We have The HP P4000 VSA 9.5 running on a ESXi 5 Cluster with Fujitsu Primergy RX300 S6 Servers. The Failover Manager is Running on vmware Server 2.02 on a Windows Server 2008R2.
Each ESX Server is hosting one VSA with a mapped RAW LUN.
All Volumes are Network RAID-10 2-Way Mirror Volumes
Now I try to describe the Problem:
If we power off one VSA or the ESX host (where the VSA is running on) it takes 40 Seconds until the Read/Write Request are redirect to the VSA that is still alive.
The network analyzer is showing that no keepalive messages are sent between the hosts.
It take up to 45 sec until the VSA that is still alive is updating the mac address for the virtual interface ipaddress of the cluster.
Because of that the Servers that are connect to the Network RAID-10 2-Way Mirror Volumes cannot perform any I/O Operations.
Anyone who can help ?
I may be wrong, but 45 seconds reminds me of spanning tree? Is there a chance spanning tree is enabled on the switch ports?