Hi Andrew, > Is this data all from the CIB process? > What was the load from the other processes like? (Ie. using top) The cib process was approximately 100% CPU rates of use, but most of the other processes did not use a CPU when they confirmed it in top. The number of the cores of the CPU assigns 2.
>> Is there the method to make these processing early? > You mean faster? Uh...For example, I thought in a parameter like "batch-limit".. Because I feel that the handling of XML is not so early... > Could you attach the CIB your cluster was using so I can have a go? I'm sorry.cib.xml was not able to send a e-mail by the limit of the attached file. I compress a file and put it in Bugzilla. - cib.xml :http://developerbugs.linux-foundation.org/attachment.cgi?id=1684 - systemtap:http://developerbugs.linux-foundation.org/attachment.cgi?id=1685 > Ah, I wondered where you go the data from. > Could you share how you used systemtap for this? step 1) I divide a crm file at first to build a cluster. - The attached file is cib_14+2_functest01.crm and cib_14+2_functest02.crm. - Because I become the error by construction with one file. step 2) After construction completion, I let a DC node break down. step 3) I started systemtap from the 1,266th line of the log file in 120 seconds of the 1,267th line and acquired information. - The attached file is ha.log.This file is ha.log of the new DC node. note ) A systemtap use procedure 1)stap -x <PID> functime.stp 2)stop is Ctl+C -- Regards, Tomo 2010/8/5 Andrew Beekhof <and...@beekhof.net> > 2010/8/5 nozawat <noza...@gmail.com>: > > Hi Andrew, > > > > I took oprofile of 120 seconds before Time-out occurs. > > However, sorted_xml function seems not to be called it so much as far as > I > > watch log of Oprofile. > > I suspect you're misinterpreting the results. > The number of calls isn't nearly as important as how much time is > spent in that function. > > Could you attach the CIB your cluster was using so I can have a go? > > > > > I took Oprofile by a sample program. > > But, the contents of Oprofile did not agree with the contents of the > sample > > program. > > Therefore I used systemtap because I was not able to do internal > bottleneck > > analysis from information of this Oprofile. > > Ah, I wondered where you go the data from. > Could you share how you used systemtap for this? > > _______________________________________________ > Pacemaker mailing list: Pacemaker@oss.clusterlabs.org > http://oss.clusterlabs.org/mailman/listinfo/pacemaker > > Project Home: http://www.clusterlabs.org > Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf > Bugs: > http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker > -- Regards, Tomo
_______________________________________________ Pacemaker mailing list: Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
