> -----Original Message----- > From: ceph-users [mailto:ceph-users-bounces at lists.ceph.com] On Behalf Of > Andreas Gerstmayr > Sent: 06 March 2017 12:58 > To: Ceph Users <ceph-users at lists.ceph.com> > Subject: [ceph-users] Current CPU recommendations for storage nodes with > multiple HDDs > > Hi, > > what is the current CPU recommendation for storage nodes with multiple > HDDs attached? In the hardware recommendations [1] it says "Therefore, > OSDs should have a reasonable amount of processing power (e.g., dual core > processors).", but I guess this is for servers with a single OSD. > > How many cores and GHz should I calculate for storage nodes with 24 HDDs > attached (one OSD per HDD)? > And what about hyper threading? > > > In the current setup we have 2x Xeon E5-2630 v3, 2.40GHz (in total 16 cores > per server, 32 with hyper threading), which works under normal workloads. > But when an entire node is down for about 10-20 minutes and then up again, > OSDs of all servers in the cluster start flapping (and all cores are at 100%). > Either the CPUs don't have enough power or we hit a bug in Ceph (11.2.0). > > [1] http://docs.ceph.com/docs/kraken/start/hardware-recommendations/ > > I can happily run 12 disks on a 4 core 3.6Ghz Xeon E3. I've never seen average CPU usage over 15-20%. The only time CPU hits 100% is for the ~10 seconds when the OSD boots up. Running Jewel BTW. So, I would say that during normal usage you should have more than enough CPU power, not sure why you start getting flapping. Do you have the correct amount of PG's for the number of OSD's? > Thanks, > Andreas > _______________________________________________ > ceph-users mailing list > ceph-users at lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |