Skip to content

Updated CI - #291

Merged
lecovi merged 13 commits into
masterfrom
upgrade-ci
Oct 3, 2026
Merged

lecovi merged 13 commits into
masterfrom
upgrade-ci

Conversation

@facundobatista

Copy link
Copy Markdown
Member

No description provided.

facundobatista and others added 13 commits September 28, 2026 19:51
Without it, the initdb self-restart on first boot was counted as an
immediate failure, marking the container unhealthy before docker
compose up --wait could give it a chance.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
docker compose up --wait failed the healthcheck almost instantly,
ignoring both start_period and retries, so drop the healthcheck and
go back to polling pg_isready via exec, silencing the transient
errors while postgres restarts during initdb.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…DNS.

The Run tests job still failed to resolve "postgres" from web even
after postgres was confirmed ready, pointing at a known compose v2
embedded-DNS bug rather than a startup race. links writes the IP into
/etc/hosts at container creation, sidestepping the DNS resolver.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Both timing fixes and links failed to resolve the persistent
"postgres" DNS failure; need actual data on web's network config
before guessing further.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found the actual root cause: docker/compose#10668, a real race in
compose v2 where a container can start before its network attach
finishes when services are brought up in parallel (the default).
--parallel 1 serializes container startup and avoids the race, so
revert the static-IP workaround and go back to the plain "postgres"
hostname.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
--parallel is a top-level docker compose flag, not an "up" subcommand
flag.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
--parallel 1 didn't fix the DNS failure either, so container->container
networking itself is unreliable in this CI environment regardless of
timing. Put postgres and mail in web's netns (network_mode: service:web)
so everything talks over localhost, sidestepping compose networking
and DNS altogether.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Connection to localhost:5432 is now refused instead of unresolved,
so the network path is fine but postgres itself isn't listening;
need its actual startup log to see why.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found the real bug: postgres:latest now resolves to postgres 18+,
whose image refuses to start against a volume mounted directly at
/var/lib/postgresql/data (it wants /var/lib/postgresql instead), so
it was crash-looping the whole time. Every DNS/connection error we
chased was just a symptom of postgres never actually being up.
Pinning to postgres:16 fixes that, so drop the now-unneeded
network_mode/static-IP/localhost workarounds and the debug step.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@lecovi lecovi self-assigned this Oct 3, 2026
@lecovi

lecovi commented Oct 3, 2026 •

Copy link
Copy Markdown
Collaborator

@facundobatista tenés idea por qué no puedo mergearlo? Veo que en el check dice:

At least 1 approving review is required by reviewers with write access.

Yo tengo acceso de "admin", no entiendo por qué no me deja... 😖

@lecovi
lecovi self-requested a review October 3, 2026 18:11

@lecovi lecovi left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@lecovi
lecovi merged commit 60adbd5 into master Oct 3, 2026
1 check passed
@lecovi
lecovi deleted the upgrade-ci branch October 3, 2026 18:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants