fix(hwlab): clean stale statefulset pods during node sync (#727)

Co-authored-by: Codex <codex@noreply.local>
2026-06-23 09:27:28 +08:00
parent c987df66f1
commit 1b3af54fba
3 changed files with 47 additions and 0 deletions
@@ -60,6 +60,7 @@ export function hwlabNodeHelp(): Record<string, unknown> {
    notes: [
      "`trigger-current` skips an already succeeded PipelineRun for the same HWLAB source commit by default.",
      "`trigger-current --confirm --wait` is the one-command CICD path: git-mirror pre-sync/pre-flush, control-plane refresh, PipelineRun create/reuse, bounded wait, and post-flush when terminal.",
+      "`control-plane sync --confirm` terminates a stale running Argo operation, deletes failed Argo hook Jobs, and recreates stale non-ready StatefulSet pods that are still pinned to an old controller revision with pull/backoff errors.",
      "`--wait` defaults to 120 seconds. If the PipelineRun is still active after 120 seconds, the CLI returns a warning plus env-reuse and git-mirror inspection commands instead of blocking.",
      "Use `--rerun` for a deliberate YAML-first config-only publish when UniDesk node/lane render inputs changed but the HWLAB source commit did not."
    ],