mirror of
https://github.com/galaxyproject/galaxy.git
synced 2026-08-29 02:26:59 +08:00
mvdbeek
fbd1356674
``start_services.sh`` exported the Service ClusterIP and returned immediately after ``kubectl apply``, then the Run tests step started right away. The Postgres pod typically needs 30+ seconds to image-pull and start its listener; runs where Galaxy boots fast enough race it and hit ``connection refused``. When that happens the entire shard errors — every Galaxy boot dies talking to its DB before any test can run. Observed on Integration shard 1 of run 25398574715 (310 errors, 0 passes, first ERROR 35s after the apply). ``kubectl rollout status`` waits for the deployment, then ``kubectl wait Ready`` confirms the pod has actually transitioned out of ContainerCreating before tests start.