upstream(mattermost-lts): update to 11.7 ESR; note restore fix, schemeid gotcha

- Mattermost now on 11.7 ESR (EOL 2027-05-15); 10.11 ESR expires 2026-08-15
- Latest patch: 11.7.5 (2026-06-18)
- Note: avoid 11.7.0–11.7.2 (schemeid bug upgrading from 10.11.17+)
- Backup/restore now uses pg_backup.sh (proper restore hook; PR #2/PR #1 fix)
- Next ESR expected ~Feb 2027

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Mbq9p2eCzZH59qfw9WsgKZ
This commit is contained in:
autonomic-bot
2026-06-19 03:53:48 +00:00
parent 15e88eaca2
commit 68bbfc72f2

View File

@ -6,21 +6,22 @@
| postgres | postgres | https://github.com/postgres/postgres | https://www.postgresql.org/docs/release/ |
## Standing notes
- mattermost-lts tracks the **10.11 ESR (Extended Support Release)**. Release calendar:
- **10.11** = current ESR, supported through **2026-08-15**; latest patch: **10.11.19** (2026-05-27)
- mattermost-lts tracks the **11.7 ESR (Extended Support Release)**. Release calendar:
- **10.11** = previous ESR, expires **2026-08-15** — migrated away 2026-06-19
- **11.7** = current ESR, supported through **2027-05-15**; latest patch: **11.7.5** (2026-06-18)
- **11.8.x** = innovation release (3-month lifecycle, EOL 2026-09-15) — do NOT target
- **10.12** = innovation release, support **EXPIRED December 2025** — do NOT target
- **11.7** = next ESR (supported through 2027-05-15) — future major upgrade, separate planning needed
- Check the release lifecycle at https://endoflife.date/mattermost before each upgrade run.
The `release-10.11` Docker Hub floating tag always points to the latest 10.11.x patch.
- Next ESR after 11.7 expected ~February 2027 (9-month cadence)
The `release-11.7` Docker Hub floating tag always points to the latest 11.7.x patch.
IMPORTANT: Do NOT use 11.7.011.7.2 — they have a schemeid migration bug in the `roles` table
when upgrading from 10.11.17+; use 11.7.3 or later (current: 11.7.5).
- postgres: stay on **15-alpine** for now. The upgrade to 16-alpine requires a major-version migration
(pg_upgrade or dump/restore with operator involvement) because the stack runs with a persistent
postgres_data volume. A 16 container against pg15 PGDATA crashes on startup. Major pg bump is a
separate operator-guided step.
- The backup/restore hooks are INLINED in compose.yml `deploy.labels` (not a separate pg_backup.sh
Docker config). This was intentional: using a Docker config caused a timing race where backupbot
could exec into the OLD postgres container (pre-config-update) during the upgrade lifecycle. Inline
commands eliminate this dependency — the backupbot reads labels from the Docker service API
(always current), and pg_dump/psql/createdb are present in any postgres:alpine container.
- Backup format (as of 2.1.12+10.11.19): gzip-compressed `backup.sql` in the postgres_data volume.
The restore hook properly terminates connections, drops + recreates the DB, then gunzips + restores.
`DROP DATABASE WITH (FORCE)` requires PostgreSQL 13+ — safe on postgres:15-alpine.
- Backup format (as of PR #2 / upcoming 3.0.0+11.7.5 release): `pg_backup.sh` Docker config on the
postgres service. Backup: `pg_dump | gzip → backup.sql`. Restore: terminate connections, FORCE-drop,
recreate DB, reimport dump. `DROP DATABASE WITH (FORCE)` requires PostgreSQL 13+ — safe on postgres:15-alpine.
The previous inline-label approach (no restore hook) was a defect: raw PGDATA restore without a reload
was a silent no-op. Fixed in PR #2 (restore fix cherry-picked from PR #1 ci/pg-restore).