grr@cbmvax.commodore.com (George Robbins) (05/21/91)
Our DS5810/32M/3.1C has recently gone from reasonable stability on
the order of 30 days uptime to hanging two days in a row, of course
at a momemnt when I don't have time for debugging or support center
phone tag...
What happens is that the system seems to be working reasonably well,
then kind of winds down into a state where the load index is very
high > 32, but almost all of the "running" jobs are actually swapped
out! At this point it seems to be arbitrary whether an existing user
doesn't notice anything wrong or all he can get is an character echo.
NFS stops working and users mounting my filesystems send nasty e-mail.
Note that doesn't seem to be one of the Ultrix hanging goodies we've
suffered before, either the /usr/new/csh eating all available virtual
memory or intensive nfs serving activity buzzing all the nfsd's.
The only recent changs have been doubling the amount of swap space
from 2 to 4 64M-byte partitions, and adding systems to the NFS export
list.
I kind of wonder if this is the memory/swapping bug that was such
grief when 3.1 first came out for the DS3100? I was hoping to hold
out for 4.2 later this summer, but maybe I do a temporary install and
generate a 3.1D kernel from some UWS tapes...
In the meantime, I guess I can also drop down to 3 swap parititions
and hope the X workload grows slowly (or chmod -x /usr/bin/dx* 8-).
--
George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr
but no way officially representing: domain: grr@cbmvax.commodore.com
Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)