[Improvement] tidb pitr improvements

[cjc7373]

In tidb addon, I use br tool to do backup and restore. During the development of pitr restore, there are some problems that needs improvement of kubeblocks.

Problem 1

br does full backup restore and pitr restore in one command. But they are separated jobs in kubeblocks.

I suggest add an option to merge full backup restore and pitr restore jobs. And the full backup name should be passed as an env variable to the pitr restore job.

• feat: add an option to merge full and pitr restore job #9404

Problem 2

The new log backup job is started before restore finished. br has a restriction that a only one log backup/restore job can be run at the same time.

I suggest check the cluster restore annotation in the backup controller, and delay backup job creation when the cluster is restoring.

• chore: delay backup job when restore is in progress #9400

Problem 3

tidb addon's restore job is issued to pd component, then pd executes restore command in all tikv nodes. So that when doing restore, tikv nodes must be all running. There's a DoReadyRestoreAfterClusterRunning option in the restore annotaion that does exactly what it says. However, this option needs to be set every in restore job. I suggest change it to an actionset-level option.

• chore: add an actionset api to delay postReady job after cluster running #9399

EDIT: I just found backup job will inherit target pod's env. So I could just target tidb pod to run backup script (and get pd info based on the env). This problem is no longer valid.

[github-actions]

This issue has been marked as stale because it has been open for 30 days with no activity