On 02/11/2013 02:50 PM, Sage Weil wrote: > There is one more wrinkle that you should test: mounting the read-only > snapshot in some fs's (ext4 at least, iirc) actually writes to the device > to replay the journal. At least this was true at one point in time. Well yes it would like to write to the device, but the device is a read-only block device, so the journal can't be replayed, and it even needs to be mounted with a special option (-o norecovery, at least when I tried that in argonaut). But I also got a nice hint - there's a tool called 'freezefs' that freezes the filesystem for a definable amount of time. Sounds good to me. > If you can use KVM + librbd, that would be the ideal route, as cloning is > fully supported there. I just tried that, and yes, it works, except for the stuck snapshot (see my other mail). > Alternatively, if you can avoid using cloning, at > least temporarily, it will be available for the kernel client in a > (kernel) release or two. hm. sounds good to me. >> Is format=2 in any respect more 'unstable' than format=1? I do have to >> decide these days if all the vm's will be running on format=1 or >> format=2. > > I wouldn't consider it any less stable than format=1. In fact, a few > things that behave strangely with format=1 (like renaming an image while > it is in use) work better with format=2. thanks for your fast reply. > sage Wolfgang _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |