Idempotency and Sun Network that Don't

[Benson Margulies]

My 3/60 seems to either drop packets or provoke dropped packets in a
singular and annoying fashion.

We use RCS. When an NFS acknowledgment packet disappears, NFS gets a
spurious error code, and fails an operation. Similiarly, a simple mv
command will fail to receive the response to the rename, even though it
has succeeded, throwing an elaborate make into disarray.

This evening, it is happening quite nearly reliably, even though there is
next to no load on the net, and the server whose responses are
disappearing is a 4/280. The symptoms vary -- sometimes I get Server foo
not responding, and sometimes I get idempotency bugs. The 3/ servers have
the same problem sat other times.

The hotline has been unhelpful. My local FE, doing his level best as
always, gave me a brand-spanking-new CPU board, but to no apparent avail.
Another site responded to an earlier message with a story about bad ether
timing crystals, but their FE left Sun, and now no one admits to knowing
anything about the story.

There's not much left in a 3/60 after one changes the CPU, so that leaves
SunOS 3.5, the tranceiver, or gremlins. Not one other 3/60 in the
facility, some used in very similiar ways, demonstrates this behavior.

Any advice short of sui- or worstation- icide gratefully accepted.

Benson I. Margulies                         Kendall Square Research Corp.
harvard!ksr!benson			    benson at ksr.com