On 09/16/2015 10:19 PM, Somnath Roy wrote: > > Sage/Greg, > > Yeah, as we expected, it is not happening probably because of recovery > settings. I reverted it back in my ceph.conf , but, still seeing this problem. > > Some observation : > ---------------------- > > 1. First of all, I don't think it is something related to my environment. I > recreated the cluster with Hammer and this problem is not there. > > 2. I have enabled the messenger/monclient log (Couldn't attach here) in one > of the OSDs and found monitor is taking long time to detect the up OSDs. If > you see the log, I have started OSD at 2015-09-16 16:13:07.042463 , but, > there is no communication (only getting KEEP_ALIVE) till 2015-09-16 > 16:16:07.180482 , so, 3 mins !! > > 3. During this period, I saw monclient trying to communicate with monitor but > not able to probably. It is sending osd_boot at 2015-09-16 16:16:07.180482 > only.. > > 2015-09-16 16:16:07.180450 7f65377fe700 10 monclient: _send_mon_message to > mon.a at 10.60.194.10:6789/0 > 2015-09-16 16:16:07.180482 7f65377fe700 1 -- 10.60.194.10:6820/20102 --> > 10.60.194.10:6789/0 -- osd_boot(osd.10 booted 0 features 72057594037927935 > v45) v6 -- ?+0 0x7f6523c19100 con 0x7f6542045680 > 2015-09-16 16:16:07.180496 7f65377fe700 20 -- 10.60.194.10:6820/20102 > submit_message osd_boot(osd.10 booted 0 features 72057594037927935 v45) v6 > remote, 10.60.194.10:6789/0, have pipe.
Is this the only monitor in your cluster? Are there others? Logs would certainly be helpful. The more the merrier, I'd think. If you can't send them to the list, please find some place where we can reach them, or drop them in cephdrop and point us to them. Thanks! -Joao > 4. BTW, the osd down scenario is detected very quickly (ceph -w output) , > problem is during coming up I guess. > > > So, something related to mon communication getting slower ? > Let me know if more verbose logging is required and how should I share the > log.. > > Thanks & Regards > Somnath > > -----Original Message----- > From: Gregory Farnum [mailto:[email protected]] > Sent: Wednesday, September 16, 2015 11:35 AM > To: Somnath Roy > Cc: ceph-devel > Subject: Re: Very slow recovery/peering with latest master > > On Tue, Sep 15, 2015 at 8:04 PM, Somnath Roy <[email protected]> wrote: >> Hi, >> I am seeing very slow recovery when I am adding OSDs with the latest master. >> Also, If I just restart all the OSDs (no IO is going on in the cluster) , >> cluster is taking a significant amount of time to reach in active+clean >> state (and even detecting all the up OSDs). >> >> I saw the recovery/backfill default parameters are now changed (to lower >> value) , this probably explains the recovery scenario , but, will it affect >> the peering time during OSD startup as well ? > > I don't think these values should impact peering time, but you could > configure them back to the old defaults and see if it changes. > -Greg > > ________________________________ > > PLEASE NOTE: The information contained in this electronic mail message is > intended only for the use of the designated recipient(s) named above. If the > reader of this message is not the intended recipient, you are hereby notified > that you have received this message in error and that any review, > dissemination, distribution, or copying of this message is strictly > prohibited. If you have received this communication in error, please notify > the sender by telephone or e-mail (as shown above) immediately and destroy > any and all copies of this message in your possession (whether hard copies or > electronically stored copies). > > N�����r��y���b�X��ǧv�^�){.n�+���z�]z���{ay� ʇڙ�,j��f���h���z� �w��� > ���j:+v���w�j�m��������zZ+�����ݢj"��!tml= > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
