info-vax@ucbvax.ARPA (02/01/85)
From: Richard B. Gopstein <GOPSTEIN@RUTGERS.ARPA> We have three 780's running V3.? on a cluster with only clustered disks. What would running V4 on ONE of these machines entail? Is there any reason why this might cause problems on the other two machines? Thanks. Rich -------
info-vax@ucbvax.ARPA (02/06/85)
From: Jerry Leichter <Leichter@YALE.ARPA>
We have three 780's running V3.? on a cluster with only clustered
disks. What would running V4 on ONE of these machines entail?
Is there any reason why this might cause problems on the other two
machines? Thanks.
Rich
-------
All the machines in a cluster must be running the same version of VMS. (In
practice, you will sometimes be able to get away with different "point
releases" - i.e., 4.x and 4.y - but there are no a priori guarantees.)
practice, you will sometimes be able to get away with different "point
releases" - i.e., 4.x and 4.y - but there are no a priori guarantees.)
There is essentially no hope in the case of 3.x and 4.0. The I/O system for
4.0 is entirely different, using the clusterwide lock manager - which isn't
in 3.x - to coordinate access to the file system. Without that coordination,
the different machines will step on each others' toes, and quickly destroy
the file structure on the disk.
You MIGHT be able to get away with a configuration in which EXACTLY ONE system
has WRITE access to the disk, and the others mount it read-only. You should
also disable caching for all accessors to the shared disk or you can get
stale data. This should work, but no guarantees.
-- Jerry
-------