coder fails to start when new version is released
I have coder running in a Kubernetes cluster, installed with Helm from https://helm.coder.com/v2
After there is a new release available, my pod will go into a crash-restart loop with an error about a missing postgres migration file. When I notice, I can upgrade the Helm release with the new version and everything starts fine.
It feels like the binary is auto-updating, then failing because of a missing migration.
If this is true, how do I stop the auto-update in my production install?
SOLVED:
My own fault - in my kubernetes install I was running an initContainer to set the initial admin user. The image was just ghcr.io/coder/coder - but I was not setting a version tag
So restarts on a new node without a container cache will pull the latest version and apply migrations, before running the original version.
After there is a new release available, my pod will go into a crash-restart loop with an error about a missing postgres migration file. When I notice, I can upgrade the Helm release with the new version and everything starts fine.
It feels like the binary is auto-updating, then failing because of a missing migration.
If this is true, how do I stop the auto-update in my production install?
SOLVED:
My own fault - in my kubernetes install I was running an initContainer to set the initial admin user. The image was just ghcr.io/coder/coder - but I was not setting a version tag
So restarts on a new node without a container cache will pull the latest version and apply migrations, before running the original version.
