[marcleblanc2]

Problem

• A customer migrating from Kustomize to Helm needs two things the chart cannot do today: init containers on any pod, and extra ports on Services. See the Slack thread

Solution

• <serviceName>.extraInitContainers on all 21 workloads

  • Rendered before any init container the chart defines

• <serviceName>.extraServicePorts appended to each component's primary Service

• Both are toYaml passthrough lists that render nothing when unset

• grafana.initContainers is replaced by grafana.extraInitContainers

• The secondary Services sourcegraph-frontend-internal, indexed-search-indexer, and worker-executors do not get extraServicePorts. Frontend-internal already listens on port 80, so an extra port 80 on frontend would otherwise collide with it

Tested

• New tests/extraInitContainersServicePorts_test.yaml sets both values on every component at once and asserts: the extra init container is index 0 on all 21 workloads and the chart's own init container is index 1 where it has one, no initContainers key renders by default, the port is appended on all 20 primary Services, and the 3 secondary Services are unchanged

• helm unittest charts/sourcegraph: 38 suites, 234 tests pass. helm lint clean. ./scripts/helm-docs.sh leaves no diff. helm template with default values is byte-identical to origin/main

• kubeconform -strict passes on the default render and on a render with both values set across frontend, otel, grafana, worker, and pgsql

• Deployed to a test instance on EKS with both values set on every deployed component: frontend, gitserver, grafana, indexed-search, precise-code-intel-worker, prometheus, searcher, syntect-server, worker

  • Every pod ran its extra init container to Completed; kubectl logs <pod> -c extra-init-check shows its output on each

  • On frontend, the extra init container ran before migrator, and a file it wrote to an extraVolumes emptyDir is readable from the frontend container

  • On grafana, a real native sidecar (restartPolicy: Always, mints RDS IAM auth tokens) runs through grafana.extraInitContainers

  • Every primary Service shows the extra port, and curl from another pod in the namespace returns 200 through it on frontend, grafana, prometheus, syntect-server, and worker

  • All pods Ready, site healthy

Amp threads

• https://ampcode.com/threads/T-c9f7829a-ccaa-4659-ad5a-dc27907b61a8 (assessment of the customer's Kustomize overlay)
• https://ampcode.com/threads/T-01a1902b-ece9-725c-863d-3e2bba35ee1d (implementation and EKS test)

Changelog

• Added <serviceName>.extraInitContainers to every workload
• Added <serviceName>.extraServicePorts to every component's primary Service

Checklist

• Follow the manual testing process
• Update changelog
• Update Kubernetes update doc

[github-actions]

Open the Diff Tour for this PR in Sourcegraph

[filiphaftek]

I run agent to search for breaking change, but looks good :+1