drbd-user March 2010 archive
Main Archive Page > Month Archives  > drbd-user archives
drbd-user: Re: [DRBD-user] Init scripts in 8.3.7

Re: [DRBD-user] Init scripts in 8.3.7

From: Gianluca Cecchi <gianluca.cecchi_at_nospam>
Date: Fri Mar 12 2010 - 10:33:20 GMT
To: drbd-user@lists.linbit.com

On Fri, Mar 12, 2010 at 10:06 AM, Lars Ellenberg

> On Thu, Mar 11, 2010 at 08:00:35PM +0000, Joseph L. Casale wrote:
> > When using drbd with cman under Redhat Cluster Suite, I need drbd to
> start
> > before cman so I edit the init scripts as follows:
> >
> > sed -i "s/^# chkconfig: 345 70 08/# chkconfig: 345 20 80/g"
> /etc/rc.d/init.d/drbd
> > sed -i "s/^# Required-Start: \$local_fs \$network \$syslog/#
> Required-Start: \$network \$syslog/g" /etc/rc.d/init.d/drbd
> > sed -i "s/^# Required-Stop: \$local_fs \$network \$syslog/#
> Required-Stop: \$network \$syslog/g" /etc/rc.d/init.d/drbd
> >
> > Are there any caveats to this I overlook? The $local_fs requirement
> causes it
> > to start after cman.
> Why would cman provide "local" fs?
> Anyways, all that matters is that
> both the start order,
> and the stop order,
> are correct.
> So if the result of that modification works for you, go for it.
under which version of OS?
In RH EL 5.4 you have 21 / 79 as start/stop numbers for cman and you don't
have any Provides or Required-* section for it (at least in version

In current F12 (cman-3.0.7-1.fc12) you have again 21 / 79 as start / stop
numbers and:
# Provides: cman
# Required-Start: $network $time
# Required-Stop: $network $time
# Default-Start:
# Default-Stop:
# Short-Description: Starts and stops cman
# Description: Starts and stops the Cluster Manager set of daemons

So if drbd starts in 20 it is ok already ootb.
Probably there is only a problem of latencies... drbd start ends and cman
suddenly starts, but drbd is not yet at Primary UpToDate level....
I have not here the config I used at that time, but I seem to remember in
F11 and F12 I had a similar problem and I solved it changing cman script
adding at the beginning of the start step a sort of drbd status check,
so that if it didn't reach a status acceptable for drbd (st Primary for this
node and ds UpToDate for this node) in N attempts, cman didn't started,
preventing problems with VMs that had their storage on drbd replicated


drbd-user mailing list