[rdiff-backup-users] Various rdiff-backup problems

[David]

Hi there.

I've been using rdiff-backup for a long time, but now seem to have a
problem with it (shortly after upgrading our backup server from Debian
Etch to Debian Lenny).

When the server that's being backed up has a large number of files (in
the millions), rdiff-backup now takes a very long time (I've left it
for 24 hours, and it was still busy), and uses up a huge amount of ram
and CPU.

I've tried turning rdiff-backup's verbosity to maximum, and there is
no indication of what could be causing the massive resource
usage/delay. The last log entry is something about removing some
metadata file. Nothing about scanning the harddrive, or compressing or
anything like that. Also, fwiw, the server's changed very little
between the first rdiff-backup backup and the second.

(I'm using rdiff-backup 1.2.8 on the server).

This is causing problems, because earlier backups never complete, so
the later backups never get a chance to run. At the moment I'm busy
updating some of the backups so they use a hardlink snapshot-like
method for storing history, instead of rdiff-backup. The disk usage
will be greater, but at least it should work better...

Another problem I've noticed is that rdiff-backup isn't very good with
backwards compatibility, eg:

1) If two different versions of rdiff-backup communicate over the
network, they will often fail. This makes it very hard to use
rdiff-backup for over-the-network backups, unless you can ensure the
same version is running on both sides. Instead, I use rsync for
over-network transfers (it's very good with network reverse
compatibility), and only ever use rdiff-backup locally to store
increments.

2) Newer versions of rdiff-backup have problems reading the
rdiff-backup store of older rdiff-backup versions. After upgrading our
backup server, I had to delete the rdiff-backup data store for many of
the backups, losing months or years worth of history (not so
important, really, but I dlike to keep them around until disk space
runs low on the backup server), because I couldn't get rdiff-backup
working (it kept on getting assertion errors with the number of
"current"-<something> files being incorrect), and Google searches & my
experimentation couldn't solve the problem.

I just thought I'd mention those problems.

David.