Has anyone had problems with long server uptimes?

[Gordon A. Lang]

We have had dhcpd in failover pairs on Sun sparc servers (Solaris 10)
running for many years without any problems.  Suddenly we started having
daily problems for over a week, and I have not found a cause.  The
problem was that while most users were getting leases, many were not.
Stopping and restarting dhcpd eliminated the problems for the rest
of the morning rush, and was fine until the next morning when it
would happen again.

"nothing changed in our environment"  (of course)

We are running 3.1.1 with USE_SOCKET defined in a strictly Cisco
"ip helper-address" environment.  During the trouble, I saw some
ordinary messages on the failover peer like "load balance to peer" but
I also saw a lot of "peer holds all free leases" and "lease ... is
duplicate" messages.  One other thing that I noticed is the pool
balancing seemed quite active.

After days of struggling with this, I noticed both servers had uptimes
in excess of 1000 days.  I rebooted both servers, and since then we have
had no more problems.  But I could find no evidence to explain anything.

Has anyone out there seen trouble like this associated with excessive
server uptimes on Sun servers running Solaris 10?

Does anyone have any other thoughts about this experience?

Thanks in advance.

--
Gordon A. Lang