chore: remove seitask runner + Workflow scenarios (cutover)#442
Conversation
The Go-native integration harness (test/integration, one CronJob per target) has replaced the seitask-runner image + the Chaos-Mesh Workflow nightly. Delete the legacy machinery: - cmd/seitask/, internal/seitask/, internal/runner/, internal/taskruntime/, runner/ (the runner binary + its templates/RBAC), scenarios/ (the Workflow CR definitions). - the seitask image-build step in ecr.yml; the runner/runner-image/ runner-push Makefile targets + RUNNER_IMG (already stale paths). internal/keygen survives (general derivation, used by TestRelease); the SeiNodeTask CRD + its controller survive (driven by the SDK). go build, vet, test, and the integration build tag all pass with no dangling refs. Also drops the committed xreview ledgers (process artifacts, not code). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PR SummaryMedium Risk Overview CI/build: Drops the RBAC/docs/CRD copy: Strips Kept: SeiNodeTask CRD + controller, Reviewed by Cursor Bugbot for commit 1e3f5a8. Bugbot is set up for automated code reviews on this repo. Configure here. |
Drop the 'Go-native' framing and the 'replaces the Chaos-Mesh Workflow / seitask-runner' narrative from doc comments (incl. now-dangling refs to the deleted machinery); keep the technical rationale (chain-as-medium, the env-alias superset the release-test image reads). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Deleting internal/runner dropped its +kubebuilder:rbac markers for chaos-mesh workflows/workflownodes, which controller-gen had aggregated into the manager role. Regenerate so verify-generated passes; the manager never used that access (it was the runner's). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…nery Coral slate (k8s + idiom) on the deletion PR: - The SeiNodeTask CRD type doc comments named the seitask-runner / Chaos Workflow layer as where to express fan-out — and the spec.target one shipped into the served CRD description (kubectl explain would cite deleted machinery). Re-point to "the orchestrating caller (one task per node)" and regenerate config/crd + manifests. - Crisp the remaining dangling "scenario" / "scenarios/major-upgrade.yaml" references in upgrade_test.go + release_test.go (keep the technical rationale, drop the provenance trailers + the broken line-flow). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Idiom + k8s confirmation pass follow-ups: - Tidy the rewrapped TestChainUpgrade doc line-flow. - Re-point CLAUDE.md's Ready/Failed-exception rationale off "the seitask-runner depends on kubectl wait" to "consumers wait on" (the contract is unchanged; the named consumer was deleted). - Drop the deleted seitask from the SDK's consumer-naming comments (labels.go, readiness.go, render.go, sdk/CLAUDE.md) — seictl stays. - gitignore the stray local seitask build artifact. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
What
The Go-native integration harness has replaced the seitask-runner + Chaos-Mesh Workflow nightly (harness CronJobs deployed in platform#1207). This deletes the legacy machinery — 7k lines removed:
cmd/seitask/,internal/seitask/,internal/runner/,internal/taskruntime/,runner/,scenarios/ecr.yml; therunner*Makefile targets +RUNNER_IMGSurvives:
internal/keygen(general key derivation, used byTestRelease); the SeiNodeTask CRD + controller (now driven by the SDK'sRunTask).Safety
internal/seitask/internal/runner/internal/taskruntime(only comment mentions remain).go build ./...,go vet ./...,go test ./..., andgo test -c -tags integration ./test/integration/all pass.sei/seitask-runnerECR repo + IAM.🤖 Generated with Claude Code