The 2026-05-24 apply was interrupted with the Helm release stuck in
pending-install, leaving only 2 of ~12 resources in TF state (any apply
errored "already exists"). Adopted the live resources back via import {}
sweep (namespace, tls-secret, uploads PVC, ESO ExternalSecret, both
ingresses, temporal Service, nfs backup PV+PVC) — plan now reaches zero.
Reconciled code to live reality (zero runtime change to running postiz):
- Removed kubernetes_deployment.temporal + kubernetes_job.temporal_search_
attr_cleanup: the temporal Deployment is gone from the cluster (only the
Service survives). Scheduled posts remain unavailable until temporal is
restored; immediate posting works.
- Removed helm_release.postiz from TF entirely: importing it would force a
helm upgrade (provider can't match merged values to config) and the
release is stuck pending-install. Left Helm-managed outside TF.
- Removed keel.sh/enrolled=true from the namespace (postiz was opted out of
Keel on 2026-05-29; this would have re-enrolled it on apply).
- Backup CronJob now dumps only the `postiz` DB (temporal/temporal_visibility
DBs don't exist) and no longer depends_on the removed helm_release.
Applied: 9 imported, 1 added (backup CronJob), 6 changed (benign), 0 destroyed.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
c2b820dc55