Hi, Le mardi 24 septembre 2013 à 18:37 +0200, Corin Langosch a écrit : > Hi there, > > do snapshots have an impact on write performance? I assume on each write all > snapshots have to get updated (cow) so the more snapshots exist the worse write > performance will get? > Not exactly : the first time a write is done on a snapshot block, yes that block (4MB per default) is duplicated on disk. So if you do 1 snapshot per RBD every day, each modified block will be duplicated once time during the day. So, it's not a big impact. But if you do frequent snapshots, one per hour for example, and your workload is a lot of 8KB random write (MySQL Innodb...), then each of this 8KB will throw a 4MB duplication on disk. Which is a big write amplification here. > Is there any way to see how much disk space a snapshot occupies? I assume > because of cow snapshots start with 0 real disk usage and grow over time as the > underlying object changes? Well, since "rados df" and "ceph df" don't report correctly space used by snapshots, no, you can't. Or I didn't find how ! Small example : you have a 8MB RBD, and make a snapshot on it. Then you still have 8MB of space used. Then you write 8KB on the first block, ceph duplicate that block and now you have 12MB used on disk. But ceph will report 8MB + 8KB used, not 12MB. Olivier _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |