Open davinchia opened 3 years ago
If we do #1 we should just clear the known logs directory/prefix, not the whole bucket.
Good point. Edited.
This can be done manually while upgrading pretty easily. This is for those who are using the default minio logging setup. Just make sure to delete the minio pvc before checking out the new version of airbyte and upgrading.
$ kubectl delete deployments airbyte-db airbyte-scheduler airbyte-server airbyte-temporal airbyte-webapp airbyte-minio
deployment.apps "airbyte-db" deleted
deployment.apps "airbyte-scheduler" deleted
deployment.apps "airbyte-server" deleted
deployment.apps "airbyte-temporal" deleted
deployment.apps "airbyte-webapp" deleted
deployment.apps "airbyte-minio" deleted
$ kubectl delete pvc airbyte-minio-pv-claim
persistentvolumeclaim "airbyte-minio-pv-claim" deleted
$ git checkout tags/{tag_name}
$ kubectl apply -k kube/overlays/stable
Tell us about the problem you're trying to solve
Today we do not perform checks whether logs already exist in the log bucket. This leads us to two error cases: 1) Application logs from previous runs get lumped into logs for the current run. Mostly inconvenient than anything else since users can scroll down to the latest log lines. 2) If the db is also reset, jobs can be assigned job paths where jobs already reside in. Also mostly inconvenient than anything else since users can scrolls down to the latest log lines.
Describe the solution you’d like
Either: 1) Empty the relevant paths (app-logging and job-logging) when we start. 2) Move the logs to another location.
Solution #1 seems fine to me.