The server maintains the locks. A cache manager obtains a lock, but
it is only good for a few minutes (~5, I think). The cache manager
must renew the lock periodically (and does so every minute or two).
This is actually an interesting topic. Very few applications use
locks, and that's a good thing. On our older RT's we stuck /usr/spool/lpd
into AFS, requiring some jiggling. We were young then...
If you watch the network for AFS protocol, the AFS FetchStatus is usually
the operation which happens the most. Execpt when some application,
like the spooler, keeps a lock on a file. The CM used to (I'd have to
check to see if its still true) renew the locks every minute... making
the lock-renew op the most often used AFS protocol operation during the time
the lock is in effect.
If lots of clients were locking files, then the lock-renew
would become the most-used AFS operations. That hasn't happened on our
net, but mileage varies from cell to cell.
mts.
