fix(deploy): gate test/staging deploys on bootstrap convergence, not just health #186

Merged

haydar merged 1 commit from fix/test-deploy-bootstrap-gate into develop 2026-10-30 19:24:22 +00:00

Conversation 0 Commits 1 Files changed 10 +653 -58

haydar commented 2026-10-30 19:17:39 +00:00

Member

• Copy link
• Copy Markdown

Test deploy run #609 deployed healthy images yet served the previous
release's Docs theme: the one-shot bootstrap container (restart: "no")
never provably completed its Wiki.js stage, and nothing in the pipeline
distinguished a finished convergence from a dead, skipped, or
still-running one — a healthy app only proves the app started.

• run.sh: emit a 'DWB bootstrap convergence complete' sentinel as the
  last line; stage failures already exit non-zero (set -eu, explicit
  exit 1, wikijs 3-attempt retry).
• wikijs image: bake DEVWORKBASE_SOURCE_REVISION (Dockerfile ARG/ENV,
  release.py build-arg); wikijs-bootstrap.sh stamps it into injectHead
  as the dwb-bootstrap-revision meta in the same theme PUT as
  dwb-theme-build and verifies it reaches the served page.
• verify_docs_theme.py: require --expect-revision for a convergence
  verdict and name the failing stage — converged / bootstrap-missing /
  bootstrap-stale / theme-stored-old / image-old (via --inspect Coolify
  app metadata) / served-truncated / no-fingerprint — with a diagnosis
  block carrying expected vs served fingerprint and revision.
• compose.prod.yml: document the re-run contract — the per-release
  digest pin is what forces recreate + start of the one-shot on every
  deploy; an exited restart:no container neither retries nor proves
  completion.
• test-deploy/staging workflows: pass --expect-revision github.sha and
  --inspect, raise --wait to 900s for multi-stage convergence time.
• scripts/test_deploy_convergence.py: end-to-end run.sh against a
  stubbed docker CLI (sentinel on success, non-zero + no sentinel on
  stage failure) and the diagnose() stage matrix; wired into the
  release-tests CI loop and release.py run_tests.

Test deploy run #609 deployed healthy images yet served the previous release's Docs theme: the one-shot bootstrap container (restart: "no") never provably completed its Wiki.js stage, and nothing in the pipeline distinguished a finished convergence from a dead, skipped, or still-running one — a healthy app only proves the app started. - run.sh: emit a 'DWB bootstrap convergence complete' sentinel as the last line; stage failures already exit non-zero (set -eu, explicit exit 1, wikijs 3-attempt retry). - wikijs image: bake DEVWORKBASE_SOURCE_REVISION (Dockerfile ARG/ENV, release.py build-arg); wikijs-bootstrap.sh stamps it into injectHead as the dwb-bootstrap-revision meta in the same theme PUT as dwb-theme-build and verifies it reaches the served page. - verify_docs_theme.py: require --expect-revision for a convergence verdict and name the failing stage — converged / bootstrap-missing / bootstrap-stale / theme-stored-old / image-old (via --inspect Coolify app metadata) / served-truncated / no-fingerprint — with a diagnosis block carrying expected vs served fingerprint and revision. - compose.prod.yml: document the re-run contract — the per-release digest pin is what forces recreate + start of the one-shot on every deploy; an exited restart:no container neither retries nor proves completion. - test-deploy/staging workflows: pass --expect-revision github.sha and --inspect, raise --wait to 900s for multi-stage convergence time. - scripts/test_deploy_convergence.py: end-to-end run.sh against a stubbed docker CLI (sentinel on success, non-zero + no sentinel on stage failure) and the diagnose() stage matrix; wired into the release-tests CI loop and release.py run_tests.

haydar added 1 commit 2026-10-30 19:17:40 +00:00

fix(deploy): gate test/staging deploys on bootstrap convergence, not just health ... 32789bfe6e

Test deploy run #609 deployed healthy images yet served the previous
release's Docs theme: the one-shot bootstrap container (restart: "no")
never provably completed its Wiki.js stage, and nothing in the pipeline
distinguished a finished convergence from a dead, skipped, or
still-running one — a healthy app only proves the app started.

- run.sh: emit a 'DWB bootstrap convergence complete' sentinel as the
  last line; stage failures already exit non-zero (set -eu, explicit
  exit 1, wikijs 3-attempt retry).
- wikijs image: bake DEVWORKBASE_SOURCE_REVISION (Dockerfile ARG/ENV,
  release.py build-arg); wikijs-bootstrap.sh stamps it into injectHead
  as the dwb-bootstrap-revision meta in the same theme PUT as
  dwb-theme-build and verifies it reaches the served page.
- verify_docs_theme.py: require --expect-revision for a convergence
  verdict and name the failing stage — converged / bootstrap-missing /
  bootstrap-stale / theme-stored-old / image-old (via --inspect Coolify
  app metadata) / served-truncated / no-fingerprint — with a diagnosis
  block carrying expected vs served fingerprint and revision.
- compose.prod.yml: document the re-run contract — the per-release
  digest pin is what forces recreate + start of the one-shot on every
  deploy; an exited restart:no container neither retries nor proves
  completion.
- test-deploy/staging workflows: pass --expect-revision github.sha and
  --inspect, raise --wait to 900s for multi-stage convergence time.
- scripts/test_deploy_convergence.py: end-to-end run.sh against a
  stubbed docker CLI (sentinel on success, non-zero + no sentinel on
  stage failure) and the diagnose() stage matrix; wired into the
  release-tests CI loop and release.py run_tests.

haydar scheduled this pull request to auto merge when all checks succeed 2026-10-30 19:19:19 +00:00

haydar merged commit 6da0c377e2 into develop 2026-10-30 19:24:22 +00:00

haydar referenced this pull request from a commit 2026-10-30 19:24:24 +00:00

Merge pull request 'fix(deploy): gate test/staging deploys on bootstrap convergence, not just health' (#186) from fix/test-deploy-bootstrap-gate into develop

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

developing4all/devworkbase!186

No description provided.