[ https://issues.apache.org/jira/browse/CASSANDRA-8549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14264468#comment-14264468 ]
Rahul Bhardwaj commented on CASSANDRA-8549: ------------------------------------------- thanks alot .. It worked..!! applying patch worked for us . But we are still facing memory issue. The processes are killed by kernel, because they are eating all memory (oom-killer). We have set JAVA heap to default (i.e. it is using 8G) because we have 64 GB RAM. > Not able to bring cassandra cluster to stable state > --------------------------------------------------- > > Key: CASSANDRA-8549 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8549 > Project: Cassandra > Issue Type: Bug > Environment: centos 6 > Reporter: Rahul Bhardwaj > Attachments: cassandra_conf.txt, system.log > > > Hi, > We are using cassandra 2.1 on prod environment (cluster with three 64 GB RAM > machines) where a batch of write operations are done per minute (around 10000 > inserts) using PHP PDO. > Couple of days back we found our cluster unstable then we restarted cassandra > service on all machines, but since then cassandra is busy in flushing bulk of > sstable_activity which causes cassandra shutdown or unresponsive due to full > heap space (java heap space 8GB). > After several attempts of restarting cassandra service two nodes started. Out > of these two nodes, one node's cassandra daemon consumed whole RAM (64 GB). > Top command shows 75G of virtual memory consumed and 69 G of Resident memory > consumption. Other node is ok. > But node1 again busy in flushing sstable_activity and not getting statrted. > please help in solving this, these issues are on prod environment. -- This message was sent by Atlassian JIRA (v6.3.4#6332)