Repository navigation
Update dockerfile dependencies - #5006
Conversation
|
Reviewed this Renovate digest bump — looks clean, no findings. What changed: Verification:
Minor note (informational, not blocking): the ca-certificates update removes trust in a compromised eMudhra intermediate CA. This only matters if a peer or destination endpoint presents a certificate chaining through that CA — an intentional upstream distrust, not a regression. |
Code reviewNo issues found. Checked for bugs and CLAUDE.md compliance. Details: This PR is a same-tag digest refresh of the pinned Dependency regression check (REVIEW.md): Alpine same-tag digest bumps correspond to routine stable-branch rebuilds (backported security/bugfix package updates only), which is the lowest-risk category of Renovate update. Note: the review environment had no outbound network access to Docker Hub or the Alpine upstream repos, so the specific point-release changelog for this digest range could not be independently verified. |
c6d722c to
75e1cc5
Compare
|
Renovate review — reviewed, looks clean ✅ Change: Verification
Upstream changes in range
Alpine also confirmed that aports, packages, and the Oct 21 releases were not tampered with during their GitLab incident. Regression risk No PII/secret exposure, no new stdout/stderr/log output, no API or build-arg surface changes. Nothing blocking. |
🔄 Flaky Test DetectedAnalysis: Infrastructure flake: the cockroachdb-v25_ch-stable job failed during Docker service bring-up because host port 40002 was already in use when starting flow-snapshot-worker (all 3 recovery attempts hit the same bind error), so no tests ever ran, while all 18 other matrix jobs on the same commit passed. ✅ Automatically retrying the workflow |
75e1cc5 to
e1493fe
Compare
Renovate review — ✅ cleanDigest-only bump of Digest verification
What changed upstream
Regression assessment: none expected.
No new stdout/stderr/log output and no credential handling in this diff, so the PII/secret-logging check is a no-op here. 🤖 Generated with Claude Code |
🔄 Flaky Test DetectedAnalysis: A single matrix job (maria-12_ch-stable) timed out after 3 minutes in the post-test cleanup wait ("wait for flow dropped") of TestApiMy/TestResyncWithSnapshotConfigOnRunningPipe, while the same 482-test package passed in every other MySQL/MariaDB job and the PR only bumps a Dockerfile base image — a timing-sensitive terminate-after-resync hang, not a regression from the change. ✅ Automatically retrying the workflow |
e1493fe to
419d00e
Compare
| # syntax=docker/dockerfile:1.28@sha256:bb22d9815c728170f72750f4e5b0d672e06176142e1d602c7e66c050100b7e5b | ||
|
|
||
| FROM golang:1.27-alpine@sha256:8a5910f31396cd4d89662f56c68b3ae31d374308270a1c3bd96672ee5ed43414 AS builder | ||
| FROM golang:1.27-alpine@sha256:d200d2dd76b03a99b52651fd1325aff7dcd35422c35c01bac56571b0507f3deb AS builder |
There was a problem hiding this comment.
Release-note check (not a blocker) — this digest moves Go 1.27.2 → 1.27.3.
Digest d200d2d is exactly the current golang:1.27.3-alpine manifest list (pushed 2026-10-27); the previous 8a5910f was a 1.27.2 build. docker-library/golang bumped 1.27 → 1.27.3 on 2026-10-20.
Go 1.27.3 contains a single security fix, CVE-2026-111943 (golang/go#82345): crypto/x509.Verify now filters out any chain that passes through one of eight blocked intermediate CA keys, so such chains are no longer returned to the caller and verification fails.
Why this is worth a look for PeerDB specifically: the flow worker makes outbound TLS connections to user-supplied peers (Postgres, MySQL, MongoDB, Snowflake, ClickHouse, Kafka, Elasticsearch, S3). If any of those endpoints presents a chain through one of the blocked intermediates, it will start failing with a certificate-verification error once this image ships, and the upstream issue documents no GODEBUG opt-out. The CA is deliberately unnamed upstream, so this cannot be pre-verified from here — just something to recognize if peer TLS errors appear post-deploy, rather than diagnosing it as a PeerDB regression.
Otherwise the bump is clean: the tag is unchanged (1.27-alpine), flow/go.mod has no toolchain pin so it simply builds with 1.27.3, and no logging/stdout surface is touched.
| fi | ||
|
|
||
| FROM alpine:3.24@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6 AS flow-base | ||
| FROM alpine:3.24@sha256:be2656061b536c1883c4034790e21d0a09c5aaaf51968b89368a3c0371ea2c9a AS flow-base |
There was a problem hiding this comment.
Release-note check (not a blocker) — this digest moves Alpine 3.24.2 → 3.24.3.
Digest be26560 is the current alpine:3.24 / alpine:3.24.3 manifest list (pushed 2026-10-23); 294b683 was 3.24.2 (2026-09-18). Per docker-library/official-images#22559, 3.24.3 is a pure security patch: musl, zlib and OpenSSL fixes plus ca-certificates refreshed to 20261020. No API/ABI breaks in an Alpine patch release.
The digest is applied consistently — both occurrences (here and stacks/peerdb-server.Dockerfile:44) were updated, and no stale alpine:3.24 pins remain anywhere in stacks/.
Minor note on the ca-certificates 20261020 refresh: line 36 does apk add --no-cache ca-certificates at build time, so the trust store already tracked the live 3.24 repo and is not actually gated by this digest pin. Same class of consideration as the Go 1.27.3 crypto/x509 change flagged on line 3.
419d00e to
cd2b786
Compare
Review: dependency digest bump — looks clean ✅Reviewed per What actually changed
Digests verified against Docker Hub: both resolve to the tags they are pinned against ( Regression assessmentGo 1.27.1 → 1.27.3 (release history) — two patch releases, no language or API changes:
Alpine 3.24.2 → 3.24.3 (release post) — patch release, security only: Only behavioral change worth being aware of: certificate chains relying on the revoked eMudhra intermediates will now fail verification — both in Go's Consistency checks
🤖 Generated with Claude Code |
🔄 Flaky Test DetectedAnalysis: Test_BigQuery_CDC_Isolated_Table_Backpressure_Does_Not_Block_Sibling hit a 2-minute WaitFor timeout waiting for BigQuery's APPENDS change feed to surface the first CDC batch, while the identical 59-test BigQuery suite passed in the bigquery_ch-stable and bigquery_ch-lts jobs of the same run — pointing to external-service latency rather than a code defect. ✅ Automatically retrying the workflow |
cd2b786 to
eb4b350
Compare
| # syntax=docker/dockerfile:1.28@sha256:bb22d9815c728170f72750f4e5b0d672e06176142e1d602c7e66c050100b7e5b | ||
|
|
||
| FROM golang:1.27-alpine@sha256:8a5910f31396cd4d89662f56c68b3ae31d374308270a1c3bd96672ee5ed43414 AS builder | ||
| FROM golang:1.27-alpine@sha256:d200d2dd76b03a99b52651fd1325aff7dcd35422c35c01bac56571b0507f3deb AS builder |
There was a problem hiding this comment.
Not a pure digest refresh — this moves the Go toolchain 1.27.1 → 1.27.3.
- old
8a5910f…=golang:1.27.1-alpine3.24, built 2026-09-21 (pinned back in Update dockerfile dependencies #4874) - new
d200d2d…=golang:1.27.3-alpine3.24, built 2026-10-27
Upstream range (docker-library/golang commit log + Go release history):
- Go 1.27.2 (2026-10-08) — 15 security backports:
cmd/gotoolchain andgolang.org/fips140checksum bypass via a malicious GOMODPROXY,crypto/tls,html/templateXSS (stale jsCtx at template-literal entry; missing ES2015+ keywords inregexpPrecederKeywords),net/http(http2 transport accepting multiple Content-Length headers),net/textproto,os. Plus compiler, linker, runtime,compress/flate,encoding/jsonandencoding/json/v2bug fixes. - Go 1.27.3 (2026-10-19) —
crypto/x509only (CVE-2026-111943 / GO-2026-6809). See the trust-store note on the alpine line below.
No API or behavioural change beyond the x509 denylist, so no regression risk I can see for flow. It also closes an existing drift: CI already pins go-version: 1.27.3 in ci.yml, flow.yml, golang-lint.yml, tilt-flow.yml and cleanup.yml, so until now the shipped binary was built with an older toolchain than the one tests ran against.
| fi | ||
|
|
||
| FROM alpine:3.24@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6 AS flow-base | ||
| FROM alpine:3.24@sha256:be2656061b536c1883c4034790e21d0a09c5aaaf51968b89368a3c0371ea2c9a AS flow-base |
There was a problem hiding this comment.
Alpine 3.24.2 → 3.24.3 (old 294b683… = 3.24.2, published 2026-09-18; new be26560… = 3.24.3, published 2026-10-23). Per docker-library/official-images#22559 this is a security point release: musl CVE-2026-107891, zlib CVE-2026-85091, 13 OpenSSL CVEs, and ca-certificates 20261020.
emSign Root CA - C1, emSign ECC Root CA - C3 and emSign ECC Root CA - G3 following the eMudhra intermediate key compromise, and the Go 1.27.3 crypto/x509 fix on line 3 adds a denylist for the affected intermediates. Combined, after this bump peer-flow will refuse TLS to any source/destination peer whose chain terminates in an emSign root, surfacing to users as x509: certificate signed by unknown authority. Narrow blast radius, but it is silent until runtime.
Three things I checked that are fine:
- The known 3.24.3 breakage (ca-certificates-20261020-r0 conflicting with an exact
ca-certificates=20260909-r0pin) only hits Dockerfiles that pin the version. Both Dockerfiles here use unpinnedapk add --no-cache ca-certificates, so no impact — consistent withdocker-buildpassing. - The AWS RDS bundle added on line 35 is an Amazon root set with its own
--checksum, unaffected by the emSign removal. - Builder and runtime move to the same Alpine 3.24.3 in this PR, so the CGO geos/musl link between the
golang:1.27-alpinebuilder and this stage stays ABI-consistent. Bothalpine:3.24pins in the repo (here and peerdb-server.Dockerfile:44) were updated together.
| cp target/${BUILD_MODE}/peerdb-server /root/target/ | ||
|
|
||
| FROM alpine:3.24@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6 | ||
| FROM alpine:3.24@sha256:be2656061b536c1883c4034790e21d0a09c5aaaf51968b89368a3c0371ea2c9a |
There was a problem hiding this comment.
Same 3.24.2 → 3.24.3 bump as flow.Dockerfile — correct and consistent.
Scope caveat on the 13 OpenSSL CVEs this release carries: the peerdb-server binary itself does not pick them up. The builder stage sets OPENSSL_STATIC=1 and installs openssl-libs-static (lines 13 and 15), so OpenSSL is statically linked from lukemathwalker/cargo-chef:latest-rust-1.99.1-alpine, whose digest this PR does not touch. The runtime bump does fix the OpenSSL used by postgresql-client and curl installed on line 46, and brings in the musl/zlib/ca-certificates fixes.
Pre-existing, not introduced here, and not a reason to hold this PR — but if the OpenSSL CVEs matter, the cargo-chef builder digest is the pin that actually needs to move.
🔄 Flaky Test DetectedAnalysis: Both failures are infrastructure, not test logic: postgres-18_other could not start flow-snapshot-worker because host port 40002 was "address already in use", and mysql-pos_ch-stable got an HTTP 500 downloading the Tilt tarball from GitHub releases — no tests ran in either job, and the other 9 matrix jobs passed on the same commit. ✅ Automatically retrying the workflow |
eb4b350 to
dafd925
Compare
|
Re-reviewed per
Both digests re-verified against Docker Hub: they are the current digests for the tags they are pinned against, and Upstream delta is security-only patch releases: Go 1.27.2 (15 CVEs, mostly Addition not in the earlier comment: this bump also closes a toolchain skew. Every CI workflow ( No stray copies of the old digests remain in the repo, and the diff adds no logging/stdout output, so there is no PII or secret-exposure surface. 🤖 Generated with Claude Code |
dafd925 to
a18978e
Compare
|
Renovate digest bump review — no findings Reviewed per What the digests actually resolve to Both new digests match what Docker Hub currently serves for the pinned tags (verified against the registry):
So despite being a digest-only bump, this quietly moves the Go toolchain 1.27.1 -> 1.27.3. Worth knowing, since it is more than a rebuild. Upstream changes in the bumped range Alpine 3.24.2 -> 3.24.3 (release notes) is a security-only point release:
Go 1.27.1 -> 1.27.3 (release history):
Regression assessment No regressions found. Both patch ranges are security plus bugfix only; 1.27.2 in fact removes regressions rather than adding any. No breaking changes land inside the 1.27.x patch range (the I checked the one Go 1.27.2 change with plausible blast radius here, the Operational note (not a blocker) Both halves of this bump distrust the compromised eMudhra intermediates, Go via the Consistency check All three pinned references in the repo were updated together ( CI was still pending at review time, so this is a static review only. The tilt matrix should be green before automerge. |
a18978e to
89f258b
Compare
| # syntax=docker/dockerfile:1.28@sha256:bb22d9815c728170f72750f4e5b0d672e06176142e1d602c7e66c050100b7e5b | ||
|
|
||
| FROM golang:1.27-alpine@sha256:8a5910f31396cd4d89662f56c68b3ae31d374308270a1c3bd96672ee5ed43414 AS builder | ||
| FROM golang:1.27-alpine@sha256:d200d2dd76b03a99b52651fd1325aff7dcd35422c35c01bac56571b0507f3deb AS builder |
There was a problem hiding this comment.
Dep-bump review — golang:1.27-alpine 8a5910f → d200d2d is Go 1.27.1 → 1.27.3. The new digest is the current 1.27.3-alpine (identical to 1.27.3-alpine3.24). The old digest was pinned on 2026-09-30 in #4874, before 1.27.2 shipped, so this crosses two patch releases rather than one.
What the range contains:
- go1.27.2 (2026-10-08) — security fixes in
gocmd,crypto/tls,html/template,net/http,net/textproto,os; bug fixes in compiler/linker/runtime,compress/flate,encoding/json,encoding/json/v2,net/http,os. - go1.27.3 (2026-10-19) — a single security fix in
crypto/x509: CVE-2026-111943 (crypto/x509: mitigation for security concerns with a specific Certificate Authority golang/go#82345).
flow binaries: go1.27.3 makes x509.Certificate.Verify filter out any chain containing one of eight blocklisted intermediate CA keys (the eMudhra intermediates Mozilla added to OneCRL on 2026-10-17). There is no GODEBUG opt-out. PeerDB dials TLS against user-supplied endpoints (Postgres/MySQL/Mongo sources; ClickHouse/Snowflake/Kafka/Elasticsearch destinations), so any peer presenting a chain through one of those intermediates will start failing handshakes once this image rolls out.
Not a reason to block — it is the intended security fix, and .github/workflows/*.yml already builds and tests on go-version 1.27.3, so this merely brings the container builder in line with CI. But it is probably worth a release-note line so support recognizes that failure mode instead of chasing it as a PeerDB regression.
No other regressions identified in the range for how this repo uses the toolchain (flow/go.mod declares go 1.27.0, so there is no language-version interaction).
| fi | ||
|
|
||
| FROM alpine:3.24@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6 AS flow-base | ||
| FROM alpine:3.24@sha256:be2656061b536c1883c4034790e21d0a09c5aaaf51968b89368a3c0371ea2c9a AS flow-base |
There was a problem hiding this comment.
Dep-bump review — alpine:3.24 294b683 → be26560 is 3.24.2 → 3.24.3. The new digest matches the 3.24.3 tag (pushed 2026-10-23); the old one was 3.24.2. The same digest is updated consistently at stacks/peerdb-server.Dockerfile:44, and no other alpine:3.24 pin exists in the repo, so the bump is complete.
Per the 3.21.9 / 3.22.7 / 3.23.7 / 3.24.3 announcement this is a pure patch release with no breaking changes inside the 3.24 series, carrying:
musl— CVE-2026-107891openssl— the 2026-09-29 advisory (11 CVEs, incl. CVE-2026-35189, CVE-2026-42772, CVE-2026-54872/3/5)zlib— CVE-2026-85091ca-certificates— eMudhra intermediate removalphp8.3–8.5— not installed in either image, no impact
The packages these stages actually install (ca-certificates + geos here; ca-certificates, postgresql-client, curl, iputils in peerdb-server.Dockerfile) see no ABI or behavior breaks.
Cross-cutting note: the ca-certificates refresh is the same eMudhra distrust as the Go 1.27.3 crypto/x509 change flagged on line 3, and it affects the Go runtime system cert pool in this image. nexus is largely insulated, since postgres-connection and the reqwest / mysql_async clients use rustls with bundled webpki-roots rather than the system store.
The pinned AWS RDS bundle ADD --checksum= is untouched by this PR, as expected. Net: strictly a security patch.
89f258b to
3db2500
Compare
Renovate review: digest bumps reviewed, looks clean ✅Diff is 3 lines across What the digests actually resolve to (Renovate labels these "digest" updates, but both move real versions):
Upstream notes for the bumped range
Compatibility checks
Sources: Go release history, Alpine 3.24.3 release post |
3db2500 to
5967ed6
Compare
|
Reviewed this digest bump — no findings, looks clean to merge. What the digests actually resolve to
So the bumped range is Alpine 3.24.2 → 3.24.3 and Go 1.27.1 → 1.27.2 → 1.27.3. Release notes for the bumped range Alpine 3.24.3 — part of the coordinated 3.21.9/3.22.7/3.23.7/3.24.3 security release. Contains OpenSSL CVE fixes (CVE-2026-35189, -42772, -54872/3/5, -72897, -75804/5/6, -77696, -84782/4) and PHP fixes (CVE-2026-109618/9). PHP is not installed in either image; the OpenSSL fixes do benefit Go 1.27.2 — security fixes in Go 1.27.3 — expedited release for CVE-2026-111943: Other checks
|
5967ed6 to
daacf7c
Compare
Review: digest bump — looks clean ✅Digest-only bump of two pinned base images in Digest verificationBoth new digests resolve to the current multi-arch manifests of the tags they are pinned against (checked via Docker Hub registry metadata) — no tag drift, no accidental cross-major move:
Both Upstream changes in the bumped range (regression check)Go 1.27.1 → 1.27.3 — two patch releases, both security-driven, no language/behaviour changes:
Alpine 3.24.2 → 3.24.3 (release post, 2026-10-21): OpenSSL CVE fixes (CVE-2026-35189, -42772, -54872/3/5, -72897, -75804/5/6, -77696, -84782, -84784) plus PHP fixes (not used here). No regressions identified in either range — these are pure patch releases. One behavioural note worth being aware of rather than a blocker: after Go 1.27.3, BonusThe Go bump actually removes drift: CI already builds and lints with Go No paths in this diff fall under a |
daacf7c to
e4e096f
Compare
|
Reviewed this Renovate digest bump — looks clean, no findings. What changed
Regression check (upstream notes)
Consistency checks
LGTM — this is a security patch bump in the desirable direction. |
e4e096f to
559704e
Compare
|
Reviewed this Renovate digest/patch bump — looks clean, no findings. Digest verification — all three pins resolve to the exact tags claimed (current Docker Hub manifest-list digests):
Regression check on the bumped range
Consistency — both Review checklist — no stdout/stderr/log changes, so the PII/secrets rule is N/A. No changed path falls under a
|
🔄 Flaky Test DetectedAnalysis: TestApiMy/TestResyncWithSnapshotConfigOnRunningPipe hit a 3-minute "wait for flow dropped" timeout in only the mysql-gtid_ch-lts cell, while the identical package/MySQL image passed in mysql-gtid_ch-latest and sibling tests using the same terminate-and-drop path passed in the same job, on a Renovate Dockerfile PR unrelated to flow termination. ✅ Automatically retrying the workflow |
559704e to
92f4e8e
Compare
| # syntax=docker/dockerfile:1.28@sha256:bb22d9815c728170f72750f4e5b0d672e06176142e1d602c7e66c050100b7e5b | ||
|
|
||
| FROM golang:1.27-alpine@sha256:8a5910f31396cd4d89662f56c68b3ae31d374308270a1c3bd96672ee5ed43414 AS builder | ||
| FROM golang:1.27-alpine@sha256:d200d2dd76b03a99b52651fd1325aff7dcd35422c35c01bac56571b0507f3deb AS builder |
There was a problem hiding this comment.
This digest bump is not a pure base-image refresh — it moves the Go toolchain used to build the shipped flow binaries from 1.27.1 → 1.27.3. (An earlier review comment on this PR stated the Go version was unchanged; that is incorrect.)
Docker Hub tag→digest mapping:
| Digest | Resolves to |
|---|---|
8a5910f3… (old) |
golang:1.27.1-alpine3.24, built 2026-09-21 |
d200d2dd… (new) |
golang:1.27.3-alpine3.24 = 1.27-alpine, built 2026-10-27 |
So the bumped range spans two security releases:
- Go 1.27.2 (2026-10-08) — security fixes to the go command and
crypto/tls,html/template,net/http,net/textproto,os. - Go 1.27.3 (2026-10-19) — security fix to
crypto/x509(CVE-2026-111943): the verifier now rejects any chain passing through eight eMudhra intermediate CA keys (added to Mozilla OneCRL on 2026-10-17).
No change requested — the bump is desirable, and it also aligns the image builder with CI, which already pins go-version: 1.27.3 across the flow / ci / golang-lint / cleanup / tilt-flow workflows; released images were previously built with an older toolchain than CI tested against.
Worth flagging to reviewers, though: the crypto/x509 change is a user-visible TLS behaviour change in the shipped binaries, not just a build-time detail. Any mirror whose source or destination certificate chains through one of those intermediates will start failing verification once this image ships. Intended security hardening, but it may belong in release notes.
🔄 Flaky Test DetectedAnalysis: Not a test failure at all — the cockroachdb-v24_ch-lts job died during environment setup when ✅ Automatically retrying the workflow |
This PR contains the following updates:
294b683→be265608a5910f→d200d2d24.22.1-alpine→24.22.2-alpineRelease Notes
nodejs/node (node)
v24.22.2: 2026-10-28, Version 24.22.2 'Krypton' (LTS), @aduh95Compare Source
Commits
f8bdcd54ed] - deps: V8: backport786c1c2(Eliau Elkouby) #662493b93fd6b2b] - deps: V8: backport1a00890(Jakob Linke) #657645aaf62faaa] - deps: V8: backportc9c0abf(Jakob Linke) #657643b8eb1a920] - deps: V8: backportebd1578(Marja Hölttä) #65764ab70b5d681] - deps: V8: cherry-pick853553f(Victor Gomes) #6624117ec999f24] - stream: keep async context of parked writes (Matteo Collina) #66907Configuration
📅 Schedule: (in timezone Etc/UTC)
🚦 Automerge: Enabled.
♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.
👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.
This PR was generated by Mend Renovate. View the repository job log.