Description
I restarted a tablet server during a stress workload, and when it came back up, it failed to start the tablet. The tablet had several hundred pending REPLICATE messages, which were enough that they wouldn't all fit in-flight at the same time. Thus, re-submitting the pending replicates failed, and the tablet marked itself as having failed to start.
Attachments
Issue Links
- relates to
-
KUDU-1779 Consensus "stuck" with all transaction trackers are at limit
- Open