Skip to content

fix: add sealed secrets#503

Open
MoeexT wants to merge 5 commits into
mainfrom
develop/sealed-secrets
Open

fix: add sealed secrets#503
MoeexT wants to merge 5 commits into
mainfrom
develop/sealed-secrets

Conversation

@MoeexT

@MoeexT MoeexT commented Jun 11, 2026

Copy link
Copy Markdown
Contributor

No description provided.

MoeexT added 5 commits June 9, 2026 10:37
- Makefile: move args file cleanup from datamate target to install target
- Makefile: milvus can now source /tmp/datamate-helm-args.sh before cleanup
- scripts/k8s/node-setup.sh: export HELM_MILVUS_TOLERATIONS in skip path
- deployment/helm/milvus/values.yaml: default tolerations to [] (no isolation)

Root cause: /tmp/datamate-helm-args.sh deleted in datamate-k8s-install,
before milvus-k8s-install could source it. When node isolation was active,
milvus received no tolerations and pods couldn't schedule on tainted nodes.

Also: Milvus chart had hardcoded tolerations (anchored YAML reference),
causing SchedulingDisabled even without node isolation configured.

Now supports all 4 scenarios:
- Open source + no isolation → no tolerations, any node (default) ✅
- Open source + isolation → tolerations applied to datamate + milvus ✅
- Commercial + no isolation → no tolerations ✅
- Commercial + isolation → tolerations for sealed-secrets + datamate + milvus ✅

Args file lifecycle:
  node-setup.sh creates → sealed-secrets sources → datamate sources →
  milvus sources → install target cleans up
- scripts/k8s/node-setup.sh: add etcd.tolerations and minio.tolerations
- Sub-charts don't inherit parent chart tolerations in Helm

Symptom: milvus-etcd and milvus-minio stuck Pending with
  '0/1 nodes are available: 1 node(s) had untolerated taint(s)'
  when node isolation is active

Root cause: --set-string tolerations[0]... only sets milvus main chart,
  but etcd and minio are sub-charts with their own tolerations: []

Fix: generate --set-string etcd.tolerations[0]... and
  --set-string minio.tolerations[0]... alongside milvus tolerations

Note: milvus-standalone CrashLoopBackOff is a Milvus v2.6.5 ARM64 bug
  (goroutine deadlock in streaming_node_manager), NOT node isolation
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant