feat(self-managed): bump chart pins to the HA-enabled releases and icms 0.8.3 - #2178
Conversation
Pin the charts released from #2052 so the stack's PodDisruptionBudget, zone spread and rollout strategy values take effect. The previously pinned charts ignored them. Signed-off-by: shobham <[email protected]>
|
Navigate logical layers of code changes, visualize relationships, and explore their blast radius. No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository: NVIDIA/nvcf/.coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (2)
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 10 remain after this review. 📝 WalkthroughWalkthroughThe change updates the Cassandra dependency and chart versions for self-managed core services. It also updates the ICMS API chart appVersion and default SIS image tag to 0.8.3. ChangesHelm chart version updates
Priority: ➖ Normal Estimated code review effort: 2 (Simple) | ~8 minutes Change: Bug fix Merge Risk: ⚪ Minimal · up to The self-managed SIS release uses the updated 0.8.3 image default, and the inspected HA wiring is consistent with the intended behavior. No concrete issue remains that blocks merging. 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
0.8.3 includes #2175, which creates the NVCA NATS streams with the HA replica count. Signed-off-by: shobham <[email protected]>
|
🎉 This PR is included in deploy/helm/icms/v2.6.0 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
|
This PR is included in version 1.29.1. The release is available on GitHub release. |
TL;DR
Bump the self-managed stack's chart pins to the chart releases cut from #2052, and move the
sischart to icms image0.8.3, which includes #2175. Without this, the stack's HA values for PodDisruptionBudgets, zone spread and rollout strategy are ignored by the old pinned charts, so services such assis,notary-serviceandnvct-apirun 2 replicas but get no PDB, no zone spread and the default 25%/25% rollout.Additional Details
Stack pins (no template or values changes):
cassandraapi-keyssisapinvct-apiinvocation-servicegrpc-proxyratelimiteress-apinotary-serviceadmin-issuer-proxyrevalnats-auth-callout-servicellm-api-gatewayllm-api-gateway, the only change between the old and new version is feat(self-managed): self-hosted control-plane high availability #2052.llm-api-gateway1.5.0 also includes fix(llm-api-gateway): stop cutting off LLM responses after 60 seconds #2140 (stop cutting off LLM responses after 60 seconds) and test(llm-api-gateway): rely on stack defaults in BDD #1881 (test-only).sischart (deploy/helm/icms/icms-api): icms image0.7.2→0.8.3invalues.yaml(tag) andChart.yaml(appVersion). 0.8.3 includes fix(icms): create NVCA NATS streams with the HA replica count #2175 (NVCA NATS streams at the HA replica count). It also includes the icms releases in between, notably GPU gating (feat(gpu-gating): Added gpu gating support in ICMS #2032, fix(icms-gpu-gating): Added gpu gating in cluster management API #2078) and the instance lookup backfill task (fix(icms): add instance lookup backfill task #1995).sischart carrying icms 0.8.3. The stack'ssispin stays at 2.5.0 here because that chart does not exist until this merges; a one-line follow-up moves the pin to it. Until then, QA can setsis.image.tag: "0.8.3".For the Reviewer
Pin changes in
helmfile.d/01-dependencies.yaml.gotmplandhelmfile.d/02-core.yaml.gotmpl, plus the icms image tag in thesischart. Please confirm icms 0.8.3 (including GPU gating) is fine to ship in the self-managed stack.For QA
make -C deploy/stacks/self-managed testpasses, includingha-value-wiring.shandha-chart-render.sh.make -C deploy/helm/icms lintpasses, andhelm templaterenders thesiscontainer image as…:0.8.3.highAvailability.mode: preferred,kubectl get pdb -Ashould now also listsis,notary-serviceandnvct-api, and each replica-safe Deployment should showmaxSurge: 1/maxUnavailable: 0and atopology.kubernetes.io/zonespread rule.Issues
Relates to #2052
Checklist
Summary by CodeRabbit