setup:
Three identical servers (HP BL460c G1), sequential serial numbers, even. Identical CPU, memory, and drivers.
Windows 2008 x64 R2 Enterprise.
All three connected to an EVA4100 SAN.
Hyper-V enabled on all three, all three nodes joined to a failover cluster. No issues with installation or initial configuration. VMs can be created and run on individual servers, no problems with failover.
Cluster shared volumes don't work correctly. The volumes only appear on the node which currently owns the shared disk. Any attempt to access the volume by other by processes on other nodes causes the process to hang indefinitely. Maintenance and redirected mode make no difference.
The event log shows event ID 5120 logged every three seconds for five minutes:
Cluster Shared Volume 'Volume1' ('Cluster Disk 2') is no longer available on this node because of 'STATUS_BAD_NETWORK_PATH(c00000be)'. All I/O will temporarily be queued until a path to the volume is reestablished.
Then one or more instances of event ID 5142:
Cluster Shared Volume 'Volume1' ('Cluster Disk 2') is no longer accessible from this cluster node because of error 'ERROR_TIMEOUT(1460)'. Please troubleshoot this node's connectivity to the storage device and network connectivity.
The nodes can communicate over both their public and private (heartbeat) LAN segments without issue. I can not only ping nodes from one another, but access file shares and remote management.
What am I missing in my configuration that CSV doesn't work?
ANSWER: You must enable the "file and printer sharing" as well as the "client for MS networks" on the heartbeat networks cards on all nodes.
That was it. After years of best practices being everything but TCP/IP disabled on the heartbeat segments, I must've missed where this change was documented. As soon as I re-enabled those clients, all the shared volumes appeared as expected.
However, this makes me wonder about what else could be wrong with my cluster configuration, or my understanding of it. Both my public and cluster private networks are configured to allow cluster communication. My understanding was that this meant both networks could be used for internal intra-cluster communication. If the private LAN failed, shouldn't the cluster have used the public LAN?
Subscribe