MaxBackups in the `ScheduledBackup` CRD not work

[tennix]

Bug Report

With tidb-operator 1.1.0-rc.1.p1, the field maxBackups in ScheduledBackup does not work as expected. After specifying maxBackups to 5, there are more than 5 ScheduledBackup objects and the corresponding backup directory in S3 bucket is also not deleted.

[tirsen]

If you plan to also delete files from S3 please note that the tidb-controller-manager can't access S3 because it generally does not have access to the secret for secret key auth or doesn't run with the correct service account for IAM role OIDC authentication. That needs to run as a pod in the namespace of the tidb cluster with the correct service account.

[DanielZhangQD]

Now it is run as a dedicated job starts with clean- and the service account in the backup CR will be set to the clean job, could you please help collect the tidb-operator log?

[shuijing198799]

If you plan to also delete files from S3 please note that the tidb-controller-manager can't access S3 because it generally does not have access to the secret for secret key auth or doesn't run with the correct service account for IAM role OIDC authentication. That needs to run as a pod in the namespace of the tidb cluster with the correct service account.

If you set MaxBackups and MaxReservedTime meanwhile, only MaxReservedTime is worked, like this code showed https://github.com/pingcap/tidb-operator/blob/master/pkg/backup/backupschedule/backup_schedule_manager.go#L254-L262 have you set the field MaxReservedTime ?

[tirsen]

I have not set MaxReservedTime. This is the CRD:

apiVersion: pingcap.com/v1alpha1
kind: BackupSchedule
metadata:
  name: backup-schedule
  namespace: ${cluster}
spec:
  maxBackups: 5
  #pause: true
  maxReservedTime: "3h"
  # 10 am UTC is 3 am PST which is off peak
  schedule: "0 10 * * *"
  # Use this for testing, every 2 mins
  #schedule: "*/2 * * * *"
  backupTemplate:
    backupType: full
    serviceAccount: tidb-backup-manager
    br:
      cluster: ${cluster}
      sendCredToTikv: false
      clusterNamespace: ${cluster}
      pd: ${cluster}-pd.${cluster}
      logLevel: debug
      # statusAddr: ${status_addr}
      # concurrency: 4
      # rateLimit: 0
      # timeAgo: ${time}
      # checksum: true
    from:
      host: ${cluster}-tidb.${cluster}.svc.cluster.local
      port: 4000
      user: ${SQM_SERVICE}_adm
      secretName: ${cluster}-tidb-backup-secret
      tlsClient:
        enabled: true
    s3:
      provider: aws
      region: us-west-2
      bucket: ${bucket}
      prefix: ${service}

[tirsen]

lol yeah I have set maxReservedTime

[tirsen]

I've confirmed this works correctly. Closing the issue.

[tirsen]

Oh I can't close it. You can close this issue. User error. :-)

[tennix]

OK, I'll close it.