Open duyleekun opened 6 years ago
Does this happen for each push? Does this happen even if you delete the gitkubed
pod from the kube-system
namespace (hence creating a new pod) ?
Once it's not clean, subsequent failed push won't clear the file (it's failed because of that not deleted file)
I resolve it by exec into the gitkube pod directly and rm the directory. Next time, I will delete the pod instead
Although It's not directly related to this issue, but my guess about failure reason to rm -rf $BUILD_ROOT
is following error I got from time to time. Retry the push resolve the error.
remote: Updating Kubernetes deployment: appworker
remote: deployment "appworker" image updated
remote: Waiting for rollout to finish: 1 old replicas are pending termination...
remote:
remote:
remote: Updating deployment failed: mofferworker
remote: $ Check deployment logs
For some reason, while executing pre-receive, my
BUILD_ROOT
dir isn't empty. My guess is that the hook doesn't always execute at the endrm -rf $BUILD_ROOT
and thegit archive $DEPLOY_NEWREV | tar -x -C $BUILD_ROOT
doesn't perform any clean up.