Subject: | |
From: | |
Reply To: | |
Date: | Thu, 14 Jun 2001 12:46:42 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Re:
> As Pete reports, cases have been see of "diagmond" consuming
> significant CPU resources when processing type 111 - entries with
> MPE XL STATUS: $FFCC0071 (device-offline).
I see the 5 to 6 diagmond processes per minute on
an HP 3000/968 with:
0) MPE/iX 7.0, C.39.06
1) no users
2) log type 111 enabled ... and NO entries
3) no external devices
4) two internal devices: one disk, one DDS
5) proper SCSI termination (inside and outside)
I also see them in separate tests after:
6) netcontrol stop & nscontrol stop
(after the diags are initialized, of course)
(and with network unplugged, in a separate test)
So...I/O errors (network or disk) don't seem to be
the culprit.
BTW, someone should look at diagmond and fix a bug / design
flaw. It opens /var/stm/data/os_decode_xref **21 times**,
and /var/stm/data/monitor_paths 4 times!
It's also inefficent in loading and running because it's
using *13* libraries:
/lib/libc.sl
/lib/libsvipc.sl
/usr/lib/libIO.sl
/usr/lib/libmcats.sl
/usr/lib/libsocket.sl
/usr/sbin/stm/lib/libdiaggs.sl
/usr/sbin/stm/lib/libdiagms.sl
/usr/sbin/stm/lib/libdiagpl.sl
/usr/sbin/stm/lib/libmpestm.sl
/usr/sbin/stm/lib/libstmsec.sl
/usr/sbin/stm/uut/lib/tllibio.sl
/usr/sbin/stm/uut/lib/tlmisc.sl
/usr/sbin/stm/uut/lib/tlpdc.sl
Note that those problems don't affect/cause the 5-per-minute
problem :)
Stan
Stan Sieler [log in to unmask]
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
|
|