On 02/28/2014 03:14 PM, Sidharta Mukerjee wrote:
I'd like to take daily snapshots (rbd snap create) of a large VM (~10 TB; image format 2). The VM grows by about 50 to 100 GB per day. These snapshots would server as my daily backups. So far, the snapshots for small VMs (eg 500 GB) take only a second or two and, from what I understand, don't take any extra space. So it seems to good to be true. Are there any consequences to taking a daily snapshot and purging snapshots that are older than say 30 days? For example, will the original VM get slower if there are 30 snapshots? Would the 30th snapshot start taking up a lot of space? It seems too good to be true that I could take snapshot of super large VMs (~10TB) almost instantaneously, without taking up any more space (or maybe just a few megabtyes extra space), and with no performance consequences to the original VM. (Note: I would only clone the snapshots in an emergency, such as some form of data recovery of an old file that a user accidentally deleted.)
A snapshot itself is cheap, it doesn't cost you any performance since it simply writes to new objects instead of writing to the old ones in RADOS.
However, they DO consume space. Everything that is written to a RBD image will consume extra space since the snapshot has to account for all changes.
Removing a snapshot however is heavier for the cluster since it has to start removing objects which could potentially increase the amount of I/O on the disks.
In my opinion going back 30 days with a snapshot is pretty long, but technically it's doable.
-Sidharta _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
-- Wido den Hollander 42on B.V. Phone: +31 (0)20 700 9902 Skype: contact42on _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
