[DRBD-user] Fencing in primary/primary setup - how to do it?

adrian at ima.pl adrian at ima.pl
Tue Mar 13 20:30:51 CET 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hello everybody,

I'm in process of setting up a two-node active/active cluster based on
Redhat Cluster and GFS filesystem. So far everything is going ok (with
small problems) but there's one thing that bugs me...
The split-brain issue. I know there was much talk about it recently,
but I don't know how to deal with it exactly. To be clear - how
to prevent it in case of Primary/Primary setup.

>From what I have seen in the configuration file, there are two main
options for that: fencing and outdate-peer.

I already have a fencing device which works with redhat cluster software
(apc masterswitch-like one). How to use it here? Can I set a script
that calls it's agent with proper arguments as an outdate-peer handler?
Or should I somehow incorporate it into delivered example (outdate-peer.sh)?
What return code should it give in case of fencing the peer?
Will it minimize the risk of split-brain situation in case of one or
another node's hangup?

Besides the configuration file there is hardly any documentation available.
Could someone calrify this?

Thanks in advance for help.

Greetings,
adrian at ima.pl
Adrian Gruntkowski




More information about the drbd-user mailing list