Hello, it's clear now. Thank you guys!
2016-01-27 10:22 GMT+01:00 Joao Eduardo Luis <[email protected]>: > On 01/27/2016 08:58 AM, Sándor Szombat wrote: > > Hello, > > > > I'm testing ceph with minimal config on our servers (I'm using > > ceph-deploy tool). We have 3 monitor nodes with 6 OSD. The worst > > scenario when only 1 monitor and 2 osd up. Unfortunatelly in this case I > > got just error when I run for example ceph -s, I think it couldn't reach > > the cluster (I turned down the leader monitor). How can I config ceph to > > work with only 1 monitor? It is possible? I read this page > > <http://docs.ceph.com/docs/hammer/rados/configuration/mon-config-ref/> > > but this isn't clear exactly. > > The monitors require a quorum to properly function. This means you need > a majority, of all the monitors in the monmap, participating in the > quorum. If you have 3 monitors in the monmap, you will always need at > least 2 of them up, running and participating in the quorum. > > When you do not have a quorum, the monitors will stop handling client > requests. This is meant to guarantee consistency, via consensus, across > all the monitors. > > The monitors elect the leader upon forming a quorum. Even if the > supposed leader is down, as long as you have a quorum you will always > have a leader. Of course, you will need 2 out of 3 monitors able to form > a quorum. If you don't have that, then you won't have a quorum, much > less a leader. > > You can have just one monitor though, but that means you will have to > modify the monmap in order to have just one monitor in it. There are > ways to do this (check the docs, under "add/rm monitors"), but I don't > think that's what you want -- especially because you will lose all your > data if you lose your monitor. Except for test clusters, just one > monitor is definitely not advised. > > Hope this helps. > > -Joao > > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
