On 07/08/13 15:14, Jeppesen, Nelson wrote:
Joao,
Have you had a chance to look at my monitor issues? I Ran ''ceph-mon -i FOO -compact' last week but it did not improve disk usage.
Let me know if there's anything else I dig up. The monitor still at 0.67-rc2 with the OSDs at .0.61.7.
Hi Nelson,
It's been a crazy week, and haven't had the opportunity to dive into the
compaction issues -- and we've been tying the last loose ends for the
dumpling release.
Btw, just noticed that you mentioned on your previous email that the
'mon compact on start = true' flag made your monitor hang. Well, that
was not a hang per se. If you try that again and take a look at IO on
the mon store, you should see the monitor doing loads of it. That's
leveldb compacting. It should take a while. A considerable while. As
I previously mentioned, 10G stores can take a while to compact -- a
220GB store will take even longer.
However, regardless of how we eventually fix this whole thing, you'll
need to compact your store. I seriously doubt there's a way out of it.
Well, there may be another way out of it, but that would involve a bit
of trickery to get the leveldb contents out of the store and into a new,
fresh store, which would seem a lot like a last resort.
But feel free to ping me on IRC and we'll try to figure something out.
-Joao
On 08/02/2013 12:15 AM, Jeppesen, Nelson wrote:
Thanks for the reply, but how can I fix this without an outage?
I tired adding 'mon compact on start = true' but the monitor just hung. Unfortunately this is a production cluster and can't take the outages (I'm assuming the cluster will fail without a monitor). I had three monitors I was hit with the store.db bug and lost two of the three.
I have tried running with 0.61.5, .0.61.7 and 0.67-rc2. None of them seem to shrink the DB.
My guess is that the compaction policies we are enforcing won't cover
the portions of the store that haven't been compacted *prior* to the
upgrade.
Even today we still know of users with stores growing over dozens of
GBs, requiring occasional restarts to compact (which is far from an
acceptable fix). Some of these stores can take several minutes to
compact when the monitors are restarted, although these guys can often
mitigate any down time by restarting monitors one at a time while
maintaining quorum. Unfortunately you don't have that luxury. :-\
If however you are willing to manually force a compaction, you should be
able to do so with 'ceph-mon -i FOO --compact'.
Now, there is a possibility this is why you've been unable to add other
monitors to the cluster. Chances are that the iterators used to
synchronize the store get stuck, or move slowly enough to make all sorts
of funny timeouts to be triggered.
I intend to look into your issue (especially the problems with adding
new monitors) in the morning to better assess what's happening.
-Joao
-----Original Message-----
From: Mike Dawson [mailto:mike.dawson at cloudapt.com]
Sent: Thursday, August 01, 2013 4:10 PM
To: Jeppesen, Nelson
Cc: ceph-users at lists.ceph.com
Subject: Re: Why is my mon store.db is 220GB?
220GB is way, way too big. I suspect your monitors need to go through a successful leveldb compaction. The early releases of Cuttlefish suffered several issues with store.db growing unbounded. Most were fixed by 0.61.5, I believe.
You may have luck stoping all Ceph daemons, then starting the monitor by itself. When there were bugs, leveldb compaction tended work better without OSD traffic hitting the monitors. Also, there are some settings to force a compact on startup like 'mon compact on start = true' and mon compact on trim = true". I don't think either are required anymore though. See some history here:
http://tracker.ceph.com/issues/4895
Thanks,
Mike Dawson
Co-Founder & Director of Cloud Architecture Cloudapt LLC
6330 East 75th Street, Suite 170
Indianapolis, IN 46250
On 8/1/2013 6:52 PM, Jeppesen, Nelson wrote:
My Mon store.db has been at 220GB for a few months now. Why is this
and how can I fix it? I have one monitor in this cluster and I suspect
that I can't add monitors to the cluster because it is too big. Thank you.
_______________________________________________
ceph-users mailing list
ceph-users at lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
_______________________________________________
ceph-users mailing list
ceph-users at lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
--
Joao Eduardo Luis
Software Engineer | http://inktank.com | http://ceph.com
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
