On Thursday, June 20, 2013, Kurt Bauer wrote: > Hi, > > we run a 3 node cluster, every node runs a mon and 4 osds. 2 defined > pools, one with replication level 2, the second with replication level 3. > We now want to relocate one node from one datacenter to another, which > means a downtime of about 4 hours for that specific node, which > shouldn't hurt the cluster, as 2 nodes are still working. > But is there anything special we should consider or do beforehand, > besides praying that nothing bad will happen with the remaining two nodes? > > Thanks for your help, > best regards, > Kurt >
Well, with four hours out then the moved OSD is going to get marked out, and the other two will re-replicate between themselves. You might want to do something like set the noout flag on the cluster to prevent that. -Greg -- Software Engineer #42 @ http://inktank.com | http://ceph.com
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
