infra/stacks/dawarich
Viktor Barzin cdb7d9a81a keel: sweep KEEL_LIFECYCLE_V1 + per-container KEEL_IGNORE_IMAGE across enrolled workloads
Every Keel-enrolled workload (policy=patch, match-tag=true, injected by the
inject-keel-annotations Kyverno policy) was fighting Terraform: Keel rewrites
the image tag and restamps keel.sh/update-time, change-cause and the rollout
revision on each poll; without ignore_changes every `tg apply` reverted those
— downgrading the image and forcing a spurious rollout that Keel then re-did.

Only llama-cpp had the full block (added 2026-05-24); the other ~73 workloads
drifted. This sweep adds, to every enrolled deployment/daemonset lifecycle:
  - container[N].image (one per container index + init_container[N]) # KEEL_IGNORE_IMAGE
  - keel.sh/match-tag, keel.sh/update-time, kubernetes.io/change-cause,
    deployment.kubernetes.io/revision  # KEEL_LIFECYCLE_V1

Verified via `tg plan` on speedtest (single-container: image downgrade
0.24.3->0.24.1 + annotation strip now gone) and changedetection (multi-container:
both container images no longer drift). AGENTS.md drift-suppression section
updated with the canonical block + marker legend.

fire-planner deferred (parallel session mid-apply per presence board).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-28 23:09:30 +00:00
..
.terraform.lock.hcl ig-poster: 69e395f2 + sync IMMICH_PG_* via ESO for CLIP scoring; postiz publish-notify n8n workflow 2026-05-09 13:16:24 +00:00
backend.tf [infra] Migrate Terraform state from local SOPS to PostgreSQL backend 2026-04-16 19:33:12 +00:00
main.tf keel: sweep KEEL_LIFECYCLE_V1 + per-container KEEL_IGNORE_IMAGE across enrolled workloads 2026-05-28 23:09:30 +00:00
providers.tf ig-poster: 69e395f2 + sync IMMICH_PG_* via ESO for CLIP scoring; postiz publish-notify n8n workflow 2026-05-09 13:16:24 +00:00
secrets [ci skip] Move Terraform modules into stack directories 2026-02-22 14:38:14 +00:00
terragrunt.hcl migrate all secrets from SOPS to Vault KV 2026-03-14 17:15:48 +00:00