Radim Kolar created CASSANDRA-4690:
--------------------------------------
Summary: Delete only workload OOM cassandra
Key: CASSANDRA-4690
URL: https://issues.apache.org/jira/browse/CASSANDRA-4690
Project: Cassandra
Issue Type: Bug
Components: Core
Affects Versions: 1.0.11
Environment: OpenJDK 6
Reporter: Radim Kolar
Priority: Blocker
Restart cassandra node and run delete only workload against CF. No other update
operations like insert or delete are used against that CF.
Cassandra do not counts deletes in serialized/live bytes. It counts them just
as ops. After some time cassandra will run almost out of Java heap space and
tries to do emergency flush, which will not be done at CF with deletes because
cassandra thinks that it has 0 live bytes. This will result in OOM after while,
but usually cassandra just hangs.
To make this workload finish, you have to issue periodic nodetool flush
commands. Their flueshes on that CF looks:
INFO [RMI TCP Connection(69)-64.6.108.239] 2012-09-19 18:22:46,366
ColumnFamilyStore.java (line 705) Enqueuing flush of
Memtable-sipdb@359811109(0/0 serialized/live bytes, 688056 ops)
INFO [FlushWriter:13] 2012-09-19 18:22:46,366 Memtable.java (line 246) Writing
Memtable-sipdb@359811109(0/0 serialized/live bytes, 688056 ops)
INFO [FlushWriter:13] 2012-09-19 18:22:49,719 Memtable.java (line 283)
Completed flushing /var/lib/cassandra/data/test/sipdb-hd-4726-Data.db (37155024
bytes)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
