[
https://issues.apache.org/jira/browse/CASSANDRA-14953?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
HUANG DUICAN updated CASSANDRA-14953:
-------------------------------------
Description:
We found that Cassandra has a lot of write accumulation in the production
environment, and our business has experienced a lot of write failures.
Through the system.log, it was found that MemtableReclaimMemory was pending at
the beginning, and then a large number of MutationStage stacks appeared at a
certain moment.
Finally, the heap memory is full, the GC time reaches tens of seconds, the
node status is DN through nodetool, but the Cassandra process is still
running.We killed the node and restarted the node, and the above situation
disappeared.
Also the number of Active MemtableReclaimMemory threads seems to stay at 1.
!image-2019-01-05-11-36-31-199.png!
a large number of MutationStage stacks appeared at a certain moment.
!image-2019-01-05-11-37-54-253.png!
long GC time:
!image-2019-01-05-11-38-21-711.png!
Why is this happening?
was:
We found that Cassandra has a lot of write accumulation in the production
environment, and our business has experienced a lot of write failures.
Through the system.log, it was found that MemtableReclaimMemory was pending at
the beginning, and then a large number of MutationStage stacks appeared at a
certain moment.
Finally, the heap memory is full, the GC time reaches tens of seconds, the node
status is DN through nodetool, but the Cassandra process is still running.We
killed the node and restarted the node, and the above situation disappeared.
Also the number of Active MemtableReclaimMemory threads seems to stay at 1.
!image-2019-01-05-11-23-23-752.png!
a large number of MutationStage stacks appeared at a certain moment.
!image-2019-01-05-11-26-56-546.png!
long GC time:
!image-2019-01-05-11-28-00-371.png!
Why is this happening?
> Failed to reclaim the memory and too many MemtableReclaimMemory pending task
> ----------------------------------------------------------------------------
>
> Key: CASSANDRA-14953
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14953
> Project: Cassandra
> Issue Type: Bug
> Components: Local/Memtable
> Environment: version : cassandra 2.1.15
> jdk: 8
> os:suse
> Reporter: HUANG DUICAN
> Priority: Major
> Attachments: cassandra_20190105.zip
>
>
> We found that Cassandra has a lot of write accumulation in the production
> environment, and our business has experienced a lot of write failures.
> Through the system.log, it was found that MemtableReclaimMemory was pending
> at the beginning, and then a large number of MutationStage stacks appeared at
> a certain moment.
> Finally, the heap memory is full, the GC time reaches tens of seconds, the
> node status is DN through nodetool, but the Cassandra process is still
> running.We killed the node and restarted the node, and the above situation
> disappeared.
>
> Also the number of Active MemtableReclaimMemory threads seems to stay at 1.
> !image-2019-01-05-11-36-31-199.png!
> a large number of MutationStage stacks appeared at a certain moment.
> !image-2019-01-05-11-37-54-253.png!
>
> long GC time:
> !image-2019-01-05-11-38-21-711.png!
>
> Why is this happening?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
