7213c2858041a7e01975da3a99aed1e7a378ef"},
{file = "PyYAML-6.0.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6c22bec3fbe2524cde73d7ada88f6566758a8f7227bfbf93a408a9d86bcc12a0"},
{file = "PyYAML-6.0.1-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:8d4e9c88387b0f5c7d5f281e55304de64cf7f9c0021a3525bd3b1c542da3b0e4"},
{file = "PyYAML-6.0.1-cp312-cp312-win32.whl", hash = "sha256:d483d2cdf104e7c9fa60c544d92981f12ad66a457afae824d146093b8c294c54"},
@@ -2708,13 +2702,13 @@ files = [
[[package]]
name = "requests"
version = "2.32.3"
version = "2.32.0"
description = "Python HTTP for Humans."
optional = false
python-versions = ">=3.8"
files = [
{file = "requests-2.32.3-py3-none-any.whl", hash = "sha256:70761cfe03c773ceb22aa2f671b4757976145175cdfca038c02654d061d6dcc6"},
{file = "requests-2.32.3.tar.gz", hash = "sha256:55365417734eb18255590a9ff9eb97e9e1da868d4ccd6402399eaf68af20a760"},
{file = "requests-2.32.0-py3-none-any.whl", hash = "sha256:f2c3881dddb70d056c5bd7600a4fae312b2a300e39be6a118d30b90bd27262b5"},
{file = "requests-2.32.0.tar.gz", hash = "sha256:fa5490319474c82ef1d2c9bc459d3652e3ae4ef4c4ebdd18a21145a47ca4b6b8"},
]
[package.dependencies]
@@ -3137,6 +3131,16 @@ files = [
{file = "wrapt-1.14.1-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:8ad85f7f4e20964db4daadcab70b47ab05c7c1cf2a7c1e51087bfaa83831854c"},
{file = "wrapt-1.14.1-cp310-cp310-win32.whl", hash = "sha256:a9a52172be0b5aae932bef82a79ec0a0ce87288c7d132946d645eba03f0ad8a8"},
{file = "wrapt-1.14.1-cp310-cp310-win_amd64.whl", hash = "sha256:6d323e1554b3d22cfc03cd3243b5bb815a51f5249fdcbb86fda4bf62bab9e164"},
{file = "wrapt-1.14.1-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:ecee4132c6cd2ce5308e21672015ddfed1ff975ad0ac8d27168ea82e71413f55"},
{file = "wrapt-1.14.1-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:2020f391008ef874c6d9e208b24f28e31bcb85ccff4f335f15a3251d222b92d9"},
{file = "wrapt-1.14.1-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2feecf86e1f7a86517cab34ae6c2f081fd2d0dac860cb0c0ded96d799d20b335"},
{file = "wrapt-1.14.1-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:240b1686f38ae665d1b15475966fe0472f78e71b1b4903c143a842659c8e4cb9"},
{file = "wrapt-1.14.1-cp311-cp311-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a9008dad07d71f68487c91e96579c8567c98ca4c3881b9b113bc7b33e9fd78b8"},
{file = "wrapt-1.14.1-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:6447e9f3ba72f8e2b985a1da758767698efa72723d5b59accefd716e9e8272bf"},
{file = "wrapt-1.14.1-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:acae32e13a4153809db37405f5eba5bac5fbe2e2ba61ab227926a22901051c0a"},
{file = "wrapt-1.14.1-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:49ef582b7a1152ae2766557f0550a9fcbf7bbd76f43fbdc94dd3bf07cc7168be"},
{file = "wrapt-1.14.1-cp311-cp311-win32.whl", hash = "sha256:358fe87cc899c6bb0ddc185bf3dbfa4ba646f05b1b0b9b5a27c2cb92c2cea204"},
{file = "wrapt-1.14.1-cp311-cp311-win_amd64.whl", hash = "sha256:26046cd03936ae745a502abf44dac702a5e6880b2b01c29aea8ddf3353b68224"},
{file = "wrapt-1.14.1-cp35-cp35m-manylinux1_i686.whl", hash = "sha256:43ca3bbbe97af00f49efb06e352eae40434ca9d915906f77def219b88e85d907"},
{file = "wrapt-1.14.1-cp35-cp35m-manylinux1_x86_64.whl", hash = "sha256:6b1a564e6cb69922c7fe3a678b9f9a3c54e72b469875aa8018f18b4d1dd1adf3"},
{file = "wrapt-1.14.1-cp35-cp35m-manylinux2010_i686.whl", hash = "sha256:00b6d4ea20a906c0ca56d84f93065b398ab74b927a7a3dbd470f6fc503f95dc3"},
@@ -3374,4 +3378,4 @@ cffi = ["cffi (>=1.11)"]
[metadata]
lock-version = "2.0"
python-versions = "^3.9"
content-hash = "9055b73352f1534f664cd8af6ebf8d93cf3bf857f115756f312ff2e3ae1bbbc1"
content-hash = "c09bcb333ab550958b33dbf4fec968c500d8e701fd4c96402cddbd9bb8048055"

View File

@@ -38,7 +38,7 @@ hostname.workspace = true
http.workspace = true
humantime.workspace = true
humantime-serde.workspace = true
hyper0.workspace = true
hyper.workspace = true
hyper1 = { package = "hyper", version = "1.2", features = ["server"] }
hyper-util = { version = "0.1", features = ["server", "http1", "http2", "tokio"] }
http-body-util = { version = "0.1" }
@@ -82,6 +82,7 @@ tokio-postgres-rustls.workspace = true
tokio-rustls.workspace = true
tokio-util.workspace = true
tokio = { workspace = true, features = ["signal"] }
tracing-opentelemetry.workspace = true
tracing-subscriber.workspace = true
tracing-utils.workspace = true
tracing.workspace = true

View File

@@ -73,9 +73,6 @@ pub(crate) enum AuthErrorImpl {
#[error("Authentication timed out")]
UserTimeout(Elapsed),
#[error("Disconnected due to inactivity after {0}.")]
ConfirmationTimeout(humantime::Duration),
}
#[derive(Debug, Error)]
@@ -106,10 +103,6 @@ impl AuthError {
pub(crate) fn user_timeout(elapsed: Elapsed) -> Self {
AuthErrorImpl::UserTimeout(elapsed).into()
}
pub(crate) fn confirmation_timeout(timeout: humantime::Duration) -> Self {
AuthErrorImpl::ConfirmationTimeout(timeout).into()
}
}
impl<E: Into<AuthErrorImpl>> From<E> for AuthError {
@@ -132,7 +125,6 @@ impl UserFacingError for AuthError {
AuthErrorImpl::IpAddressNotAllowed(_) => self.to_string(),
AuthErrorImpl::TooManyConnections => self.to_string(),
AuthErrorImpl::UserTimeout(_) => self.to_string(),
AuthErrorImpl::ConfirmationTimeout(_) => self.to_string(),
}
}
}
@@ -151,7 +143,6 @@ impl ReportableError for AuthError {
AuthErrorImpl::IpAddressNotAllowed(_) => crate::error::ErrorKind::User,
AuthErrorImpl::TooManyConnections => crate::error::ErrorKind::RateLimit,
AuthErrorImpl::UserTimeout(_) => crate::error::ErrorKind::User,
AuthErrorImpl::ConfirmationTimeout(_) => crate::error::ErrorKind::User,
}
}
}

View File

@@ -565,7 +565,7 @@ mod tests {
stream::{PqStream, Stream},
};
use super::{auth_quirks, jwt::JwkCache, AuthRateLimiter};
use super::{auth_quirks, AuthRateLimiter};
struct Auth {
ips: Vec<IpPattern>,
@@ -611,16 +611,12 @@ mod tests {
}
static CONFIG: Lazy<AuthenticationConfig> = Lazy::new(|| AuthenticationConfig {
jwks_cache: JwkCache::default(),
thread_pool: ThreadPool::new(1),
scram_protocol_timeout: std::time::Duration::from_secs(5),
rate_limiter_enabled: true,
rate_limiter: AuthRateLimiter::new(&RateBucketInfo::DEFAULT_AUTH_SET),
rate_limit_ip_subnet: 64,
ip_allowlist_check_enabled: true,
is_auth_broker: false,
accept_jwts: false,
webauth_confirmation_timeout: std::time::Duration::from_secs(5),
});
async fn read_message(r: &mut (impl AsyncRead + Unpin), b: &mut BytesMut) -> PgMessage {

View File

@@ -8,7 +8,7 @@ use anyhow::{bail, ensure, Context};
use arc_swap::ArcSwapOption;
use dashmap::DashMap;
use jose_jwk::crypto::KeyInfo;
use serde::{de::Visitor, Deserialize, Deserializer};
use serde::{Deserialize, Deserializer};
use signature::Verifier;
use tokio::time::Instant;
@@ -261,6 +261,10 @@ impl JwkCacheEntryLock {
let sig = base64::decode_config(signature, base64::URL_SAFE_NO_PAD)
.context("Provided authentication token is not a valid JWT encoding")?;
ensure!(
header.typ == "JWT",
"Provided authentication token is not a valid JWT encoding"
);
let kid = header.key_id.context("missing key id")?;
let mut guard = self
@@ -295,7 +299,7 @@ impl JwkCacheEntryLock {
verify_ec_signature(header_payload.as_bytes(), &sig, key)?;
}
jose_jwk::Key::Rsa(key) => {
verify_rsa_signature(header_payload.as_bytes(), &sig, key, &header.algorithm)?;
verify_rsa_signature(header_payload.as_bytes(), &sig, key, &jwk.prm.alg)?;
}
key => bail!("unsupported key type {key:?}"),
};
@@ -307,11 +311,13 @@ impl JwkCacheEntryLock {
tracing::debug!(?payload, "JWT signature valid with claims");
if let Some(aud) = expected_audience {
ensure!(
payload.audience.0.iter().any(|s| s == aud),
"invalid JWT token audience"
);
match (expected_audience, payload.audience) {
// check the audience matches
(Some(aud1), Some(aud2)) => ensure!(aud1 == aud2, "invalid JWT token audience"),
// the audience is expected but is missing
(Some(_), None) => bail!("invalid JWT token audience"),
// we don't care for the audience field
(None, _) => {}
}
let now = SystemTime::now();
@@ -377,7 +383,7 @@ fn verify_rsa_signature(
data: &[u8],
sig: &[u8],
key: &jose_jwk::Rsa,
alg: &jose_jwa::Algorithm,
alg: &Option<jose_jwa::Algorithm>,
) -> anyhow::Result<()> {
use jose_jwa::{Algorithm, Signing};
use rsa::{
@@ -388,7 +394,7 @@ fn verify_rsa_signature(
let key = RsaPublicKey::try_from(key).map_err(|_| anyhow::anyhow!("invalid RSA key"))?;
match alg {
Algorithm::Signing(Signing::Rs256) => {
Some(Algorithm::Signing(Signing::Rs256)) => {
let key = VerifyingKey::<sha2::Sha256>::new(key);
let sig = Signature::try_from(sig)?;
key.verify(data, &sig)?;
@@ -402,6 +408,9 @@ fn verify_rsa_signature(
/// <https://datatracker.ietf.org/doc/html/rfc7515#section-4.1>
#[derive(serde::Deserialize, serde::Serialize)]
struct JwtHeader<'a> {
/// must be "JWT"
#[serde(rename = "typ")]
typ: &'a str,
/// must be a supported alg
#[serde(rename = "alg")]
algorithm: jose_jwa::Algorithm,
@@ -411,12 +420,11 @@ struct JwtHeader<'a> {
}
/// <https://datatracker.ietf.org/doc/html/rfc7519#section-4.1>
#[derive(serde::Deserialize, Debug)]
#[allow(dead_code)]
#[derive(serde::Deserialize, serde::Serialize, Debug)]
struct JwtPayload<'a> {
/// Audience - Recipient for which the JWT is intended
#[serde(rename = "aud", default)]
audience: OneOrMany,
#[serde(rename = "aud")]
audience: Option<&'a str>,
/// Expiration - Time after which the JWT expires
#[serde(deserialize_with = "numeric_date_opt", rename = "exp", default)]
expiration: Option<SystemTime>,
@@ -439,59 +447,6 @@ struct JwtPayload<'a> {
session_id: Option<&'a str>,
}
/// `OneOrMany` supports parsing either a single item or an array of items.
///
/// Needed for <https://datatracker.ietf.org/doc/html/rfc7519#section-4.1.3>
///
/// > The "aud" (audience) claim identifies the recipients that the JWT is
/// > intended for. Each principal intended to process the JWT MUST
/// > identify itself with a value in the audience claim. If the principal
/// > processing the claim does not identify itself with a value in the
/// > "aud" claim when this claim is present, then the JWT MUST be
/// > rejected. In the general case, the "aud" value is **an array of case-
/// > sensitive strings**, each containing a StringOrURI value. In the
/// > special case when the JWT has one audience, the "aud" value MAY be a
/// > **single case-sensitive string** containing a StringOrURI value. The
/// > interpretation of audience values is generally application specific.
/// > Use of this claim is OPTIONAL.
#[derive(Default, Debug)]
struct OneOrMany(Vec<String>);
impl<'de> Deserialize<'de> for OneOrMany {
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
where
D: Deserializer<'de>,
{
struct OneOrManyVisitor;
impl<'de> Visitor<'de> for OneOrManyVisitor {
type Value = OneOrMany;
fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result {
formatter.write_str("a single string or an array of strings")
}
fn visit_str<E>(self, v: &str) -> Result<Self::Value, E>
where
E: serde::de::Error,
{
Ok(OneOrMany(vec![v.to_owned()]))
}
fn visit_seq<A>(self, mut seq: A) -> Result<Self::Value, A::Error>
where
A: serde::de::SeqAccess<'de>,
{
let mut v = vec![];
while let Some(s) = seq.next_element()? {
v.push(s);
}
Ok(OneOrMany(v))
}
}
deserializer.deserialize_any(OneOrManyVisitor)
}
}
fn numeric_date_opt<'de, D: Deserializer<'de>>(d: D) -> Result<Option<SystemTime>, D::Error> {
let d = <Option<u64>>::deserialize(d)?;
Ok(d.map(|n| SystemTime::UNIX_EPOCH + Duration::from_secs(n)))
@@ -585,6 +540,7 @@ mod tests {
key: jose_jwk::Key::Ec(pk),
prm: jose_jwk::Parameters {
kid: Some(kid),
alg: Some(jose_jwa::Algorithm::Signing(jose_jwa::Signing::Es256)),
..Default::default()
},
};
@@ -598,6 +554,7 @@ mod tests {
key: jose_jwk::Key::Rsa(pk),
prm: jose_jwk::Parameters {
kid: Some(kid),
alg: Some(jose_jwa::Algorithm::Signing(jose_jwa::Signing::Rs256)),
..Default::default()
},
};
@@ -606,6 +563,7 @@ mod tests {
fn build_jwt_payload(kid: String, sig: jose_jwa::Signing) -> String {
let header = JwtHeader {
typ: "JWT",
algorithm: jose_jwa::Algorithm::Signing(sig),
key_id: Some(&kid),
};

View File

@@ -14,15 +14,17 @@ use crate::{
EndpointId,
};
use super::jwt::{AuthRule, FetchAuthRules};
use super::jwt::{AuthRule, FetchAuthRules, JwkCache};
pub struct LocalBackend {
pub(crate) jwks_cache: JwkCache,
pub(crate) node_info: NodeInfo,
}
impl LocalBackend {
pub fn new(postgres_addr: SocketAddr) -> Self {
LocalBackend {
jwks_cache: JwkCache::default(),
node_info: NodeInfo {
config: {
let mut cfg = ConnCfg::new();

View File

@@ -89,12 +89,7 @@ pub(super) async fn authenticate(
// Wait for web console response (see `mgmt`).
info!(parent: &span, "waiting for console's reply...");
let db_info = tokio::time::timeout(auth_config.webauth_confirmation_timeout, waiter)
.await
.map_err(|_elapsed| {
auth::AuthError::confirmation_timeout(auth_config.webauth_confirmation_timeout.into())
})?
.map_err(WebAuthError::from)?;
let db_info = waiter.await.map_err(WebAuthError::from)?;
if auth_config.ip_allowlist_check_enabled {
if let Some(allowed_ips) = &db_info.allowed_ips {

View File

@@ -6,10 +6,7 @@ use compute_api::spec::LocalProxySpec;
use dashmap::DashMap;
use futures::future::Either;
use proxy::{
auth::backend::{
jwt::JwkCache,
local::{LocalBackend, JWKS_ROLE_MAP},
},
auth::backend::local::{LocalBackend, JWKS_ROLE_MAP},
cancellation::CancellationHandlerMain,
config::{self, AuthenticationConfig, HttpConfig, ProxyConfig, RetryConfig},
console::{
@@ -77,10 +74,10 @@ struct LocalProxyCliArgs {
#[clap(long, default_value = "127.0.0.1:5432")]
compute: SocketAddr,
/// Path of the local proxy config file
#[clap(long, default_value = "./local_proxy.json")]
#[clap(long, default_value = "./localproxy.json")]
config_path: Utf8PathBuf,
/// Path of the local proxy PID file
#[clap(long, default_value = "./local_proxy.pid")]
#[clap(long, default_value = "./localproxy.pid")]
pid_path: Utf8PathBuf,
}
@@ -109,7 +106,7 @@ struct SqlOverHttpArgs {
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let _logging_guard = proxy::logging::init_local_proxy()?;
let _logging_guard = proxy::logging::init().await?;
let _panic_hook_guard = utils::logging::replace_panic_hook_with_tracing_panic_hook();
let _sentry_guard = init_sentry(Some(GIT_VERSION.into()), &[]);
@@ -138,7 +135,7 @@ async fn main() -> anyhow::Result<()> {
// in order to trigger the appropriate SIGHUP on config change.
//
// This also claims a "lock" that makes sure only one instance
// of local_proxy runs at a time.
// of local-proxy runs at a time.
let _process_guard = loop {
match pid_file::claim_for_current_process(&args.pid_path) {
Ok(guard) => break guard,
@@ -164,6 +161,12 @@ async fn main() -> anyhow::Result<()> {
16,
));
// write the process ID to a file so that compute-ctl can find our process later
// in order to trigger the appropriate SIGHUP on config change.
let pid = std::process::id();
info!("process running in PID {pid}");
std::fs::write(args.pid_path, format!("{pid}\n")).context("writing PID to file")?;
let mut maintenance_tasks = JoinSet::new();
let refresh_config_notify = Arc::new(Notify::new());
@@ -176,9 +179,9 @@ async fn main() -> anyhow::Result<()> {
// trigger the first config load **after** setting up the signal hook
// to avoid the race condition where:
// 1. No config file registered when local_proxy starts up
// 1. No config file registered when local-proxy starts up
// 2. The config file is written but the signal hook is not yet received
// 3. local_proxy completes startup but has no config loaded, despite there being a registerd config.
// 3. local-proxy completes startup but has no config loaded, despite there being a registerd config.
refresh_config_notify.notify_one();
tokio::spawn(refresh_config_loop(args.config_path, refresh_config_notify));
@@ -264,18 +267,14 @@ fn build_config(args: &LocalProxyCliArgs) -> anyhow::Result<&'static ProxyConfig
allow_self_signed_compute: false,
http_config,
authentication_config: AuthenticationConfig {
jwks_cache: JwkCache::default(),
thread_pool: ThreadPool::new(0),
scram_protocol_timeout: Duration::from_secs(10),
rate_limiter_enabled: false,
rate_limiter: BucketRateLimiter::new(vec![]),
rate_limit_ip_subnet: 64,
ip_allowlist_check_enabled: true,
is_auth_broker: false,
accept_jwts: true,
webauth_confirmation_timeout: Duration::ZERO,
},
proxy_protocol_v2: config::ProxyProtocolV2::Rejected,
require_client_ip: false,
handshake_timeout: Duration::from_secs(10),
region: "local".into(),
wake_compute_retry_config: RetryConfig::parse(RetryConfig::WAKE_COMPUTE_DEFAULT_VALUES)?,
@@ -305,7 +304,7 @@ async fn refresh_config_inner(path: &Utf8Path) -> anyhow::Result<()> {
let mut jwks_set = vec![];
for jwks in data.jwks.into_iter().flatten() {
for jwks in data.jwks {
let mut jwks_url = url::Url::from_str(&jwks.jwks_url).context("parsing JWKS url")?;
ensure!(

View File

@@ -8,7 +8,6 @@ use aws_config::web_identity_token::WebIdentityTokenCredentialsProvider;
use aws_config::Region;
use futures::future::Either;
use proxy::auth;
use proxy::auth::backend::jwt::JwkCache;
use proxy::auth::backend::AuthRateLimiter;
use proxy::auth::backend::MaybeOwned;
use proxy::cancellation::CancelMap;
@@ -18,7 +17,6 @@ use proxy::config::AuthenticationConfig;
use proxy::config::CacheOptions;
use proxy::config::HttpConfig;
use proxy::config::ProjectInfoCacheOptions;
use proxy::config::ProxyProtocolV2;
use proxy::console;
use proxy::context::parquet::ParquetUploadArgs;
use proxy::http;
@@ -104,9 +102,6 @@ struct ProxyCliArgs {
default_value = "http://localhost:3000/authenticate_proxy_request/"
)]
auth_endpoint: String,
/// if this is not local proxy, this toggles whether we accept jwt or passwords for http
#[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
is_auth_broker: bool,
/// path to TLS key for client postgres connections
///
/// tls-key and tls-cert are for backwards compatibility, we can put all certs in one dir
@@ -149,6 +144,9 @@ struct ProxyCliArgs {
/// size of the threadpool for password hashing
#[clap(long, default_value_t = 4)]
scram_thread_pool_size: u8,
/// Require that all incoming requests have a Proxy Protocol V2 packet **and** have an IP address associated.
#[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
require_client_ip: bool,
/// Disable dynamic rate limiter and store the metrics to ensure its production behaviour.
#[clap(long, default_value_t = true, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
disable_dynamic_rate_limiter: bool,
@@ -231,15 +229,6 @@ struct ProxyCliArgs {
/// Configure if this is a private access proxy for the POC: In that case the proxy will ignore the IP allowlist
#[clap(long, default_value_t = false, value_parser = clap::builder::BoolishValueParser::new(), action = clap::ArgAction::Set)]
is_private_access_proxy: bool,
/// Configure whether all incoming requests have a Proxy Protocol V2 packet.
// TODO(conradludgate): switch default to rejected or required once we've updated all deployments
#[clap(value_enum, long, default_value_t = ProxyProtocolV2::Supported)]
proxy_protocol_v2: ProxyProtocolV2,
/// Time the proxy waits for the webauth session to be confirmed by the control plane.
#[clap(long, default_value = "2m", value_parser = humantime::parse_duration)]
webauth_confirmation_timeout: std::time::Duration,
}
#[derive(clap::Args, Clone, Copy, Debug)]
@@ -393,27 +382,9 @@ async fn main() -> anyhow::Result<()> {
info!("Starting mgmt on {mgmt_address}");
let mgmt_listener = TcpListener::bind(mgmt_address).await?;
let proxy_listener = if !args.is_auth_broker {
let proxy_address: SocketAddr = args.proxy.parse()?;
info!("Starting proxy on {proxy_address}");
Some(TcpListener::bind(proxy_address).await?)
} else {
None
};
// TODO: rename the argument to something like serverless.
// It now covers more than just websockets, it also covers SQL over HTTP.
let serverless_listener = if let Some(serverless_address) = args.wss {
let serverless_address: SocketAddr = serverless_address.parse()?;
info!("Starting wss on {serverless_address}");
Some(TcpListener::bind(serverless_address).await?)
} else if args.is_auth_broker {
bail!("wss arg must be present for auth-broker")
} else {
None
};
let proxy_address: SocketAddr = args.proxy.parse()?;
info!("Starting proxy on {proxy_address}");
let proxy_listener = TcpListener::bind(proxy_address).await?;
let cancellation_token = CancellationToken::new();
let cancel_map = CancelMap::default();
@@ -459,17 +430,21 @@ async fn main() -> anyhow::Result<()> {
// client facing tasks. these will exit on error or on cancellation
// cancellation returns Ok(())
let mut client_tasks = JoinSet::new();
if let Some(proxy_listener) = proxy_listener {
client_tasks.spawn(proxy::proxy::task_main(
config,
proxy_listener,
cancellation_token.clone(),
cancellation_handler.clone(),
endpoint_rate_limiter.clone(),
));
}
client_tasks.spawn(proxy::proxy::task_main(
config,
proxy_listener,
cancellation_token.clone(),
cancellation_handler.clone(),
endpoint_rate_limiter.clone(),
));
// TODO: rename the argument to something like serverless.
// It now covers more than just websockets, it also covers SQL over HTTP.
if let Some(serverless_address) = args.wss {
let serverless_address: SocketAddr = serverless_address.parse()?;
info!("Starting wss on {serverless_address}");
let serverless_listener = TcpListener::bind(serverless_address).await?;
if let Some(serverless_listener) = serverless_listener {
client_tasks.spawn(serverless::task_main(
config,
serverless_listener,
@@ -699,7 +674,7 @@ fn build_config(args: &ProxyCliArgs) -> anyhow::Result<&'static ProxyConfig> {
)?;
let http_config = HttpConfig {
accept_websockets: !args.is_auth_broker,
accept_websockets: true,
pool_options: GlobalConnPoolOptions {
max_conns_per_endpoint: args.sql_over_http.sql_over_http_pool_max_conns_per_endpoint,
gc_epoch: args.sql_over_http.sql_over_http_pool_gc_epoch,
@@ -714,16 +689,12 @@ fn build_config(args: &ProxyCliArgs) -> anyhow::Result<&'static ProxyConfig> {
max_response_size_bytes: args.sql_over_http.sql_over_http_max_response_size_bytes,
};
let authentication_config = AuthenticationConfig {
jwks_cache: JwkCache::default(),
thread_pool,
scram_protocol_timeout: args.scram_protocol_timeout,
rate_limiter_enabled: args.auth_rate_limit_enabled,
rate_limiter: AuthRateLimiter::new(args.auth_rate_limit.clone()),
rate_limit_ip_subnet: args.auth_rate_limit_ip_subnet,
ip_allowlist_check_enabled: !args.is_private_access_proxy,
is_auth_broker: args.is_auth_broker,
accept_jwts: args.is_auth_broker,
webauth_confirmation_timeout: args.webauth_confirmation_timeout,
};
let config = Box::leak(Box::new(ProxyConfig {
@@ -733,7 +704,7 @@ fn build_config(args: &ProxyCliArgs) -> anyhow::Result<&'static ProxyConfig> {
allow_self_signed_compute: args.allow_self_signed_compute,
http_config,
authentication_config,
proxy_protocol_v2: args.proxy_protocol_v2,
require_client_ip: args.require_client_ip,
handshake_timeout: args.handshake_timeout,
region: args.region.clone(),
wake_compute_retry_config: config::RetryConfig::parse(&args.wake_compute_retry)?,

View File

@@ -20,7 +20,7 @@ use tokio_postgres::tls::MakeTlsConnect;
use tokio_postgres_rustls::MakeRustlsConnect;
use tracing::{error, info, warn};
pub const COULD_NOT_CONNECT: &str = "Couldn't connect to compute node";
const COULD_NOT_CONNECT: &str = "Couldn't connect to compute node";
#[derive(Debug, Error)]
pub(crate) enum ConnectionError {

View File

@@ -1,8 +1,5 @@
use crate::{
auth::{
self,
backend::{jwt::JwkCache, AuthRateLimiter},
},
auth::{self, backend::AuthRateLimiter},
console::locks::ApiLocks,
rate_limiter::{RateBucketInfo, RateLimitAlgorithm, RateLimiterConfig},
scram::threadpool::ThreadPool,
@@ -10,7 +7,6 @@ use crate::{
Host,
};
use anyhow::{bail, ensure, Context, Ok};
use clap::ValueEnum;
use itertools::Itertools;
use remote_storage::RemoteStorageConfig;
use rustls::{
@@ -34,7 +30,7 @@ pub struct ProxyConfig {
pub allow_self_signed_compute: bool,
pub http_config: HttpConfig,
pub authentication_config: AuthenticationConfig,
pub proxy_protocol_v2: ProxyProtocolV2,
pub require_client_ip: bool,
pub region: String,
pub handshake_timeout: Duration,
pub wake_compute_retry_config: RetryConfig,
@@ -42,16 +38,6 @@ pub struct ProxyConfig {
pub connect_to_compute_retry_config: RetryConfig,
}
#[derive(Copy, Clone, Debug, ValueEnum, PartialEq)]
pub enum ProxyProtocolV2 {
/// Connection will error if PROXY protocol v2 header is missing
Required,
/// Connection will parse PROXY protocol v2 header, but accept the connection if it's missing.
Supported,
/// Connection will error if PROXY protocol v2 header is provided
Rejected,
}
#[derive(Debug)]
pub struct MetricCollectionConfig {
pub endpoint: reqwest::Url,
@@ -81,10 +67,6 @@ pub struct AuthenticationConfig {
pub rate_limiter: AuthRateLimiter,
pub rate_limit_ip_subnet: u8,
pub ip_allowlist_check_enabled: bool,
pub jwks_cache: JwkCache,
pub is_auth_broker: bool,
pub accept_jwts: bool,
pub webauth_confirmation_timeout: tokio::time::Duration,
}
impl TlsConfig {
@@ -268,26 +250,18 @@ impl CertResolver {
let common_name = pem.subject().to_string();
// We need to get the canonical name for this certificate so we can match them against any domain names
// seen within the proxy codebase.
//
// In scram-proxy we use wildcard certificates only, with the database endpoint as the wildcard subdomain, taken from SNI.
// We need to remove the wildcard prefix for the purposes of certificate selection.
//
// auth-broker does not use SNI and instead uses the Neon-Connection-String header.
// Auth broker has the subdomain `apiauth` we need to remove for the purposes of validating the Neon-Connection-String.
//
// Console Web proxy does not use any wildcard domains and does not need any certificate selection or conn string
// validation, so let's we can continue with any common-name
let common_name = if let Some(s) = common_name.strip_prefix("CN=*.") {
s.to_string()
} else if let Some(s) = common_name.strip_prefix("CN=apiauth.") {
s.to_string()
} else if let Some(s) = common_name.strip_prefix("CN=") {
s.to_string()
// We only use non-wildcard certificates in web auth proxy so it seems okay to treat them the same as
// wildcard ones as we don't use SNI there. That treatment only affects certificate selection, so
// verify-full will still check wildcard match. Old coding here just ignored non-wildcard common names
// and passed None instead, which blows up number of cases downstream code should handle. Proper coding
// here should better avoid Option for common_names, and do wildcard-based certificate selection instead
// of cutting off '*.' parts.
let common_name = if common_name.starts_with("CN=*.") {
common_name.strip_prefix("CN=*.").map(|s| s.to_string())
} else {
bail!("Failed to parse common name from certificate")
};
common_name.strip_prefix("CN=").map(|s| s.to_string())
}
.context("Failed to parse common name from certificate")?;
let cert = Arc::new(rustls::sign::CertifiedKey::new(cert_chain, key));

View File

@@ -90,8 +90,6 @@ use tokio::task::JoinError;
use tokio_util::sync::CancellationToken;
use tracing::warn;
extern crate hyper0 as hyper;
pub mod auth;
pub mod cache;
pub mod cancellation;

View File

@@ -1,13 +1,7 @@
use tracing::Subscriber;
use tracing_opentelemetry::OpenTelemetryLayer;
use tracing_subscriber::{
filter::{EnvFilter, LevelFilter},
fmt::{
format::{Format, Full},
time::SystemTime,
FormatEvent, FormatFields,
},
prelude::*,
registry::LookupSpan,
};
/// Initialize logging and OpenTelemetry tracing and exporter.
@@ -29,7 +23,9 @@ pub async fn init() -> anyhow::Result<LoggingGuard> {
.with_writer(std::io::stderr)
.with_target(false);
let otlp_layer = tracing_utils::init_tracing("proxy").await;
let otlp_layer = tracing_utils::init_tracing("proxy")
.await
.map(OpenTelemetryLayer::new);
tracing_subscriber::registry()
.with(env_filter)
@@ -40,45 +36,6 @@ pub async fn init() -> anyhow::Result<LoggingGuard> {
Ok(LoggingGuard)
}
/// Initialize logging for local_proxy with log prefix and no opentelemetry.
///
/// Logging can be configured using `RUST_LOG` environment variable.
pub fn init_local_proxy() -> anyhow::Result<LoggingGuard> {
let env_filter = EnvFilter::builder()
.with_default_directive(LevelFilter::INFO.into())
.from_env_lossy();
let fmt_layer = tracing_subscriber::fmt::layer()
.with_ansi(false)
.with_writer(std::io::stderr)
.event_format(LocalProxyFormatter(Format::default().with_target(false)));
tracing_subscriber::registry()
.with(env_filter)
.with(fmt_layer)
.try_init()?;
Ok(LoggingGuard)
}
pub struct LocalProxyFormatter(Format<Full, SystemTime>);
impl<S, N> FormatEvent<S, N> for LocalProxyFormatter
where
S: Subscriber + for<'a> LookupSpan<'a>,
N: for<'a> FormatFields<'a> + 'static,
{
fn format_event(
&self,
ctx: &tracing_subscriber::fmt::FmtContext<'_, S, N>,
mut writer: tracing_subscriber::fmt::format::Writer<'_>,
event: &tracing::Event<'_>,
) -> std::fmt::Result {
writer.write_str("[local_proxy] ")?;
self.0.format_event(ctx, writer, event)
}
}
pub struct LoggingGuard;
impl Drop for LoggingGuard {

View File

@@ -10,7 +10,6 @@ pub(crate) mod wake_compute;
pub use copy_bidirectional::copy_bidirectional_client_compute;
pub use copy_bidirectional::ErrorSource;
use crate::config::ProxyProtocolV2;
use crate::{
auth,
cancellation::{self, CancellationHandlerMain, CancellationHandlerMainInternal},
@@ -94,19 +93,15 @@ pub async fn task_main(
connections.spawn(async move {
let (socket, peer_addr) = match read_proxy_protocol(socket).await {
Ok((socket, Some(addr))) => (socket, addr.ip()),
Err(e) => {
error!("per-client task finished with an error: {e:#}");
return;
}
Ok((_socket, None)) if config.proxy_protocol_v2 == ProxyProtocolV2::Required => {
error!("missing required proxy protocol header");
Ok((_socket, None)) if config.require_client_ip => {
error!("missing required client IP");
return;
}
Ok((_socket, Some(_))) if config.proxy_protocol_v2 == ProxyProtocolV2::Rejected => {
error!("proxy protocol header not supported");
return;
}
Ok((socket, Some(addr))) => (socket, addr.ip()),
Ok((socket, None)) => (socket, peer_addr.ip()),
};

View File

@@ -1,6 +1,5 @@
use crate::{
auth::backend::ComputeCredentialKeys,
compute::COULD_NOT_CONNECT,
compute::{self, PostgresConnection},
config::RetryConfig,
console::{self, errors::WakeComputeError, locks::ApiLocks, CachedNodeInfo, NodeInfo},
@@ -16,7 +15,7 @@ use crate::{
use async_trait::async_trait;
use pq_proto::StartupMessageParams;
use tokio::time;
use tracing::{debug, info, warn};
use tracing::{error, info, warn};
use super::retry::ShouldRetryWakeCompute;
@@ -117,6 +116,7 @@ where
node_info.set_keys(user_info.get_keys());
node_info.allow_self_signed_compute = allow_self_signed_compute;
// let mut node_info = credentials.get_node_info(ctx, user_info).await?;
mechanism.update_connect_config(&mut node_info.config);
let retry_type = RetryType::ConnectToCompute;
@@ -139,10 +139,10 @@ where
Err(e) => e,
};
debug!(error = ?err, COULD_NOT_CONNECT);
error!(error = ?err, "could not connect to compute node");
let node_info = if !node_info.cached() || !err.should_retry_wake_compute() {
// If we just recieved this from cplane and didn't get it from cache, we shouldn't retry.
// If we just recieved this from cplane and dodn't get it from cache, we shouldn't retry.
// Do not need to retrieve a new node_info, just return the old one.
if should_retry(&err, num_retries, connect_to_compute_retry_config) {
Metrics::get().proxy.retries_metric.observe(
@@ -191,7 +191,7 @@ where
}
Err(e) => {
if !should_retry(&e, num_retries, connect_to_compute_retry_config) {
// Don't log an error here, caller will print the error
error!(error = ?e, num_retries, retriable = false, "couldn't connect to compute node");
Metrics::get().proxy.retries_metric.observe(
RetriesMetricGroup {
outcome: ConnectOutcome::Failed,
@@ -202,7 +202,7 @@ where
return Err(e.into());
}
warn!(error = ?e, num_retries, retriable = true, COULD_NOT_CONNECT);
warn!(error = ?e, num_retries, retriable = true, "couldn't connect to compute node");
}
};

View File

@@ -5,7 +5,6 @@
mod backend;
pub mod cancel_set;
mod conn_pool;
mod http_conn_pool;
mod http_util;
mod json;
mod sql_over_http;
@@ -20,8 +19,7 @@ use anyhow::Context;
use futures::future::{select, Either};
use futures::TryFutureExt;
use http::{Method, Response, StatusCode};
use http_body_util::combinators::BoxBody;
use http_body_util::{BodyExt, Empty};
use http_body_util::Full;
use hyper1::body::Incoming;
use hyper_util::rt::TokioExecutor;
use hyper_util::server::conn::auto::Builder;
@@ -83,28 +81,7 @@ pub async fn task_main(
}
});
let http_conn_pool = http_conn_pool::GlobalConnPool::new(&config.http_config);
{
let http_conn_pool = Arc::clone(&http_conn_pool);
tokio::spawn(async move {
http_conn_pool.gc_worker(StdRng::from_entropy()).await;
});
}
// shutdown the connection pool
tokio::spawn({
let cancellation_token = cancellation_token.clone();
let http_conn_pool = http_conn_pool.clone();
async move {
cancellation_token.cancelled().await;
tokio::task::spawn_blocking(move || http_conn_pool.shutdown())
.await
.unwrap();
}
});
let backend = Arc::new(PoolingBackend {
http_conn_pool: Arc::clone(&http_conn_pool),
pool: Arc::clone(&conn_pool),
config,
endpoint_rate_limiter: Arc::clone(&endpoint_rate_limiter),
@@ -365,7 +342,7 @@ async fn request_handler(
// used to cancel in-flight HTTP requests. not used to cancel websockets
http_cancellation_token: CancellationToken,
endpoint_rate_limiter: Arc<EndpointRateLimiter>,
) -> Result<Response<BoxBody<Bytes, hyper1::Error>>, ApiError> {
) -> Result<Response<Full<Bytes>>, ApiError> {
let host = request
.headers()
.get("host")
@@ -409,7 +386,7 @@ async fn request_handler(
);
// Return the response so the spawned future can continue.
Ok(response.map(|b| b.map_err(|x| match x {}).boxed()))
Ok(response.map(|_: http_body_util::Empty<Bytes>| Full::new(Bytes::new())))
} else if request.uri().path() == "/sql" && *request.method() == Method::POST {
let ctx = RequestMonitoring::new(
session_id,
@@ -432,7 +409,7 @@ async fn request_handler(
)
.header("Access-Control-Max-Age", "86400" /* 24 hours */)
.status(StatusCode::OK) // 204 is also valid, but see: https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods/OPTIONS#status_code
.body(Empty::new().map_err(|x| match x {}).boxed())
.body(Full::new(Bytes::new()))
.map_err(|e| ApiError::InternalServerError(e.into()))
} else {
json_response(StatusCode::BAD_REQUEST, "query is not supported")

View File

@@ -1,8 +1,6 @@
use std::{io, sync::Arc, time::Duration};
use std::{sync::Arc, time::Duration};
use async_trait::async_trait;
use hyper_util::rt::{TokioExecutor, TokioIo, TokioTimer};
use tokio::net::{lookup_host, TcpStream};
use tracing::{field::display, info};
use crate::{
@@ -29,13 +27,9 @@ use crate::{
Host,
};
use super::{
conn_pool::{poll_client, Client, ConnInfo, GlobalConnPool},
http_conn_pool::{self, poll_http2_client},
};
use super::conn_pool::{poll_client, Client, ConnInfo, GlobalConnPool};
pub(crate) struct PoolingBackend {
pub(crate) http_conn_pool: Arc<super::http_conn_pool::GlobalConnPool>,
pub(crate) pool: Arc<GlobalConnPool<tokio_postgres::Client>>,
pub(crate) config: &'static ProxyConfig,
pub(crate) endpoint_rate_limiter: Arc<EndpointRateLimiter>,
@@ -109,44 +103,32 @@ impl PoolingBackend {
pub(crate) async fn authenticate_with_jwt(
&self,
ctx: &RequestMonitoring,
config: &AuthenticationConfig,
user_info: &ComputeUserInfo,
jwt: String,
) -> Result<(), AuthError> {
jwt: &str,
) -> Result<ComputeCredentials, AuthError> {
match &self.config.auth_backend {
crate::auth::Backend::Console(console, ()) => {
config
.jwks_cache
.check_jwt(
ctx,
user_info.endpoint.clone(),
&user_info.user,
&**console,
&jwt,
)
.await
.map_err(|e| AuthError::auth_failed(e.to_string()))?;
Ok(())
crate::auth::Backend::Console(_, ()) => {
Err(AuthError::auth_failed("JWT login is not yet supported"))
}
crate::auth::Backend::Web(_, ()) => Err(AuthError::auth_failed(
"JWT login over web auth proxy is not supported",
)),
crate::auth::Backend::Local(_) => {
config
crate::auth::Backend::Local(cache) => {
cache
.jwks_cache
.check_jwt(
ctx,
user_info.endpoint.clone(),
&user_info.user,
&StaticAuthRules,
&jwt,
jwt,
)
.await
.map_err(|e| AuthError::auth_failed(e.to_string()))?;
// todo: rewrite JWT signature with key shared somehow between local proxy and postgres
Ok(())
Ok(ComputeCredentials {
info: user_info.clone(),
keys: crate::auth::backend::ComputeCredentialKeys::None,
})
}
}
}
@@ -192,55 +174,14 @@ impl PoolingBackend {
)
.await
}
// Wake up the destination if needed
#[tracing::instrument(fields(pid = tracing::field::Empty), skip_all)]
pub(crate) async fn connect_to_local_proxy(
&self,
ctx: &RequestMonitoring,
conn_info: ConnInfo,
) -> Result<http_conn_pool::Client, HttpConnError> {
info!("pool: looking for an existing connection");
if let Some(client) = self.http_conn_pool.get(ctx, &conn_info) {
return Ok(client);
}
let conn_id = uuid::Uuid::new_v4();
tracing::Span::current().record("conn_id", display(conn_id));
info!(%conn_id, "pool: opening a new connection '{conn_info}'");
let backend = self
.config
.auth_backend
.as_ref()
.map(|()| ComputeCredentials {
info: conn_info.user_info.clone(),
keys: crate::auth::backend::ComputeCredentialKeys::None,
});
crate::proxy::connect_compute::connect_to_compute(
ctx,
&HyperMechanism {
conn_id,
conn_info,
pool: self.http_conn_pool.clone(),
locks: &self.config.connect_compute_locks,
},
&backend,
false, // do not allow self signed compute for http flow
self.config.wake_compute_retry_config,
self.config.connect_to_compute_retry_config,
)
.await
}
}
#[derive(Debug, thiserror::Error)]
pub(crate) enum HttpConnError {
#[error("pooled connection closed at inconsistent state")]
ConnectionClosedAbruptly(#[from] tokio::sync::watch::error::SendError<uuid::Uuid>),
#[error("could not connection to postgres in compute")]
PostgresConnectionError(#[from] tokio_postgres::Error),
#[error("could not connection to local-proxy in compute")]
LocalProxyConnectionError(#[from] LocalProxyConnError),
#[error("could not connection to compute")]
ConnectionError(#[from] tokio_postgres::Error),
#[error("could not get auth info")]
GetAuthInfo(#[from] GetAuthInfoError),
@@ -252,20 +193,11 @@ pub(crate) enum HttpConnError {
TooManyConnectionAttempts(#[from] ApiLockError),
}
#[derive(Debug, thiserror::Error)]
pub(crate) enum LocalProxyConnError {
#[error("error with connection to local-proxy")]
Io(#[source] std::io::Error),
#[error("could not establish h2 connection")]
H2(#[from] hyper1::Error),
}
impl ReportableError for HttpConnError {
fn get_error_kind(&self) -> ErrorKind {
match self {
HttpConnError::ConnectionClosedAbruptly(_) => ErrorKind::Compute,
HttpConnError::PostgresConnectionError(p) => p.get_error_kind(),
HttpConnError::LocalProxyConnectionError(_) => ErrorKind::Compute,
HttpConnError::ConnectionError(p) => p.get_error_kind(),
HttpConnError::GetAuthInfo(a) => a.get_error_kind(),
HttpConnError::AuthError(a) => a.get_error_kind(),
HttpConnError::WakeCompute(w) => w.get_error_kind(),
@@ -278,8 +210,7 @@ impl UserFacingError for HttpConnError {
fn to_string_client(&self) -> String {
match self {
HttpConnError::ConnectionClosedAbruptly(_) => self.to_string(),
HttpConnError::PostgresConnectionError(p) => p.to_string(),
HttpConnError::LocalProxyConnectionError(p) => p.to_string(),
HttpConnError::ConnectionError(p) => p.to_string(),
HttpConnError::GetAuthInfo(c) => c.to_string_client(),
HttpConnError::AuthError(c) => c.to_string_client(),
HttpConnError::WakeCompute(c) => c.to_string_client(),
@@ -293,8 +224,7 @@ impl UserFacingError for HttpConnError {
impl CouldRetry for HttpConnError {
fn could_retry(&self) -> bool {
match self {
HttpConnError::PostgresConnectionError(e) => e.could_retry(),
HttpConnError::LocalProxyConnectionError(e) => e.could_retry(),
HttpConnError::ConnectionError(e) => e.could_retry(),
HttpConnError::ConnectionClosedAbruptly(_) => false,
HttpConnError::GetAuthInfo(_) => false,
HttpConnError::AuthError(_) => false,
@@ -306,7 +236,7 @@ impl CouldRetry for HttpConnError {
impl ShouldRetryWakeCompute for HttpConnError {
fn should_retry_wake_compute(&self) -> bool {
match self {
HttpConnError::PostgresConnectionError(e) => e.should_retry_wake_compute(),
HttpConnError::ConnectionError(e) => e.should_retry_wake_compute(),
// we never checked cache validity
HttpConnError::TooManyConnectionAttempts(_) => false,
_ => true,
@@ -314,38 +244,6 @@ impl ShouldRetryWakeCompute for HttpConnError {
}
}
impl ReportableError for LocalProxyConnError {
fn get_error_kind(&self) -> ErrorKind {
match self {
LocalProxyConnError::Io(_) => ErrorKind::Compute,
LocalProxyConnError::H2(_) => ErrorKind::Compute,
}
}
}
impl UserFacingError for LocalProxyConnError {
fn to_string_client(&self) -> String {
"Could not establish HTTP connection to the database".to_string()
}
}
impl CouldRetry for LocalProxyConnError {
fn could_retry(&self) -> bool {
match self {
LocalProxyConnError::Io(_) => false,
LocalProxyConnError::H2(_) => false,
}
}
}
impl ShouldRetryWakeCompute for LocalProxyConnError {
fn should_retry_wake_compute(&self) -> bool {
match self {
LocalProxyConnError::Io(_) => false,
LocalProxyConnError::H2(_) => false,
}
}
}
struct TokioMechanism {
pool: Arc<GlobalConnPool<tokio_postgres::Client>>,
conn_info: ConnInfo,
@@ -395,99 +293,3 @@ impl ConnectMechanism for TokioMechanism {
fn update_connect_config(&self, _config: &mut compute::ConnCfg) {}
}
struct HyperMechanism {
pool: Arc<http_conn_pool::GlobalConnPool>,
conn_info: ConnInfo,
conn_id: uuid::Uuid,
/// connect_to_compute concurrency lock
locks: &'static ApiLocks<Host>,
}
#[async_trait]
impl ConnectMechanism for HyperMechanism {
type Connection = http_conn_pool::Client;
type ConnectError = HttpConnError;
type Error = HttpConnError;
async fn connect_once(
&self,
ctx: &RequestMonitoring,
node_info: &CachedNodeInfo,
timeout: Duration,
) -> Result<Self::Connection, Self::ConnectError> {
let host = node_info.config.get_host()?;
let permit = self.locks.get_permit(&host).await?;
let pause = ctx.latency_timer_pause(crate::metrics::Waiting::Compute);
// let port = node_info.config.get_ports().first().unwrap_or_else(10432);
let res = connect_http2(&host, 10432, timeout).await;
drop(pause);
let (client, connection) = permit.release_result(res)?;
Ok(poll_http2_client(
self.pool.clone(),
ctx,
&self.conn_info,
client,
connection,
self.conn_id,
node_info.aux.clone(),
))
}
fn update_connect_config(&self, _config: &mut compute::ConnCfg) {}
}
async fn connect_http2(
host: &str,
port: u16,
timeout: Duration,
) -> Result<(http_conn_pool::Send, http_conn_pool::Connect), LocalProxyConnError> {
// assumption: host is an ip address so this should not actually perform any requests.
// todo: add that assumption as a guarantee in the control-plane API.
let mut addrs = lookup_host((host, port))
.await
.map_err(LocalProxyConnError::Io)?;
let mut last_err = None;
let stream = loop {
let Some(addr) = addrs.next() else {
return Err(last_err.unwrap_or_else(|| {
LocalProxyConnError::Io(io::Error::new(
io::ErrorKind::InvalidInput,
"could not resolve any addresses",
))
}));
};
match tokio::time::timeout(timeout, TcpStream::connect(addr)).await {
Ok(Ok(stream)) => {
stream.set_nodelay(true).map_err(LocalProxyConnError::Io)?;
break stream;
}
Ok(Err(e)) => {
last_err = Some(LocalProxyConnError::Io(e));
}
Err(e) => {
last_err = Some(LocalProxyConnError::Io(io::Error::new(
io::ErrorKind::TimedOut,
e,
)));
}
};
};
let (client, connection) = hyper1::client::conn::http2::Builder::new(TokioExecutor::new())
.timer(TokioTimer::new())
.keep_alive_interval(Duration::from_secs(20))
.keep_alive_while_idle(true)
.keep_alive_timeout(Duration::from_secs(5))
.handshake(TokioIo::new(stream))
.await?;
Ok((client, connection))
}

View File

@@ -1,342 +0,0 @@
use dashmap::DashMap;
use hyper1::client::conn::http2;
use hyper_util::rt::{TokioExecutor, TokioIo};
use parking_lot::RwLock;
use rand::Rng;
use std::collections::VecDeque;
use std::sync::atomic::{self, AtomicUsize};
use std::{sync::Arc, sync::Weak};
use tokio::net::TcpStream;
use crate::console::messages::{ColdStartInfo, MetricsAuxInfo};
use crate::metrics::{HttpEndpointPoolsGuard, Metrics};
use crate::usage_metrics::{Ids, MetricCounter, USAGE_METRICS};
use crate::{context::RequestMonitoring, EndpointCacheKey};
use tracing::{debug, error};
use tracing::{info, info_span, Instrument};
use super::conn_pool::ConnInfo;
pub(crate) type Send = http2::SendRequest<hyper1::body::Incoming>;
pub(crate) type Connect =
http2::Connection<TokioIo<TcpStream>, hyper1::body::Incoming, TokioExecutor>;
#[derive(Clone)]
struct ConnPoolEntry {
conn: Send,
conn_id: uuid::Uuid,
aux: MetricsAuxInfo,
}
// Per-endpoint connection pool
// Number of open connections is limited by the `max_conns_per_endpoint`.
pub(crate) struct EndpointConnPool {
// TODO(conrad):
// either we should open more connections depending on stream count
// (not exposed by hyper, need our own counter)
// or we can change this to an Option rather than a VecDeque.
//
// Opening more connections to the same db because we run out of streams
// seems somewhat redundant though.
//
// Probably we should run a semaphore and just the single conn. TBD.
conns: VecDeque<ConnPoolEntry>,
_guard: HttpEndpointPoolsGuard<'static>,
global_connections_count: Arc<AtomicUsize>,
}
impl EndpointConnPool {
fn get_conn_entry(&mut self) -> Option<ConnPoolEntry> {
let Self { conns, .. } = self;
loop {
let conn = conns.pop_front()?;
if !conn.conn.is_closed() {
conns.push_back(conn.clone());
return Some(conn);
}
}
}
fn remove_conn(&mut self, conn_id: uuid::Uuid) -> bool {
let Self {
conns,
global_connections_count,
..
} = self;
let old_len = conns.len();
conns.retain(|conn| conn.conn_id != conn_id);
let new_len = conns.len();
let removed = old_len - new_len;
if removed > 0 {
global_connections_count.fetch_sub(removed, atomic::Ordering::Relaxed);
Metrics::get()
.proxy
.http_pool_opened_connections
.get_metric()
.dec_by(removed as i64);
}
removed > 0
}
}
impl Drop for EndpointConnPool {
fn drop(&mut self) {
if !self.conns.is_empty() {
self.global_connections_count
.fetch_sub(self.conns.len(), atomic::Ordering::Relaxed);
Metrics::get()
.proxy
.http_pool_opened_connections
.get_metric()
.dec_by(self.conns.len() as i64);
}
}
}
pub(crate) struct GlobalConnPool {
// endpoint -> per-endpoint connection pool
//
// That should be a fairly conteded map, so return reference to the per-endpoint
// pool as early as possible and release the lock.
global_pool: DashMap<EndpointCacheKey, Arc<RwLock<EndpointConnPool>>>,
/// Number of endpoint-connection pools
///
/// [`DashMap::len`] iterates over all inner pools and acquires a read lock on each.
/// That seems like far too much effort, so we're using a relaxed increment counter instead.
/// It's only used for diagnostics.
global_pool_size: AtomicUsize,
/// Total number of connections in the pool
global_connections_count: Arc<AtomicUsize>,
config: &'static crate::config::HttpConfig,
}
impl GlobalConnPool {
pub(crate) fn new(config: &'static crate::config::HttpConfig) -> Arc<Self> {
let shards = config.pool_options.pool_shards;
Arc::new(Self {
global_pool: DashMap::with_shard_amount(shards),
global_pool_size: AtomicUsize::new(0),
config,
global_connections_count: Arc::new(AtomicUsize::new(0)),
})
}
pub(crate) fn shutdown(&self) {
// drops all strong references to endpoint-pools
self.global_pool.clear();
}
pub(crate) async fn gc_worker(&self, mut rng: impl Rng) {
let epoch = self.config.pool_options.gc_epoch;
let mut interval = tokio::time::interval(epoch / (self.global_pool.shards().len()) as u32);
loop {
interval.tick().await;
let shard = rng.gen_range(0..self.global_pool.shards().len());
self.gc(shard);
}
}
fn gc(&self, shard: usize) {
debug!(shard, "pool: performing epoch reclamation");
// acquire a random shard lock
let mut shard = self.global_pool.shards()[shard].write();
let timer = Metrics::get()
.proxy
.http_pool_reclaimation_lag_seconds
.start_timer();
let current_len = shard.len();
let mut clients_removed = 0;
shard.retain(|endpoint, x| {
// if the current endpoint pool is unique (no other strong or weak references)
// then it is currently not in use by any connections.
if let Some(pool) = Arc::get_mut(x.get_mut()) {
let EndpointConnPool { conns, .. } = pool.get_mut();
let old_len = conns.len();
conns.retain(|conn| !conn.conn.is_closed());
let new_len = conns.len();
let removed = old_len - new_len;
clients_removed += removed;
// we only remove this pool if it has no active connections
if conns.is_empty() {
info!("pool: discarding pool for endpoint {endpoint}");
return false;
}
}
true
});
let new_len = shard.len();
drop(shard);
timer.observe();
// Do logging outside of the lock.
if clients_removed > 0 {
let size = self
.global_connections_count
.fetch_sub(clients_removed, atomic::Ordering::Relaxed)
- clients_removed;
Metrics::get()
.proxy
.http_pool_opened_connections
.get_metric()
.dec_by(clients_removed as i64);
info!("pool: performed global pool gc. removed {clients_removed} clients, total number of clients in pool is {size}");
}
let removed = current_len - new_len;
if removed > 0 {
let global_pool_size = self
.global_pool_size
.fetch_sub(removed, atomic::Ordering::Relaxed)
- removed;
info!("pool: performed global pool gc. size now {global_pool_size}");
}
}
pub(crate) fn get(
self: &Arc<Self>,
ctx: &RequestMonitoring,
conn_info: &ConnInfo,
) -> Option<Client> {
let endpoint = conn_info.endpoint_cache_key()?;
let endpoint_pool = self.get_or_create_endpoint_pool(&endpoint);
let client = endpoint_pool.write().get_conn_entry()?;
tracing::Span::current().record("conn_id", tracing::field::display(client.conn_id));
info!(
cold_start_info = ColdStartInfo::HttpPoolHit.as_str(),
"pool: reusing connection '{conn_info}'"
);
ctx.set_cold_start_info(ColdStartInfo::HttpPoolHit);
ctx.success();
Some(Client::new(client.conn, client.aux))
}
fn get_or_create_endpoint_pool(
self: &Arc<Self>,
endpoint: &EndpointCacheKey,
) -> Arc<RwLock<EndpointConnPool>> {
// fast path
if let Some(pool) = self.global_pool.get(endpoint) {
return pool.clone();
}
// slow path
let new_pool = Arc::new(RwLock::new(EndpointConnPool {
conns: VecDeque::new(),
_guard: Metrics::get().proxy.http_endpoint_pools.guard(),
global_connections_count: self.global_connections_count.clone(),
}));
// find or create a pool for this endpoint
let mut created = false;
let pool = self
.global_pool
.entry(endpoint.clone())
.or_insert_with(|| {
created = true;
new_pool
})
.clone();
// log new global pool size
if created {
let global_pool_size = self
.global_pool_size
.fetch_add(1, atomic::Ordering::Relaxed)
+ 1;
info!(
"pool: created new pool for '{endpoint}', global pool size now {global_pool_size}"
);
}
pool
}
}
pub(crate) fn poll_http2_client(
global_pool: Arc<GlobalConnPool>,
ctx: &RequestMonitoring,
conn_info: &ConnInfo,
client: Send,
connection: Connect,
conn_id: uuid::Uuid,
aux: MetricsAuxInfo,
) -> Client {
let conn_gauge = Metrics::get().proxy.db_connections.guard(ctx.protocol());
let session_id = ctx.session_id();
let span = info_span!(parent: None, "connection", %conn_id);
let cold_start_info = ctx.cold_start_info();
span.in_scope(|| {
info!(cold_start_info = cold_start_info.as_str(), %conn_info, %session_id, "new connection");
});
let pool = match conn_info.endpoint_cache_key() {
Some(endpoint) => {
let pool = global_pool.get_or_create_endpoint_pool(&endpoint);
pool.write().conns.push_back(ConnPoolEntry {
conn: client.clone(),
conn_id,
aux: aux.clone(),
});
Arc::downgrade(&pool)
}
None => Weak::new(),
};
tokio::spawn(
async move {
let _conn_gauge = conn_gauge;
let res = connection.await;
match res {
Ok(()) => info!("connection closed"),
Err(e) => error!(%session_id, "connection error: {}", e),
}
// remove from connection pool
if let Some(pool) = pool.clone().upgrade() {
if pool.write().remove_conn(conn_id) {
info!("closed connection removed");
}
}
}
.instrument(span),
);
Client::new(client, aux)
}
pub(crate) struct Client {
pub(crate) inner: Send,
aux: MetricsAuxInfo,
}
impl Client {
pub(self) fn new(inner: Send, aux: MetricsAuxInfo) -> Self {
Self { inner, aux }
}
pub(crate) fn metrics(&self) -> Arc<MetricCounter> {
USAGE_METRICS.register(Ids {
endpoint_id: self.aux.endpoint_id,
branch_id: self.aux.branch_id,
})
}
}

View File

@@ -5,13 +5,13 @@ use bytes::Bytes;
use anyhow::Context;
use http::{Response, StatusCode};
use http_body_util::{combinators::BoxBody, BodyExt, Full};
use http_body_util::Full;
use serde::Serialize;
use utils::http::error::ApiError;
/// Like [`ApiError::into_response`]
pub(crate) fn api_error_into_response(this: ApiError) -> Response<BoxBody<Bytes, hyper1::Error>> {
pub(crate) fn api_error_into_response(this: ApiError) -> Response<Full<Bytes>> {
match this {
ApiError::BadRequest(err) => HttpErrorBody::response_from_msg_and_status(
format!("{err:#?}"), // use debug printing so that we give the cause
@@ -64,24 +64,17 @@ struct HttpErrorBody {
impl HttpErrorBody {
/// Same as [`utils::http::error::HttpErrorBody::response_from_msg_and_status`]
fn response_from_msg_and_status(
msg: String,
status: StatusCode,
) -> Response<BoxBody<Bytes, hyper1::Error>> {
fn response_from_msg_and_status(msg: String, status: StatusCode) -> Response<Full<Bytes>> {
HttpErrorBody { msg }.to_response(status)
}
/// Same as [`utils::http::error::HttpErrorBody::to_response`]
fn to_response(&self, status: StatusCode) -> Response<BoxBody<Bytes, hyper1::Error>> {
fn to_response(&self, status: StatusCode) -> Response<Full<Bytes>> {
Response::builder()
.status(status)
.header(http::header::CONTENT_TYPE, "application/json")
// we do not have nested maps with non string keys so serialization shouldn't fail
.body(
Full::new(Bytes::from(serde_json::to_string(self).unwrap()))
.map_err(|x| match x {})
.boxed(),
)
.body(Full::new(Bytes::from(serde_json::to_string(self).unwrap())))
.unwrap()
}
}
@@ -90,14 +83,14 @@ impl HttpErrorBody {
pub(crate) fn json_response<T: Serialize>(
status: StatusCode,
data: T,
) -> Result<Response<BoxBody<Bytes, hyper1::Error>>, ApiError> {
) -> Result<Response<Full<Bytes>>, ApiError> {
let json = serde_json::to_string(&data)
.context("Failed to serialize JSON response")
.map_err(ApiError::InternalServerError)?;
let response = Response::builder()
.status(status)
.header(http::header::CONTENT_TYPE, "application/json")
.body(Full::new(Bytes::from(json)).map_err(|x| match x {}).boxed())
.body(Full::new(Bytes::from(json)))
.map_err(|e| ApiError::InternalServerError(e.into()))?;
Ok(response)
}

Some files were not shown because too many files have changed in this diff Show More