[Bug]: Docker Compose preview deployments cause brief downtime on all other previews due to shared project name #9014
Description
Error Message and Logs
Found orphan containers ([worker-b41m0wpeftiwetp27jq0k1pv-pr-480 app-b41m0wpeftiwetp27jq0k1pv-pr-480 ...]) for this project.
Docker events during deploy show disconnect/reconnect churn on coolify network:
1773679620 disconnect coolify
1773679632 connect coolify
1773679640 disconnect coolify
All other running PR previews return 502/504 for several seconds during each new deploy.
Steps to Reproduce
1.Create a Docker Compose application with Traefik routing
2.Enable preview deployments with domain template {{pr_id}}.example.com
3.Deploy PR 1 — works fine
4.Deploy PR 2 — works, but PR 1 gets gateway timeout for several seconds
5.Deploy PR 3 — PR 1 and 2 both get brief gateway timeout
6.Pattern repeats with every new deploy
Example Repository URL
No response
Coolify Version
v4.0.0-beta.467
Are you using Coolify Cloud?
No (self-hosted)
Operating System and Version (self-hosted)
Ubuntu 24.04.3 LTS
Additional Information
ApplicationDeploymentJob.php → deploy_docker_compose_buildpack() sets --project-name to $this->application->uuid for ALL PR previews. The PR number is never appended. Service names, container names, and networks get PR-specific suffixes, but the project name does not.
Docker Compose sees other PRs' containers as orphans (same project, different services) and performs network reconciliation, briefly disconnecting them from the coolify network.
Related issue: #8030
--providers.docker.network=coolify in Traefik helps containers recover automatically but does not prevent the brief downtime
Workaround: manual docker restart coolify-proxy after each deploy