Thanks a zillion to Derek Fraser (d.fraser@fml.co.uk) and Simon Shickman
(simon@horizon.cc.huji.ac.il) for their help on the problem.
>From Derek:
Three of our SS10 51's have been crashing when they have been incremental
backed up. I sent of the savecore dump for these crashes and the problem has been
traced to the ethernet card "le0". When 4.1.3_U1 was installed the patch
100768 is installed. Below is a description of patch 100768 taken from the README
file.
Patch 100768 for the "le" driver is designed such that "le" interface switches
from "back-to-back packet" mode to "packet-at-a-time" mode at a collision rate
of 10 % by default (lecollhi = 100 Decimal). When the collision rate drops
below 5% by default (lecolllo = 50 Decimal), the "le" interface switches
back to "back-to-back packet" mode.
Network traffic is very high and when the backup is run its even higher.
This causes the le0 driver to change to "packet-at-a-time" mode.
The le0 buffer then gets overrun and freezes causing
the machine to hangup. The workround that I have been given
is a "backout" of the patch 100768 for 4.1.3_U1.
The backout patch is as follows
login to the SS10 51 as root and make a copy of the orginal vmunix then type :-
adb -w /vmunix
lecollenable?W 0
$q
reboot
I have also been given a patch(README below). Both of these solutions work.
I hope this helps.
Simon also stated that Sun suggested upgrading to 4.1.3_U1/B. I had
done this, but updated to 101508-06 (most current). I also had an
4.1.3 box hang, but it had patch 100768.
So ... I am no utilziing 4.1.3_U1/B wiht patch 101757 that backs out
100768.
Frank
-- Frank 'Scruff' Miller | Internet: Frank.Miller@master.cna.tek.com Software Engineer/UNIX Systems | "I went into her cave and returned without Tektronix CNA Division | the ears of a rabbit."
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:05 CDT