On 12/17/2013 09:54 AM, Karan Singh wrote:
Umar
*Ceph is stable for production* , there are a large number of ceph
clusters deployed and running smoothly in PRODUCTIONS and countless in
testing / pre-production.
Since you are facing problems with your ceph testing , it does not mean
CEPH is unstable.
I would suggest put some time troubleshooting your problem.
What i see from your logs --
1) you have 2 Mons thats a problem ( either have 1 or have 3 to form
quorum ) . Add 1 more monitor node
Just to clarify this point a bit, one doesn't need an odd number of
monitors in a ceph cluster to reach quorum. This is a common misconception.
The requirement to reach quorum is simply to have a majority of monitors
able to talk to each other. If one has 2 monitors and both are able to
talk to each other they'll be able to form a quorum.
Odd-numbers are advised however because one can tolerate as much
failures with less infrastructure. E.g.,
- for n = 1, failure of 1 monitor means loss of quorum
- for n = 2, failure of 1 monitor means loss of quorum
- for n = 3, failure of 1 monitor is okay; failure of 2 monitors means
loss of quorum
- for n = 4, failure of 1 monitor is okay; failure of 2 monitors means
loss of quorum
- for n = 5, failure of 2 monitors is okay; failure of 3 monitors means
loss of quorum
- for n = 6, failure of 2 monitors is okay; failure of 3 monitors means
loss of quorum
etc.
So you can see how you don't get any benefits, from an availability
perspective, by having either 2, 4 or 6 monitors when compared to having
1, 3, 5. If your target however is replication, then 2 is better than 1.
-Joao
--
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
