Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Andreas Schultz <aschultz at ...> writes:
>
> Hi Luis,
>
> First, please send your mails in plain text. Not all people are willing to
> bother with HTML mails.
>
Sorry...
> On Thursday 16 September 2004 17:17, Luis F. V. Gomes wrote:
> > Hi
> >
> > Our team trying to setup a mailserver cluster with heartbeat and DRBD to
> > mount /var/spool/mail, but we are having a problem.
> > If we choose any of the cluster members and simply pull the power plug out
> > and later turn it on again, everything works fine. But if the primary is
> > rebooted or halted using the normal UNIX commands, the secondary becomes
> > primary but when the original primary server is up again, the current
> > primary turns to cs:StandAlone forever logging the following messages:
> >
> >
> > drbd0: Current Primary shall become sync TARGET! Aborting to prevent data
> > corruption drbd0: error receiving ReportParams, l: 72!
>
> There was a thread about a similar problem some days ago. The suggestion then
> was that the network was shutdown before drbd. This leads to a split brain
> situation where both system have localy consistent data, but are still out of
> sync.
>
> Andreas
Yes. You are right. I searched for this subject in the list archives but I
couldn't find a final solution.
The problem occurs due to the way /etc/rc works (at least in Fedora and Red
Hat): it will shutdown drbd only if there is a file named /var/lock/subsys/drbd
(or drbd.init). So, I had to modify the drbd script in /etc/init.d to include:
1) At the end of 'start)' block in 'case "$1" in'
touch /var/lock/subsys/drbd
2) At the end of 'stop)' block in 'case "$1" in'
rm -f /var/lock/subsys/drbd
Now drbd is shut down gracefully before the network interfaces and then halting
and rebooting the primary is no longer a problem.
Thank you.
Luís
[Rest of message suppressed]