Repository navigation
fix(cloud-functions): default nvcf-api to 1.28.2 for the Cassandra replica-down fixes - #3012
Conversation
…plica-down fixes 1.28.2 carries #2918: range read timeouts retry on the next host when the ncp profile selects RangeReadRetryPolicy, and retriable Cassandra failures return 503 with Retry-After instead of a 500 with CQL. It also picks up #2966 (/info version after config refresh). Relates to #2893 Signed-off-by: shobham <shobham@nvidia.com>
|
No actionable comments were generated in the recent review.
|
| Layer / File(s) | Summary |
|---|---|
Update chart and image versions deploy/helm/cloud-functions/nvcf-api/Chart.yaml, deploy/helm/cloud-functions/nvcf-api/values.yaml |
The chart’s appVersion and default api.image.tag change from 1.28.0 to 1.28.2. |
Priority:
Estimated code review effort: 1 (Trivial) | ~3 minutes
Change: Bug fix
Possibly related PRs
- NVIDIA/nvcf#2939: Updates the same chart fields from
1.27.7to1.28.0. - NVIDIA/nvcf#2764: Updates the same nvcf-api chart fields in an earlier release bump.
- NVIDIA/nvcf#2918: Implements behavior included in the
1.28.2API image selected by this update.
Suggested reviewers: balajinvda
Merge Risk: 
Minimal · up to 86960
-
No actionable merge-blocking risk was identified in this version-only chart update.
Pre-merge checks: 
4
-
Full details: Title check
-
Explanation
The title uses the required Conventional Commits format, includes the required scope for a customer-impacting fix, and accurately describes the nvcf-api version update for Cassandra replica-down fixes.
-
Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Comment @coderabbitai help to get the list of available commands.
|
🎉 This PR is included in deploy/helm/cloud-functions/v1.33.2 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
|
This PR is included in version 1.37.3. The release is available on GitHub release. |
TL;DR
Moves the cloud-functions chart's default nvcf-api image from 1.28.0 to 1.28.2, so self-managed installs get the Cassandra replica-down fixes from #2918. No chart release so far carries them.
Additional Details (optional for docs, build, test, refactor, ci, chore, style, and revert PRs)
deploy/helm/cloud-functions/nvcf-api/Chart.yamlappVersionandvalues.yamlimage.tag:1.28.0->1.28.2. Nothing else changes.RangeReadRetryPolicy, so a range read timeout caused by a dead replica is retried once on another coordinator. Retriable Cassandra failures return 503 withRetry-After(gRPCUNAVAILABLE) instead of a 500 with CQL in it./infoversion stays correct after a config refresh).range_request_timeoutfrom Cassandra chart 0.23.2. Stacks should pin both.For the Reviewer
apipin to it in fix(self-managed): backport the NATS reconnect and Cassandra replica-down fixes [self-managed 1.1] #3013, together with Cassandra 0.23.2.For QA (optional for docs, build, test, refactor, ci, chore, style, and revert PRs)
helm lint deploy/helm/cloud-functions/nvcf-apipasses.kill -9the Cassandra process on one node. Expect no stall longer than about 5 s, and no 500s from read timeouts.Issues
Relates to #2893.
Checklist