Hi gang,
Last week I installed the v4.0d 'jumbo' patch, duv40das00002-19980717,
to two very stable v4.0d systems, apparently without incident, and
have since then experienced some very odd performance behavior.
First things first: one system is a 250 4/266, 128Mb, AdvFS, heavy
NFS server, modest interactive load, the other is a new 533 clone,
256Mb, AdvFS, very heavy interactive use (MATLAB,) although only
4-5 interactive users, modest NFS load. Both systems use NIS, although
inseperate NIS domains (e.g.. they share no resources and do not interact
at all, except as noted below.) Both systems were new installs of v4.0d,
not upgrades; no errors or warnings were reported during the upgrade.
Both systems are managed exclusively by the author... ahah! you say,
perhaps this is the problem :-)
Both of these systems have experienced debilitating 'slowdowns' on
a couple of occasions since upgrading. Performance of ALL kinds seems
to go to nil; interactive access cannot be had, although they do
respond promptly to pinging; NFS timeouts are rampant; no response
can be had from console keyboard input, whether a session is already in
progress or xdm is running, waiting for a console login (the screen
remains dark, with the server not apparently receiving notification of
a 'wakeup' call.
Since these machine are license servers for a number of products,
these outages are proving nearly catastrophic. A significant number
of my other workstations (~5,0) are suffering timeouts relating to
either the ensuing NFS performance problems, license service requests,
or both.
No errors have appeared in ANY of the system logs, either hardware or
software. I am currently running top, netstat, etc., in a couple of
interactive windows, as well as xload, in an effort to identify the
culprit, with no success to date (since I haven't yet experienced a
'freeze' condition since starting these monitoring tools.
Has anyone seen behavior like this? I peeked at the archives and didn't
come across anything obvious (or even subtle,) that might illuminate
the problem.
As always, thanks in advance,
Seth
Received on Tue Nov 17 1998 - 16:38:40 NZDT