I've got a cluster setup with 2 W2K8 R2 SP1 with the latest patches installed on HP DL380 G8 servers using a P2000 iSCSI storage array. Drivers and firmware on the storage array are at there lastest provided by HP.
This an install that we are still integrating to support an home grown application with MSDTC and SQL, also integrated to the cluster.
From a networking perspective, I use 3 available NIC on each server and both servers uses the same configuration. One nic is for the cluster sync, directly connected between servers, one nic to the P2000 storage array, directly connected and the other is connected to a switch facing the corporate network where the customer requests comes in.
There is a scenario that is still an issue, where I loose the connection to the client facing network, some of the disks do not flip over to the other nodes. The services do want to flip from the cluster's perspective, but fails to start because the failed nodes is still connected to the storage and keeps iSCSI reservation to the disks mapped to the service.
Is there something that is configurable, other then making the physical connection to the storage fail ?
The issue seems to happen the most when node 2 flips to node 1 when there was a flip from node 1 to node 2 in a short period of time.... Say within 5 minutes..... (network reboot of switches one after the other as an example).
thanks in advance