Hi, Am 11.12.2014 00:22, schrieb Sanders, Bill: > Thank you for the reply, Florian. > > Yes, MON data is on the RAID drive. Is it recommended to get its own drive? What do MON writes look like in terms of size/frequency? > > So slow IO's and monitor elections are a symptom of not enough disk performance? Even though the disks don't appear to be maxed out? Is there a clear way to tell given some fio or rados bench test? > > We don't have SSD's available at the moment, so is it at all logical to co-locate a few journals on the same spinners? If we keep Ceph on this system, I think we will eventually be putting SSD's in, but as of right now our issues are more stability based rather than performance-based. > IIRC most workload was small writes and updates of leveldb taking many iops. So not enough disk performance in terms of iops the disks can handle. You could check the iops for the disks holding mon data dir with iostat -k 1 on the monitor nodes and see what happens when there is elections. We had this case and observed the problem went away as soon as we put our monitors on ssd. Maybe it helps if you put them on separate/faster disks. Regarding journal on same disk as osd data, all writes go through the journal first and are later written to osd. This is a problem because in this setup you'll easily observe spindle contention. -- Mit freundlichen Grüßen, Florian Wiessner Smart Weblications GmbH Martinsberger Str. 1 D-95119 Naila fon.: +49 9282 9638 200 fax.: +49 9282 9638 205 24/7: +49 900 144 000 00 - 0,99 EUR/Min* http://www.smart-weblications.de -- Sitz der Gesellschaft: Naila Geschäftsführer: Florian Wiessner HRB-Nr.: HRB 3840 Amtsgericht Hof *aus dem dt. Festnetz, ggf. abweichende Preise aus dem Mobilfunknetz _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |