nick song created KAFKA-10786:
---------------------------------
Summary: ReplicaAlterLogDirsThread gets stuck during the
reassignment of Kafka partition
Key: KAFKA-10786
URL: https://issues.apache.org/jira/browse/KAFKA-10786
Project: Kafka
Issue Type: Bug
Components: log
Affects Versions: 2.0.0
Reporter: nick song
Attachments: attachment 1.png, attachment 2.png, attachment 3.png
Topic config：Configs for topic 'athena_8603' are
leader.replication.throttled.replicas=9:7,9:6,10:8,10:7,8:6,8:5,11:9,11:8,follower.replication.throttled.replicas=9:13,10:0,8:15,11:14,retention.ms=86400000,delete.retention.ms=60000
Reassignment of replica athena_8603-1-15 is still in progress
When I reassigning the topic partition, I found that some tasks have been in
progress, lasting more than ten hours. After investigation, it was found that
ReplicaAlterLogDirsThread was running all the time and occupies a high CPU
usage rate (Attachment 1).
Check the thread information (Attachment 2) and find that the log data is being
copied. Check the log directory (Attachment 3) and find that the index of the
future directory is older than the original log. Is it because the
configuration delete.retention.ms=60000 caused the data to be deleted while
copying ? This causes the replication thread to get stuck. Is there any
solution?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
