dragonfly system backups

[Dmitri Nikulin]

On 4/9/07, Matthew Dillon <dillon at apollo.backplane.com> wrote:
    Well, I hit a little snafu.  Nothing big.  It just turns out that even
    with the memory efficiency enhancements I made to fsck, it still fails
    (runs out of memory) when you hit 50 million inodes or so.  Not to
    mention the fact that it takes 12 hours to run, but that's a different
    issue :-)
    UFS is definitely getting creaky.  I had to reformat my LAN backup
    machine's /backup partition and start all over again.  I am now doing
    a full level 0 backup of all the machines and I will adjust the backup
    system to keep the number of inodes down to a reasonable fsckable
    level.. that is, not keep as many days worth of backups as it was
    keeping before.
    Talk about information overload!
Is there a really good reason not to tar the backups? It seems
impossible to have 50 million tarballs/days of backups, so I infer
that you're just doing a straight copy. Or maybe now's the time to
homebrew a more intelligent backup system. Or just make the
journalling layer useful for backups.
---
Dmitri Nikulin
Centre for Synchrotron Science
Monash University
Victoria 3800, Australia