Repository navigation
24.3 Stable: use the clickhouse/minio-minio mirror image - #2701
Conversation
minio/minio was deleted from Docker Hub, so every with_minio=True test fails in cluster.start() on the image pull. Point docker_compose_minio.yml at the clickhouse/minio-minio mirror (upstream ClickHouse#119999). The mirror has no RELEASE.2023-09-30T07-02-29Z tag, so bump to RELEASE.2024-07-31T05-46-26Z, the same image releases/24.8.14 uses.
dl.min.io returns 410 Gone for every minio/mc binary, so the stateless-test image can no longer be rebuilt. Copy minio and mc from clickhouse/minio-minio:RELEASE.2024-07-31T05-46-26Z instead, the same image the integration tests use. The newer mc and minio need the same changes upstream made when it moved stateless tests to 2024 minio (ClickHouse#67975): the renamed mc policy commands, and not sending a "/" prefix in S3ObjectStorage::listObjects.
|
This is an automated comment for commit 922a4fc with description of existing statuses. It's updated for the latest CI running ✅ Click here to open a full report in a separate page Successful checks
|
On runners where buildx attaches a provenance attestation, the per-arch tag becomes an index, and `docker manifest create` rejects it. Rebuild stateless-test to replace the amd64 tags pushed that way.
Azurite's transitive dependency @dabh/diagnostics >= 2.0.5 pulls @so-ric/colorspace, which uses syntax Node.js 12 from Ubuntu 22.04 can't parse, so azurite failed to start and the server aborted on Azure errors.
Builds dropped from the run config (e.g. tsan/msan by ci_exclude_ tags) have no reports, so the build check stayed pending and failed on every PR. ci.py now saves the run config for the job, and the build report skips missing builds that are not part of the run.
This comment was marked as outdated.
This comment was marked as outdated.
CI triage for
|
| Check | Classification |
|---|---|
GrypeScan (server) / Grype Scan |
infra: outdated grype version |
GrypeScan (server, -alpine) / Grype Scan |
infra: same cause |
GrypeScan (keeper) / Grype Scan |
infra: same cause |
FinishCheck |
knock-on effect of the Grype failures |
Grype Scan (all 3 jobs)
The ClickHouse images were pulled fine. Grype itself failed to load its vulnerability database (log):
WARN error updating db error=unable to activate new vulnerability database: failed to hydrate database: failed to open db: unable to migrate: constraint failed: UNIQUE constraint failed: operating_systems.name, ...
ERROR failed to load vulnerability db: database does not exist
The keeper and alpine jobs log the same two lines. releases/24.3.18 still pins GRYPE_VERSION: "v0.92.2-arm64v8" (.github/workflows/grype_scan.yml:23, .github/grype/run_grype_scan.sh:6), and that grype release can no longer read the current upstream DB schema. Other branches have already fixed this with 1ebf3b1 ("Fix Grype Scan: bump grype to v0.115.0 (v0.92.2 can't load current DB)"). Both antalya-26.3 and releases/24.8.14 now pin v0.115.0-arm64v8. This PR does not touch anything Grype-related, so the scan fails the same way for any run on this base branch.
FinishCheck
File ".github/create_workflow_report.py", line 445, in get_cves
results.append(json.loads(content))
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
This happens because the Grype jobs uploaded empty result files. It will go green once Grype works.
Suggested next step (optional, not required for this PR)
To get Grype green on the 24.3 line, cherry-pick 1ebf3b1. It is a 2-line bump, v0.92.2 → v0.115.0, in .github/workflows/grype_scan.yml and .github/grype/run_grype_scan.sh. It could go on this PR or in a separate CI fix PR against releases/24.3.18. I have not pushed it, since you only asked for triage. Mention me if you want it on this branch. Otherwise, the functional checks for this MinIO-image change are fully green.
@blau-ai , do it in a separate PR into |
|
@zvonand Done. I opened the Grype bump as a separate PR into
Nothing was pushed to this PR (#2701). The Grype checks on #2717 will show whether the bump fixes the scan; I have not run it locally. |
minio/minio was deleted from Docker Hub, so every with_minio=True test fails in cluster.start() on the image pull. Point docker_compose_minio.yml at the clickhouse/minio-minio mirror (upstream ClickHouse#119999). The mirror has no RELEASE.2023-09-30T07-02-29Z tag, so bump to RELEASE.2024-07-31T05-46-26Z, the same image releases/24.8.14 uses.
Changelog category (leave one):
CI/CD Options
Exclude tests:
Regression jobs to run: