Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Stefan Seifert wrote: > On Tuesday, 01. April 2008, Florian Haas wrote: > >> The User's Guide, as it stands today, does not include a how-to for using >> the drbd master/slave RA. Until it does, take a look at >> http://wiki.linux-ha.org/DRBD/HowTov2 where what you are attempting to do >> is described nicely. > > Please note, that I have not to date seen a message of someone who is > successfully using the master/slave RA for drbd. I have failed to succeed > myself after months of playing around. Using the way described in the user's > guide gave me a working configuration quickly and really, I don't see any > problem the master/slave RA could solve. > > I have multiple resources having to run on the same node as the drbd master as > well as some that can only run on the slave (like the nfs client) and it does > just what it should do. > > My advice would be: spare yourself the trouble with the master/slave RA. It's > not worth it. I agree this is somewhat true with older versions. Make sure you do this with a recent pacemaker version and don't use heartbeat's crm any more. It's working! Make sure you get the config right. There are examples on the linux-ha mailinglists and on the linux-ha page. The propably most important improvement you get when using the Master/Slave RA is the monitoring of the slave. This way you know when the slave is not okay and can react. With the old drbddisk RA, the slave is not monitored at all and the cluster might try to use a bad slave. You could of course monitor it some other way, but do you want that when using a cluster software already? Regards Dominik