Describe the bug
In some cases, we will find that some unhealthy tablets are not scheduled, resulting in failure to be repaired all the time.
This may be because there are a large number of tablets in the cluster that cannot be repaired.
For example, a single-replica tablet encounters a disk damage.
However, these tablets still occupy the scheduling thread of the replica repair, so that other tablets
that can be repaired cannot be scheduled.
2条答案
按热度按时间qlzsbp2j1#
Improve the priority of the tablet which waiting for a long time?
9bfwbjaz2#
Set the tablet repair timeout, and move to the low-level advanced queue if the timeout expires, and wait for the next retry?