On Tue, 9 Dec 2008, Keith Lofstrom wrote:
One of my offsite VPN client machines was moved to a slower internet
link. Meanwhile, in October it went through a major automated upgrade.
As a result, the nightly rsync job took longer than 24 hours. The next
night's backup started before the first completed, which attempted to
move all those files again, and slowed the link down further.
Cascading failures. I was incredibly busy, so I turned off dirvish to
that machine rather than fix the problem.
I use Dirvish for >24h backups, too.
I use Dirvish from the Debian packages, which have a cron job. I'm
attaching the original Debian cron job and my modified one; you'll see my
modifications are centered around creating a PID file in
/var/lock/dirvish-cronjob ; this allows me to be sure that I match not
just any rsync, but the actual dirvish cron job.
A problem with Keith's suggestion is that if any user at all is running
rsync, then the dirvish cron job will fail to start.
I hereby permit distribution for this modification under the same terms as
Dirvish itself.
My way also has the cron job output something indicating why it did not
run overnight, so that you get a nightly email knowing what happened.
-- Asheesh.
--
Truthful, adj.:
Dumb and illiterate.
-- Ambrose Bierce, "The Devil's Dictionary"#! /bin/sh
#
# daily cron job for the dirvish package
#
if [ ! -x /usr/sbin/dirvish-expire ]; then exit 0; fi
if [ ! -s /etc/dirvish/master.conf ]; then exit 0; fi
mount_check() {
mntout=`tempfile -p mount`
mount $1 >$mntout 2>&1
if [ ! -d $1/lost+found ]; then # only works for "real" filesystems :-)
# (Yes, I know about reiserfs.)
echo "'mount $1' failed?! Stopping."
echo "mount output:"
cat $mntout
rm -f $mntout
exit 2
fi
if stat $1 | grep 'Inode: 2[^0-9]' >/dev/null; then # ditto
rm -f $mntout
return 0 # ok
fi
echo "$1 isn't inode 2 ?! Mount must have failed; stopping."
echo ''
stat $1
echo "mount output:"
cat $mntout
rm -f $mntout
umount $1
exit 2
}
## Example of how to mount and umount a backup partition...
# mount_check /backup
/usr/sbin/dirvish-expire --quiet && /usr/sbin/dirvish-runall --quiet
rc=$?
# umount /backup || rc=$?
exit $rc
#! /bin/sh
#
# daily cron job for the dirvish package
#
if [ ! -x /usr/sbin/dirvish-expire ]; then exit 0; fi
if [ ! -s /etc/dirvish/master.conf ]; then exit 0; fi
mount_check() {
mntout=`tempfile -p mount`
mount $1 >$mntout 2>&1
if [ ! -d $1/lost+found ]; then # only works for "real" filesystems :-)
# (Yes, I know about reiserfs.)
echo "'mount $1' failed?! Stopping."
echo "mount output:"
cat $mntout
rm -f $mntout
exit 2
fi
if stat $1 | grep 'Inode: 2[^0-9]' >/dev/null; then # ditto
rm -f $mntout
return 0 # ok
fi
echo "$1 isn't inode 2 ?! Mount must have failed; stopping."
echo ''
stat $1
echo "mount output:"
cat $mntout
rm -f $mntout
umount $1
exit 2
}
## Example of how to mount and umount a backup partition...
# mount_check /backup
## Asheesh's locking addition
fail() {
echo "Cron job currently running; I'm outta here."
exit 1;
}
die_if_dirvish_locked() {
OTHER_PID=$(cat /var/lock/dirvish-cronjob 2>/dev/null)
# if the PID file exists:
[ -f /var/lock/dirvish-cronjob ] &&
ps "$OTHER_PID" 2>&1 >/dev/null && fail
}
lock_dirvish() {
MY_PID=$$
echo "$MY_PID" > /var/lock/dirvish-cronjob
}
unlock_dirvish() {
rm -f /var/lock/dirvish-cronjob
}
die_if_dirvish_locked
lock_dirvish
/usr/sbin/dirvish-expire --quiet && /usr/sbin/dirvish-runall --quiet
rc=$?
# umount /backup || rc=$?
unlock_dirvish
exit $rc
