We run an OSF/1 3.0 system as the main Internet/Email/News server
for our campus. It runs your basic set of Internet tools - INN, HTTPD,
sendmail, pine, Tin, wu-ftpd, Mosaic, X, and the like.
We've recently been plagued with abysmal performance, with load averages
of 40/50/60 not uncommon at peak periods. This increased load average
seems disproportionate to the increased usage.
We've watched the system carefully with Top and see nothing out of the
ordinary. We've pretty much ruled out any particular process or server
as the cause.
Watching it with the Digital Performance Solution software shows something
very odd. Every 5 minutes, like clockwork, we see large spikes in the
Performance Solution strip charts. Almost everything spikes at once:
context switches, interrupts, page-ins, _every_ disk device, IP packets
sent. There's nothing we can think of that we do that could cause
such universal load on the system.
What, within OSF, could be running on a 5 minute timer that could cause
this?
What other tools are available to watch the system? Is there are any
reasonable way to determine what is causing that spike? Do we need
to just keep turning things off until it stops?
Thanks in advance for all your responses.
- Saul
--
Saul Tannenbaum, Manager, Academic Systems | "It's still rocket
stannenb_at_emerald.tufts.edu | science" - Vint Cerf
Tufts University Computing and |
Communications Services |
Received on Thu Feb 23 1995 - 15:25:47 NZDT