[cisco-voip] 7970 phones not failing over in CCM 4.1.2

[Mike Newell]

I have a CCM 4.1.2 system with a publisher on one subnet, two subscribers
on another subnet, and mostly 7970 phones plus one 7936 conference phone
on a third subnet.  All are local through a single router at present.

When I look at the configuration for each phone I see all three CCM
servers' IP addresses listed in the order Subscriber 1 (with a lock next
to it), Subscriber 2, then Publisher.  During system testing I unplugged
Subscriber 1 to test failover.  The 7936 failed over as expected to
Subscriber 2.  None of the other phones failed over.

To see what is going on I did a sniffer trace on a 7970's switch port.
The phone tries valiantly to connect to Subscriber 1 - both via TCP to the
CCM registration port and via TFTP.  Of course it fails.  It then tries an
SRST connection to its default gateway which also fails because we're not
running SRST on the core router.  It then starts the cycle again - always
trying to connect to Subscriber 1.

So I went into the menus and erased the config.  I then power cycled - not
just ##*##'d - the phone.  It did the DHCP dance then again tried to
contact Subscriber 1.  Cisco Option 150 did indeed return the set
{Subscriber 1, Subscriber 2}.  Doesn't seem to phase the phone...

This is the latest 7970 load - 6.0.2SR1.  The system is set for
authentication plus encryption.  I run another similar system with
7940/7960 phones that does not exhibit this problem (it's a production
system so I can't just plug in a 7970 and fail its subscriber to see what
happens).

No failover is a Bad Thing(tm).

I contacted the TAC last week and they are "looking into it".  So far
their only comment has been to check the LMHOSTS/HOSTS files on the
publisher and subscribers; these are correct and I'm using IP addresses
throughout anyway.  Othern' that they've come up with zero.

Anybody out there run into a similar problem or have any ideas?  The
system is scheduled for deployment in two weeks and this is pretty much
the only problem I've seen in testing - and it's a BIG one!!

Thanks!!

Mike