I have been configuring the dual-master HA resource in SLES 11 SP 2 (with current updates for both the OS and HAE). I used the same configuration as per SLES11 SP1 + HAE and SLES 11 etc - all of which worked fine. The settings are easliy found on the web.
Basically I set up a Master resource with a DRBD primitive resource within. The Master resource Meta- Attributes contain the ‘Master-Max=2’ and ‘Notify=True’ settings (I also tried setting Interleave and Globally-Unique. Globally unique seemed to work ok, however only one drbd resource could be active at any one time). The DRBD primitive resource is of Type ‘LINBIT’ ‘DRBD’ and apart from the setting to the specific DRBD resource, there are the 2 ‘Monitor’ entries, one for Master node set with Interval=20 and TimeOut=20, and the other Slave node set with Interval=30 and TimeOut=20.
For some reason, on SLES11 SP2 + HAE, the above configuration occasionally works in bringing up the DRBD resources in Master state on both nodes simultaneously. Mostly, the resources simply fail at a state (as shown in the Pacemaker GUI) (rc=1), meaning a general failure, when the resource tries to Promote the DRBD resources to Primary status.
At one point, having the ‘Notify=True’ and ‘Ordered=True’ seemed to work. But it did not last. All this was a never showed in SLES11 SP1 + HAE. I have tried many things, but still cannot get the DRBD resources starting within the cluster with both nodes comming up with the DRBD resources in Master state.
The DRBD resources work fine if I issue a ‘rcdrbd start’ and using ‘drbdadm primary all’ commands on both nodes.
Has anyone seen this type of behaviour with SLES11 SP2 + HAE and DRBD dual-primary resources?
Any insite is welcome, this is driving me crazy!