Let me assume that the graph shows the CPU idle time. How do I know that the spikes are during the replication
It is possible that you observe CPU spikes soon after the replication because that is when you will have very few cache hits . Because searches are done live. Even if the index is very large the download is not CPU intensive. We have done perf studies end which showed low CPU usage with large index downloads ( 6GB) To verify this , just setup a temporary slave which does not serve any requests and do a replication and see the CPU usage On Tue, Apr 7, 2009 at 3:02 PM, sunnyfr <johanna...@gmail.com> wrote: > > Hi Noble > > I turnd off autoWarming to zero. > And yes it's during it replicate, it takes all the data index. > > Because it merges too much, too much update 2000docs every 30mn, it always > merge my index. > So the replication bring back all my data/index. > which use a big part of the cpu like u can see on the graph, the first > part>> > http://www.nabble.com/file/p22925561/cpu_.jpg cpu_.jpg > and on this graph and first part of the graph (blue part) it's just > replication no request at all. > normally i've 20 request per second > what would you reckon ? > > > > Noble Paul നോബിള് नोब्ळ् wrote: >> >> hi sunnyfr, >> >> I wish to clarify something. >> >> you say that the performance is poor "during" the replication. >> >> I suspect that the performance is poor soon after the replication. The >> reason being , replication is a low CPU activity. If you think >> otherwise let me know how you found it out. >> >> If the perf is low soon after the replication is completed. I mean the >> index files are downloaded and the searcher is getting opened, it is >> understandable. That is the time when warming is done. have you setup >> auto warming? >> >> On Mon, Apr 6, 2009 at 11:12 PM, sunnyfr <johanna...@gmail.com> wrote: >>> >>> Hi, >>> >>> Sorry I can't find and issue, during my replication my respond time query >>> goes very slow. >>> I'm using replication handler, is there a way to slow down debit or ??? >>> >>> 11G index size >>> 8G ram >>> 20 requests/sec >>> Java HotSpot(TM) 64-Bit Server VM >>> >>> <lst name="jvm"> >>> <str name="version">10.0-b22</str> >>> <str name="name">Java HotSpot(TM) 64-Bit Server VM</str> >>> <int name="processors">4</int> >>> >>> <str>-Xms4G</str> >>> <str>-Xmx5G</str> >>> <str>-XX:ScavengeBeforeFullGC</str> >>> <str>-XX:+UseConcMarkSweepGC</str> >>> <str>-XX:+HeapDumpOnOutOfMemoryError</str> >>> <str>-Xloggc:/data/solr/logs/gc.log</str> >>> <str>-XX:+PrintGCDetails</str> >>> <str>-XX:+PrintGCTimeStam</str> >>> - >>> <str> >>> >>> Is it a problem ?? >>> <double name="systemLoadAverage">0.21</double> >>> <str name="uname">(error executing: uname -a)</str> >>> <str name="ulimit">(error executing: ulimit -n)</str> >>> <str name="uptime">(error executing: uptime)</str> >>> >>> Thanks >>> >>> -- >>> View this message in context: >>> http://www.nabble.com/solr-1.4-memory-jvm-tp22913742p22913742.html >>> Sent from the Solr - User mailing list archive at Nabble.com. >>> >>> >> >> >> >> -- >> --Noble Paul >> >> > > -- > View this message in context: > http://www.nabble.com/solr-1.4-memory-jvm-tp22913742p22925561.html > Sent from the Solr - User mailing list archive at Nabble.com. > > -- --Noble Paul
