Fix backup/restore #14

2026-06-02T19:23:46Z

notplants commented

2026-06-02 19:23:46 +00:00

I have been using Salami (some call it llm) to make automated tests to test backup and restore for co-op cloud recipes (more info on this coming soon), and have been finding some bugs in backup/restore that don't actually work and probably have not been tested

this recipe was one of them. before the fix, the restore was appearing to restore but was I believe a silent no-op. this PR has a fix for it (based on pg_backup.sh of other recipes, with one additional safeguard of killing any live connections to the database before doing the database recreate)

notes:

also increased postgres version (I wanted to use pgautoupgrade, but it looks like the bitnami image bundles a plugin that requires pgvector as the database... and couldn't find a version of pgautoupgrade that also had pgvector without a custom image... so including pgvector as the image and using the homegrown auto-upgrader seems like the easiest path for now... until perhaps in a future PR switching to a non-bitnami image)
also increased start_period for healthcheck to make automated tests easier
https://hub.docker.com/r/bitnami/discourse now requires a paid subscription. this PR switches to use https://hub.docker.com/r/bitnamilegacy/discourse, which also says it will no longer be updated.
... probably we should switch to using https://github.com/discourse/discourse_docker published somewhere else... but will save that for another PR as this was a few other bits

I have been using Salami (some call it llm) to make automated tests to test backup and restore for co-op cloud recipes (more info on this coming soon), and have been finding some bugs in backup/restore that don't actually work and probably have not been tested this recipe was one of them. before the fix, the restore was appearing to restore but was I believe a silent no-op. this PR has a fix for it (based on pg_backup.sh of other recipes, with one additional safeguard of killing any live connections to the database before doing the database recreate) notes: - also increased postgres version (I wanted to use pgautoupgrade, but it looks like the bitnami image bundles a plugin that requires pgvector as the database... and couldn't find a version of pgautoupgrade that also had pgvector without a custom image... so including pgvector as the image and using the homegrown auto-upgrader seems like the easiest path for now... until perhaps in a future PR switching to a non-bitnami image) - also increased start_period for healthcheck to make automated tests easier - https://hub.docker.com/r/bitnami/discourse now requires a paid subscription. this PR switches to use https://hub.docker.com/r/bitnamilegacy/discourse, which also says it will no longer be updated. ... probably we should switch to using https://github.com/discourse/discourse_docker published somewhere else... but will save that for another PR as this was a few other bits

notplants added 3 commits 2026-06-02 19:23:47 +00:00

chore: upgrade to 0.8.0+3.5.0 0f873433ba

fix(backup): add pg_backup.sh + proper backup/restore hooks, 20m start_period ec7bbdf786

improved comments

continuous-integration/drone/pr Build is failing

Details

5091fd999e

notplants marked the pull request as work in progress 2026-06-02 19:55:06 +00:00

notplants added 1 commit 2026-06-02 20:18:52 +00:00

fix(db): switch postgres image to pgvector/pgvector:pg17 + bump PG_BACKUP_VERSION

continuous-integration/drone/pr Build is failing

Details

continuous-integration/drone/tag Build is passing

Details

b0f9ae743a

notplants commented

2026-06-02 20:20:16 +00:00

I realize this PR contains a few different things and might be hard to read.

If someone else is also wanting to collab on maintaining this recipe, then I can break it into parts. Or I can just merge it as its probably all stuff that should happen at some point.

I realize this PR contains a few different things and might be hard to read. If someone else is also wanting to collab on maintaining this recipe, then I can break it into parts. Or I can just merge it as its probably all stuff that should happen at some point.

notplants marked the pull request as ready for review 2026-06-02 20:41:56 +00:00

notplants commented

2026-06-02 20:56:39 +00:00

cc recent committers if you want to weigh in @decentral1se @jeppebundsgaard @3wordchant

notplants merged commit 7e4d85a810 into main

2026-06-15 17:33:39 +00:00

notplants referenced this issue from a commit

2026-06-15 17:33:45 +00:00

Merge pull request 'Fix backup/restore' (#14) from fix-restore into main

notplants deleted branch fix-restore

2026-06-15 17:33:45 +00:00

notplants referenced this pull request

2026-06-16 18:33:30 +00:00

fix(db): make pg_upgrade 13->17 idempotent and use the cluster's real install user #15

notplants commented

2026-06-16 18:34:26 +00:00

ran into an issue in the old bespoke upgrade script while deploying this change to an old deployment (with a different database install user). created a fix and explanation here: #15 and releasing a patch release shortly

ran into an issue in the old bespoke upgrade script while deploying this change to an old deployment (with a different database install user). created a fix and explanation here: https://git.coopcloud.tech/coop-cloud/discourse/pulls/15 and releasing a patch release shortly

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: coop-cloud/discourse#14