On Jun 24, 2008, at 10:08 AM, Binger David wrote:

> Here is a remote backup script I've installed here as a three cron
> jobs
> running at different frequencies. (rsync >=3 required)
> So far, it seems to work as intended.
> Is there anything we should do to improve this?

Yes there is.

I misunderstood the behavior under the --append-verify flag.
If, as usual when a Durus database that gets packed, the
sending side's file is smaller than the one on the receiving
side, nothing happens.  That's not good for backups.

I think we need to use the '--append' flag between packs,
and to just do a straight rsync, without --append or --append-verify,
when a pack has been detected.

>
>
>
> #!/usr/bin/env python
> """
> Backup remote Durus database.
> This assumes that you have configured your system so that
> this user can ssh to the remote system without entering a
> password.
> """
> from commands import getoutput
> import sys, os
> try:
>    remote_host, remote_path, local_path = sys.argv[1:]
> except:
>    print("%s   " % sys.argv[0])
>    raise SystemExit
>
> local_prepackstat_path = local_path + '.prepackstat'
> local_prepackstat = ''
>
> rsync_flag = '--append-verify'

# The default is a full copy.
rsync_flag = ''


>
>
> if os.path.exists(local_prepackstat_path):
>    local_prepackstat = open(local_prepackstat_path).read()
>
> remote_prepackstat = getoutput("ssh %s stat -t %s.prepack" %
> (remote_host, remote_path))
> if local_prepackstat:
>    if local_prepackstat == remote_prepackstat:

# Take the fast option if we have evidence that a pack has not
happened since the last run.
>
>        rsync_flag = '--append'
> f = open(local_prepackstat_path, 'w')
> f.write(remote_prepackstat)
> f.close()
>
> command = 'rsync %s --rsh=ssh %s:%s %s' % (
>        rsync_flag, remote_host, remote_path, local_path)
> #print(command)
> os.system(command)


Note that rsync version 3 still has an advantage over earlier versions
because it offers the fastest option of --append without verify.
This might be useful if you need very frequent updates and/or if
your database is too large to verify at that frequency.