[Sunhelp] Odd possible ARP issue?

[Steven Levin]

I manage a fairly heterogeneous group of Windows NT4, Solaris 2.8/x86, and
Solaris 2.7/SPARC systems.

Our 2.7/SPARC system is actually a pair of E4500's running the Veritas
Cluster Server for Oracle.  Each system has a couple of Fast Ethernet
interfaces plugged into the switch for redundancy.

Suddenly, without any warning, none of the Windows NT servers could talk to
the Veritas cluster.  They could ping/connect to the alternate interface,
but since the primary interface wasn't down (not only did it show good link,
but my x86 boxes, my routers, etc., could all ping the primary interface),
it didn't fail over.

We rebooted the Windows boxen, and also manually deleted ARP values, all to
no avail.  In the heat of the moment, we rebooted the pair of 4500's, which
fixed the problem.

Needless to say, I'm fairly unhappy with reboot as a fix.  And I'm totally
lost as to what the issue might have been.  Any input would be greatly
appreciated.

Thanks!

Steve
--
Steven Levin                               steven.levin at ipix.com
Manager and Tech Lead, Web Operations             (650) 388-0490
Internet Pictures Corporation                http://www.ipix.com
             "Life is to achieve the impossible"