changing crush map on the fly?
Hey, Just to confirm my understanding: If I set up a 3-osd cluster really fast with an EC42 pool, and I set the crush map to osd failover domain, the data will be distributed among the osd's, and of course there won't be protection against host failure. And yes, I know that's a bad idea, but I need the extra storage really fast, and it's a backup of other data. So availability is important, but now critical. If I then add 5 more hosts a week later, I can just edit the crush map and change the failover domain from osd to host, put the crush map back in, and ceph should automatically distribute all the pg's over the osd's again to be fully host-fault tolerant, right? Am I understanding this correctly? Angelo.
You are correct, but that will involve massive data movement. You can change the failure domain osd/host/rack/datacenter/etc... You can change the replica_count=2,3,4,5,6 You *CAN'T* change the EC value eg. 4+2 to something else Kind regards, Nino On Fri, Jun 23, 2023 at 12:40 AM Angelo Höngens <angelo@hongens.nl> wrote:
Hey,
Just to confirm my understanding: If I set up a 3-osd cluster really fast with an EC42 pool, and I set the crush map to osd failover domain, the data will be distributed among the osd's, and of course there won't be protection against host failure. And yes, I know that's a bad idea, but I need the extra storage really fast, and it's a backup of other data. So availability is important, but now critical.
If I then add 5 more hosts a week later, I can just edit the crush map and change the failover domain from osd to host, put the crush map back in, and ceph should automatically distribute all the pg's over the osd's again to be fully host-fault tolerant, right?
Am I understanding this correctly?
Angelo. _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
participants (2)
-
Angelo Höngens
-
Nino Kotur