On 07/25/2016 03:45 PM, Joshua M. Boniface wrote:
My understanding is that you need an odd number of monitors to reach quorum. This seems to match what you're seeing: with 3, there is a definite leader, but with 4, there isn't. Have you tried starting both the 4th and 5th simultaneously and letting them both vote?
Just to make sure this statement does not pollute the list archives and
erroneously mislead someone else in the future: what you said is incorrect.
This subject has been explained several times in the past, but let me
explain it again so it's obvious why your statement is incorrect.
A quorum can be formed by *any* number of monitors (greater that zero).
You can have a quorum of 1 monitor; of 2 monitors; or 3 monitors; of 4
monitors, etc.
The recommendation of an odd number of monitors is solely for fault
tolerance purposes.
You need a majority of monitors to form a quorum. I.e., if you have
- 1 monitor: you need 1 monitor up to form a quorum
- 2 monitors: you need 2 monitors up to form a quorum
- 3 monitors: you need 2 monitors up to form a quorum
- 4 monitors: you need 3 monitors up to form a quorum
- 5 monitors: you need 3 monitors up to form a quorum
In other words,
- 1 monitor: you can tolerate 0 failures
- 2 monitors: you can tolerate 0 failures
- 3 monitors: you can tolerate 1 failure
- 4 monitors: you can tolerate 1 failure
- 5 monitors: you can tolerate 2 failures
As you can see, odd numbers are the ones that give you the highest
tolerance to failures while optimizing the number of nodes - well,
except for 1, which provides absolutely no tolerance to failures.
Sergio's monitors may not be forming quorum, but that has nothing to do
with it being 4 monitors vs 3 monitors or 5 monitors. There's another
cause that needs to be found.
-Joao
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
