My original query was:
We have a Sun 4/280, 128MB, Xylogics 7053 controller running 2 Hitachi disks
Xylogics 472 tape controller running a 9-track tape. We run SunOS 4.1 with
some patches. The machine is heavily loaded around the clock.
The last three nights we have had crashes around midnight during multiuser
incremental file backups. The crash messages from the 3 incidents are:
Other facts: The crashes all happened during dump(8), but not necessarily
first dump of the evening...The system reboots and the operators rerun the
dumps with no further problem that night...dump(8) is running at nice -5...
dumps are the only time the 9-track tape gets used.
I have found no patch that seems to match this problem. The FEs suspect
memory. I wonder about the tape controller. What do you think?
Thanks to all repliers. The suggestions were:
1. Test memory and the tape drive.
2. Upgrade the OS.
3. Apply patches 100330 and 100274.
4. Don't run at nice -5, as per the dump(8) man page.
The 100330 patch doesn't apply to SunOS 4.1, which I run. I had already
applied 100274; it took only a couple of days after installing the xd
controller to find that I needed that patch.
Upgrading is a major hassle on this machine. And I hope to replace it soon.
I had not seen the comment about nice(1) in the dump(8) man page, so I tried
removing the nicing. And for 2 weeks there have been no crashes. So my
conclusion is that the nicing drives some kernel bug which caused my crashes.
I don't know what bug or what patch to apply, and I would believe that a later
release of the OS would solve it, but I am content to leave the nice off.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:50 CDT