I have setup a production sql server cluster, as ususall I
setup MSCS first and do the failover test.

I use dual fiber channel card for redundancy, but I meet
prob:
Broke one fiber connect. Passed it did not fail over
because it is redundant. Broke second fiber connection.
CLUSTER FAILED!! Had to reboot DB1(node1) and it failed
over

DB2(node2) was hang on there looks like know nothing about
the fail.

Question:

1.Is this a correct behavior that MSCS suppose to do?
2.If not, any ideas how to fix it?

Thanks
David Liu
Sr. DBA
moneris.com
MCDBA, MCSE

The heart beat cable is used to detect which node is live. If a node issues
a select @@servername and gets no response from the other node it assumes
the other node is dead and takes over. If the heart beat cable, or the
private network link goes down, there is no way to know which node is live
and both nodes will attempt to launch.

What you see is expected behavior.

--
Hilary Cotter
Looking for a SQL Server replication book?
http://www.nwsu.com/0974973602.html


"David Liu" <david.z.liu (AT) moneris (DOT) com> wrote

"Hilary Cotter" <hilary.cotter (AT) gmail (DOT) com> wrote

While this is correct, I don't think this is the problem David is
describing.

I don't think he's referring to the network cables, but the fibre cables to
the SAN itself.

In this case, I think you need to look at whatever software is doing the
multipathing to the SAN from the HBA.