I have a Ceph node with one pool are configured with size is 3 and 98 OSD. So if 1 OSD is down we do not worry about losing data but how many OSD are down we will care about this issue? Anyway, please help me to understand what is happened when one OSD is down, any copy data task to a new OSD?
1 Answers
0
votes
That depends which OSDs are down. If ceph has enough time and space to recover a failed OSD then your cluster could survive two failed OSDs of an acting set. But then again, it also depends on your actual configuration (ceph osd tree) and rulesets. Also keep in mind that in order to rebalance after an OSD failed your cluster can fill up quicker since it lost a whole OSD. The recovery starts when an OSD has been down for 10 minutes, then it is marked as "out" and the remapping begins.
