Details
-
Bug
-
Status: Resolved
-
Minor
-
Resolution: Fixed
-
3.8.0, 3.7.1
Description
In MirrorCheckpointTask#start we call Scheduler#execute, which blocks for up to admin.timeout.ms waiting for the task to complete. Since restoring the checkpoint and offset sync state may take some time, and the rest of the task is prepared for this, we should replace the Scheduler#execute call with a single-shot asynchronous method that starts the job once but does not wait for it to complete.
Attachments
Issue Links
- links to