Subject: | |
From: | |
Reply To: | |
Date: | Fri, 13 Jun 1997 17:39:27 +0200 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
ed, culp wrote:
>
> I have recently had a problem with our 987.
> Several times a day it would become very slow for 30 to 60
> seconds with much disc i/o. Glance showed SNMP was using cpu time.
> I fixed the problem by adding the following line to the system
> startup procedure:
>
> RUN SNMPUTIL.NET.SYS;INFO='SNMPCONTROL STOP'
If no other node on your network expects your 987 to answer
SNMP requests, then this should be ok for you.
>
> Is this normal or is there something misconfigured?
I am anything but an SNMP expert (maybe James Hofmeister can shed
a little light here), but there is 2 things with this process
SNMP.NET.SYS that I am aware of:
- SNMP contains part of a resource monitoring functionality, it can
tell basic data about CPU & disc space consumption to requesting
systems. (afaik some of our OpenView products make use of this).
Part of this is watchdogging available disc space, and other data
about your volumes per calls to the HPVOLINFO intrinsic (you could
trace that with Glance).
This can cause a higher CPU consumption and a bunch of disc I/Os
for a limited time frame, especially if you have a lot of disc
drives on your system.
- What might make this a "pain" for your users is the fact that
SNMP on 5.0 and 5.5 is running at a too high process priority,
so it preempts your normal CS-queue users, while it is busy doing
this monitoring.
For the latter problem you can get a patch (MPEJXE0) for both
5.0 and 5.5 from your next RC.
Hope this helps,
Goetz
|
|
|