On Tue, Mar 11, 2003 at 10:28:38PM +0700, Jason Smith wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hi. Like many others, I'm interested in using LVM2 as a componet for a > Linux-based file server. i only have experience with the first lvm implementation, not with lvm2. > When needed, I had hoped to sacrifice disk space for long-term filesystem > snapshots, by creating multiple snapshots and rotating them out on > perhaps a weekly basis. The intent is to emulate the convenience and > power of similar snapshot features in commercial file servers. i used to do 4. Every hour, every 8 hour, every day, and every week. I gave up because it was too unstable. But since that there has been found a few bugs in lvm1 snapshots. > My problem is, this does not currently look possible without a severe > performance hit. If a volume has, for example, five active snapshots, > then the deltas seem to get written five times, once per snapshot volume. How would you not do that ? You have to store the snapshots some how. > Alternatively, I could use meta-snapshots, but they are not currently > implemented, and that idea sounds klugy. meta snapshots ? > So then do the LVM developers and users recommend against multiple > concurrent snapshots of a volume? Should I keep hope for this great > feature, only found in commercial file servers? well, if you need it you can either implement it yourself, or possibly find someone who can do it for you. > Thanks for any input. Until this get made you can simulate this, though with the exception that it will take a little time, and use linear to the number of snapshots extra space. You take a snapshot. Then you create another LV, mkfs it, and rsync the data from the snapshot to the new LV. When that is done you have a "snapshot" that doesnt take cpu time and performance, just space. JonB _______________________________________________ linux-lvm mailing list linux-lvm@sistina.com http://lists.sistina.com/mailman/listinfo/linux-lvm read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
