On 2010-03-21, at 20:12, Matt Helsley wrote: > These are the same kinds of problems encountered during backup. You > can play fast and loose -- like taking a backup while everything is > running -- or you can play it conservative and freeze things. > > I think btrfs snapshots are just one possible solution and it's not > overkill. > > For some filesystems it might make sense to use the filesystem > freezer to > ensure that no files are deleted while the backup takes place. > Combined > with tools like rsync or rdiff backup these operations could be low > bandwidth > and low latency if well-known live-migration techniques are used. > > Or use dm snapshots. If you are using snapshots, then even an open-unlinked file will not be deleted from the filesystem until it is closed, because the inode will still be available on disk even without the filename. That would be a good reason to also store the file handle (e.g. inode+generation for simple filesystems) in the checkpoint file, so that you can re- open this file by the file handle after the process is restarted. Since Aneesh is starting to add an interface for this to the kernel anyway, I don't think it would be very hard to dump/restore a handful of extra bytes with each file. Conversely, now is the time for getting the open-by-handle APIs correct for this code. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
