Houliang Qi created IOTDB-1317:
----------------------------------
Summary: [Cluster] Log CatchUp always failed du to not check the
follower's match index
Key: IOTDB-1317
URL: https://issues.apache.org/jira/browse/IOTDB-1317
Project: Apache IoTDB
Issue Type: Bug
Components: Cluster
Reporter: Houliang Qi
Fix For: 0.12.1
When the follower lags behind the leader, it will catch up with the leader
through the log. When it can't find the log to catch up with in the memory of
the leader, it will go to the file on the local disk to find the satisfied log.
The logs found are between the *old match index* and the *commit index* saved
by the leader. However, at this time, the match index of the follower may have
been updated and the *old match index* of the follower may have been deleted
(in order to prevent a large number of logs from being kept on disk, there is a
log deletion mechanism). So at this time, if you send the above logs to the
follower, the follower will fail to find the matching log. At this time, this
storage group will appear to write stuck phenomenon.
The solution is to re-check the match index of the follower after reading the
log from the disk.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
