Have you changed the default Memtable settings? Are you running on nodes with a single 1TB drive? Are you monitoring your I/O load on the nodes?
On Thu, Jul 22, 2010 at 6:40 PM, 王一锋 <wangyif...@aspire-tech.com> wrote: > The version we are using is 0.6.1 > > 2010-07-23 > ________________________________ > > ________________________________ > 发件人： 王一锋 > 发送时间： 2010-07-23 09:38:15 > 收件人： user > 抄送： > 主题： Re: Re: Re: What is consuming the heap? > Yes, we are doing a lot of inserts. > > But how can CASSANDRA-1042 cause an OutOfMemory? > And we are using multigetSlice(). We are not doing any get_range_slice() at > all. > > 2010-07-23 > ________________________________ > > ________________________________ > 发件人： Jonathan Ellis > 发送时间： 2010-07-21 21:17:21 > 收件人： user > 抄送： > 主题： Re: Re: What is consuming the heap? > On Tue, Jul 20, 2010 at 11:33 PM, Peter Schuller > <peter.schul...@infidyne.com> wrote: >>> INFO [GC inspection] 2010-07-21 01:01:49,661 GCInspector.java (line 110) >>> GC for ConcurrentMarkSweep: 11748 ms, 413673472 reclaimed leaving >>> 9779542600 used; max is 10873667584 >>> ERROR [Thread-35] 2010-07-21 01:02:10,941 CassandraDaemon.java (line 78) >>> Fatal exception in thread Thread[Thread-35,5,main] >>> java.lang.OutOfMemoryError: Java heap space >>> INFO [GC inspection] 2010-07-21 01:02:10,958 GCInspector.java (line 110) >>> GC for ConcurrentMarkSweep: 10043 ms, 259576 reclaimed leaving 10172790816 >>> used; max is 10873667584 >> >> So that confirms a "legitimate" out-of-memory condition in the sense >> that CMS is reclaiming extremely little and the live set after a >> concurrent mark/sweep is indeed around the 10 gig. > Are you doing a lot of inserts? You might be hitting > https://issues.apache.org/jira/browse/CASSANDRA-1042 > -- > Jonathan Ellis > Project Chair, Apache Cassandra > co-founder of Riptano, the source for professional Cassandra support > http://riptano.com
