Cloud installer alpha.185/186 fails on v1.22/v1.33: bootstrapper cannot resolve apecloud-pg

[khalildai]

Summary

Cloud Installer And Upgrade Test failed deterministically for both v2.3.0-alpha.185 and .186 on fresh installs for K3s v1.22 and v1.33. The upgrade path succeeds, so this is scoped to fresh bootstrap.

Evidence

• alpha.185 run: https://github.com/apecloud/apecloud-cd/actions/runs/29795201683
  • upgrade-test: success
  • installer-test-v1.22: failure at check kb-cloud-bootstrapper pod Completed
  • installer-test-v1.33: same failure
• alpha.186 run: https://github.com/apecloud/apecloud-cd/actions/runs/29796883103
  • upgrade-test: success
  • both fresh installer jobs fail at the same step

On v1.33, seven kb-cloud-bootstrapper-* pods remain Error. Their logs repeatedly show:

get meta db dsn: ...apecloud-pg.kb-cloud:5432/kubeblockscloud?sslmode=disable
psql: error: could not translate host name "apecloud-pg.kb-cloud" to address: Name does not resolve

There are no image-pull failures, disk pressure, or other kb-system / kb-cloud error pods in the collected diagnostics. The v1.22 job shows the same seven-bootstrapper Error pattern; log retrieval itself ends with kubelet EOF.

Expected

Fresh install should create/resolve the apecloud-pg.kb-cloud service before bootstrapper DB initialization, and bootstrapper should complete on both supported K3s versions.

Suggested investigation

1. Compare fresh-install ordering and PostgreSQL Service creation between alpha.184 and alpha.185.
2. Verify whether the service was renamed, moved, or gated while the bootstrapper DSN remained hard-coded.
3. Make bootstrapper wait for DNS/service readiness before running psql; preserve diagnostics for each failed pod.

Do not rerun .185/.186 unchanged: two consecutive releases and four fresh-install jobs reproduce the same signature.