Running heartbeat 2.0.7 as provided by Debian 4.0 official packages.

Two nodes (cmphoto2 and cmphoto3), running heartbeat in the 'old
haresource' mode rather than with crm, as I couldn't get my head
wrapped around the whole XML configuration.

My haresource file only has one uncommented line:

cmphoto2 drbddisk::photo_archive 192.168.4.127/24/vnews
Filesystem::/dev/drbd0::/mnt/shared_archive::jfs winbind samba

Under an orderly shutdown (i.e. /etc/init.d/heartbeat stop, or
shutdown -h now, or hitting the power button [which goes via ACPI to
initiate a controlled shutdown]), cmphoto3 properly takes over all
resources from cmphoto2.

When pulling power from cmphoto2, cmphoto3 dies when starting
Filesystem, as the filesystem on /dev/drbd0 is dirty and requires a
replay of the journal. Since Filesystem didn't start, it bails out of
all resources. Repowering cmphoto2 also leaves everything in shambles,
as it will properly think it needs to take over, but since the
Filesystem is dirty and won't mount, it also bails out of all
resources.

I found that if I edit the file
/usr/lib/ocf/resource.d/heartbeat/Filesystem line 354 (where it
decides which filesystem types don't need to be fscked), and REMOVE
the entry for jfs then the cluster responds as I wish to 'pull the
power from cmphoto2' events. If I understand the coding correctly,
this means that it is running a fsck on this filesystem every time it
is started, however my understanding is that with jfs this shouldn't
take an inordinate amount of time.

My questions are:
1.) Is forcing a fsck at every start of the Filesystem resource the
best way to solve this problem?
2.) Is there a way to do this that doesn't involve editing the
Filesystem script, but rather some setting in one of the configuration
files (so that when Debian gets around to upgrading to 2.0.8 or
whatever, I don't have to reapply my fix by hand [or worse yet, forget
to do so and leave my cluster vulnerable to 'cmphoto2 goes away
unexpectedly'])?

Thanks in advance to any advice!
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to