Failover issue with Link Status on heartbeat connection

Ok, new to the world of clustering, but I have a SQL 2008 cluster up
and running happily with a SAS system for storage, and able to fail
over manually as needed. After setting it all up, we've moved onto
proper testing simulating hardware failovers. We're currently using a
crossover cable for the heartbeat link, but are finding as soon as
that connection loses link, we lose external connectivity to the
cluster!

The failover appears to work correctly, all services come back up on
the 2nd node without issue, and I can connect to services LOCALLY.

Externally from clients however, I can ping the IP, but not connect to
the SQL service!

As soon as link is restored on the 2nd (heartbeat) NIC, I can
instantly connect to the SQL Service again! Its definitely not IP
based to do with the 2nd node, as we simulate a power failure (remove
power to one node) which causes the link to drop on the heartbeat
connection. No ammount of waiting will bring the external connectivity
back up, however as soon as we re-power the failed node, the link
comes on almost right away, during the BIOS stage, etc, and
immediately, everything is normal again.

We've worked around this issue for now by putting a switch on the
heartbeat link, but of course this means a failure of the switch will
also be able to re-create this situation.

Have I miss-configured something somewhere? The heartbeat links are
using TCP/IP only (no netbios, or anything else) separate IP ranges,
no DNS, and i've checked the priority order, all is good there on both
nodes. Note that this affects either node, regardless of which one
fails..

Any thoughts?

I would have to say yes, you've misconfigured something somewhere as this is
not the expected behavior of clustering. You should be able to drop your
heartbeat connection without affecting the public lan link.

Is your heartbeat NIC using a different subnet than the public? If not, it
should be on a different subnet

Does your heartbeat NIC have a default gateway set? It shouldn't.

Try updating NIC drivers?

Regards,
John

Visit my blog: http://msmvps.com/blogs/jtoner

<vidguide@gmail.com> wrote in message
news:33f1a66c-b414-487b-bcf7-d8f783c45377@j18g2000prm.googlegroups.com...
> Ok, new to the world of clustering, but I have a SQL 2008 cluster up
> and running happily with a SAS system for storage, and able to fail
> over manually as needed. After setting it all up, we've moved onto
> proper testing simulating hardware failovers. We're currently using a
> crossover cable for the heartbeat link, but are finding as soon as
> that connection loses link, we lose external connectivity to the
> cluster!
>
> The failover appears to work correctly, all services come back up on
> the 2nd node without issue, and I can connect to services LOCALLY.
>
> Externally from clients however, I can ping the IP, but not connect to
> the SQL service!
>
> As soon as link is restored on the 2nd (heartbeat) NIC, I can
> instantly connect to the SQL Service again! Its definitely not IP
> based to do with the 2nd node, as we simulate a power failure (remove
> power to one node) which causes the link to drop on the heartbeat
> connection. No ammount of waiting will bring the external connectivity
> back up, however as soon as we re-power the failed node, the link
> comes on almost right away, during the BIOS stage, etc, and
> immediately, everything is normal again.
>
> We've worked around this issue for now by putting a switch on the
> heartbeat link, but of course this means a failure of the switch will
> also be able to re-create this situation.
>
> Have I miss-configured something somewhere? The heartbeat links are
> using TCP/IP only (no netbios, or anything else) separate IP ranges,
> no DNS, and i've checked the priority order, all is good there on both
> nodes. Note that this affects either node, regardless of which one
> fails..
>
> Any thoughts?