Hi Stephane I followed your advice and I increased kernel.sched_migration_cost by a factor of 10. At beginning the situation improved and the load average of the machine decrease to 5.78, 7.79, 10.71 but now came back to 36.36, 33.80, 29.75. I'm going to try 10000000 over the night and I'll check again tomorrow.
I will try numactl ASAP. Thank you Carmelo On Tue, 2015-04-21 at 14:07 +0200, THIELL Stephane wrote: > Hi Carmelo, > > On 04/21/15 11:40, Carmelo Ponti (CSCS) wrote: > > > > checking the load with top I can see a lot of processes called > > migration/[0-31] which are loading the servers: > > You could try to increase the value of kernel.sched_migration_cost on > your robinhood server (x 10 would be a good start). This tuning seems > recommended on a server with a lot of running threads: > > $ sysctl -w kernel.sched_migration_cost=5000000 > > You could also play with tools like numactl to see if cpu pinning helps... > > Let us know if that helps, > > Stephane > -- ---------------------------------------------------------------------- Carmelo Ponti System Engineer CSCS Swiss Center for Scientific Computing Via Trevano 131 Email: [email protected] CH-6900 Lugano http://www.cscs.ch Phone: +41 91 610 82 15/Fax: +41 91 610 82 82 ---------------------------------------------------------------------- ------------------------------------------------------------------------------ BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT Develop your own process in accordance with the BPMN 2 standard Learn Process modeling best practices with Bonita BPM through live exercises http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_ source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF _______________________________________________ robinhood-support mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/robinhood-support
