fix(deploy): bump postgres to 18-alpine and move volume mount#41
Merged
Conversation
Renovate PR #11 bumps the postgres docker tag to v18 but leaves the volume mount at /var/lib/postgresql/data. The 18+ official images changed the on-disk layout so PostgreSQL data lives in a major-version- specific subdirectory under /var/lib/postgresql, and mounting the old "data" path now triggers an "unused mount/volume" startup error (see docker-library/postgres#1259): postgres-1 | Error: in 18+, these Docker images are configured to store database data in a format which is compatible with "pg_ctlcluster" ... Counter to that, there appears to be PostgreSQL data in: /var/lib/postgresql/data (unused mount/volume) The fix is to mount the parent directory (the new convention) so pg18 manages the per-version data subdir itself. The Helm chart is unaffected — it provisions postgres via cloudnative-pg, which manages its own storage layout. BREAKING CHANGE for local dev: existing `postgres_data` volumes from pg16 must be wiped before this change takes effect: docker compose -f deploy/docker-compose.yml \ -f deploy/docker-compose.dev.yml down -v CI starts each run with a fresh volume so it is unaffected.
3 tasks
Haibread
added a commit
that referenced
this pull request
May 10, 2026
…ge (#56) The Helm chart was still pinning the CNPG cluster to postgres 16 while the docker-compose stack had moved to postgres:18-alpine in PR #41. Project audit flagged this as a P0 — version drift between local dev and a Helm-managed cluster means schema and tooling behaviour differ in production from what we test in dev. - deploy/helm/ai-registry/values.yaml: postgresVersion "16" → "18". Renders to ghcr.io/cloudnative-pg/postgresql:18 in the CNPG Cluster manifest (verified via `helm template`). Comment now documents the parity with PR #41 so the next bump understands the relationship. - docs/runbook.md: the two `pg-probe` snippets that spin up an ephemeral pod for psql-into-the-cluster troubleshooting were pulling postgres:16-alpine. Bumped to 18-alpine to match. Verified: `helm lint` passes, `helm template` produces `imageName: ghcr.io/cloudnative-pg/postgresql:18` for the Cluster resource. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Merged
3 tasks
Haibread
added a commit
that referenced
this pull request
May 14, 2026
…gn, ADR 0001, UI/UX plan (#60) Sweep over the project's markdown to bring it back in line with the post-Phase-7 + post-audit codebase. Documentation was lagging in multiple places — the project audit's docs front had flagged most of this; this PR ships the fixes. CHANGELOG.md Repointed the `Unreleased` section to cover BOTH the existing Phase 7 access-control + change-approval bundle (PRs #28-#32 + #37) AND the project-audit follow-ups (PRs #52-#59) that just landed. The audit follow-ups subsection summarises each merged PR with its scope and verification, plus a deferred-items list (DisallowUnknownFields, rate-limiter janitor, per-handler child spans, eager markdown chunk). Added a note that a real version stamp is overdue. CLAUDE.md - Tech stack: drop the "sqlc or pgx" hedge — the codebase is pgx + hand-written SQL only. - Auth bullet: stop claiming hashed API keys are supported; they're parked under v0.4.x. Mirrored on Decision B in the table. - Deployment bullet: stop claiming docker-compose has dev + prod profiles. Only dev + ci exist; the prod profile is parked. - Conventions: drop the obsolete `claude/ai-registry-setup-KMC3l` bootstrap branch reference; describe the actual feat/fix/docs/chore convention. - Configuration section: "**both** of the following" → "**all three**" (env / YAML / default). Original wording predated the YAML layer and was only updated on the precedence list, not the lede. PLAN.md - Phase 5 hardening section: the four-line `**TODO — Phase 5:**` list was misleading — the items weren't unfinished Phase 5 work, they were carried forward into v0.4.x. Replaced with a "Parked from Phase 5 (now tracked under v0.4.x)" header that points at the README + CLAUDE.md Decision B for the live status. - v0.2.2 section: the entire DoD checklist was rendered as `- [ ]` even though the section header said "✅ SHIPPED". Replaced the unchecked checklist with a past-tense "What landed" summary, a "Carried forward" subsection for the OTel-spans gap that v0.2.2 only partially closed (PR #58 finished it), and a DoD-met note. README.md - Tech stack: PostgreSQL 16 → 18, matching the dev compose (postgres:18-alpine since PR #41) and the Helm CNPG cluster (PR #56). - Infra bullet: replace the "docker-compose (dev / ci / prod)" claim with an accurate description of the two real overlays (dev, ci) and a parked-under-v0.4.x note for prod. design.md - Typography table: drop the "Geist (next/font)" row — that loader was removed with the rest of the Next.js stack in Phase 6 (ADR 0004). The web app uses Tailwind's default `font-sans` / `font-mono` system stacks; explained in a follow-up paragraph. - Admin sidebar ASCII diagram: add the Workspaces nav item that Phase 7 introduced; rename "Activity" → "Audit" to match the actual nav label; flag the API Keys item as a placeholder. docs/adr/0001-workspaces-under-publishers.md - Drop the "Migration numbers ... are **placeholders**" preamble — the actual numbered migrations (000008, 000009, 000010) are on disk and the placeholder language is no longer accurate. - Mark Step 1 + Step 2 as shipped with the actual entry points (Go-side `db.BackfillWorkspaces` instead of the original `make backfill-workspaces`). - Add an explicit "Status note (2026-05-10)" callout that Step 3 (the finalising migration that drops `publisher_id` and flips `workspace_id` to NOT NULL) was scoped at design time but never landed; production keeps both columns coexisting and code paths still read `publisher_id` directly. Verified by grep against `internal/store/mcp.go`. docs/ui-ux-implementation-plan.md - Added a "Status: largely shipped — retrospective document" banner at the top. The document plans 10 batches, the vast majority of which shipped across v0.2.x and v0.3.x. Without the banner the file reads as forward-looking work and gives a misleading picture to anyone landing on it via search. Points the reader at PLAN.md and README.md for current open work. Out of scope (audit findings deliberately not addressed here): docs/ui-ux-proposals.md is a decision record (proposals + accepted / deferred verdicts), so staleness is by design — not a bug. The runbook, db-backup, future-multi-environment, ADRs 0002/0003/0004, and test/load README are all current. No code changes; no test impact. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Renovate #11 bumps the postgres docker tag to v18 but leaves the volume mount unchanged. The 18+ official images changed the on-disk layout — PostgreSQL data now lives in a major-version-specific subdirectory under `/var/lib/postgresql` — and mounting the old `/var/lib/postgresql/data` path now triggers an `unused mount/volume` startup error (see docker-library/postgres#1259).
The actual error from CI:
```
postgres-1 | Error: in 18+, these Docker images are configured to store database data in a
format which is compatible with "pg_ctlcluster" ... Counter to that, there appears to
be PostgreSQL data in: /var/lib/postgresql/data (unused mount/volume)
```
(This was only visible because #38 fixed the dump-logs step in the E2E workflow earlier today.)
Fix: mount the parent directory `/var/lib/postgresql` so pg18 manages its per-version data subdir. Applied to both `docker-compose.yml` and `docker-compose.dev.yml`. The Helm chart is unaffected — it provisions postgres via cloudnative-pg which manages its own storage.
Existing `postgres_data` volumes from pg16 must be wiped before this change takes effect:
```
docker compose -f deploy/docker-compose.yml -f deploy/docker-compose.dev.yml down -v
```
CI starts each run with a fresh volume, so it's unaffected.
Test plan