Repost/Update: RH9 ran fine, FC1 locks system completely!

Exile In Paradise exile at weylan-yutani.com
Fri Dec 5 03:26:42 UTC 2003


This past weekend I upgraded a smooth-running RH9 to FC1.
All of the hardware remained the same.

After upgrading and applying all of the updates such as glibc,
I have begun experiencing complete system lockups.

Hardware:
Motherboard: Supermicro P6DGU (440GX chipset)
CPUs: Dual Pentium III 500MHz (Katmai)
RAM: 768MB
Drives: 1 x 9GB Seagate and DLT-4000 on motherboard AIC7890
2 x 20GB Western Digital on onboard IDE (requires ide=nodma to boot)
1 x 10x DVD-ROM on IDE primary channel as slave
NVidia GeForce 4 MX440
SB Live! Platimum (emu10k1 primary sound+MIDI)
Ensoniq 5880 AudioPCI (add on MIDI bus)
3Com 3C905B

The symptoms:
Each morning at 5:01am localtime (Central TZ) the system locks requiring
power off, restart, filesystem check, and metadisk resynch.
In addition to the consistent 5am lockups, I have had 3 instances of
lockups at other times. The only noticed symptom then was high load on
the CPU at the time.

Suspicions:
Cron/Anacron kicks off cron.daily at 4am central.
Some process then is causing enough load to lock the system around 5am.
The system is already 100% busy with SETI at home and some extra process
is causing load that locks the system due to a bug.
I did see one note that said the 2129 kernel fixed some 440GX bugs?
This morning I did update to 2919 kernel which I am fervently hoping has
fixed this problem.

Thanks in advance for any help that can be lent to fixing this.

Update Dec 04:
Yesterday I went through all cron steps in cron.daily and ran
each one without a lockup. I also modified /usr/bin/run-parts
to call logger for each script start and stop. Then I kept an Xconsole
window open wide to see what would still be on the screen after it locks
up completely.

I was also able to lock my computer twice by starting Quanta.
Somewhere after the kbuildsycoca running message it would lock.

So, last night, I went through and rpm verified my system starting with
kde first, then realizing I needed to do everything.

Every package that failed to rpm -V cleanly was reinstalled from the FC1
CD's after re-verifying the RPM GPG keys, or re-examined if it was only
config file changes being reported.

Many old stupid sloppy admin things were cleaned up on the box...
basically the cruft of many addon RPMs, and continuous upgrades since
RH6.2 or so. The result is that rpm -Va only shows things I can explain,
not the cruft of 1000 dead packages.

I did have some ximian packages I removed and after all of verifying etc
I can run Quanta without causing a lockup.

I also re-ran prelink on anything that reported problems, since I am not
familar with prelinking and wanted to get rpm -V to report no problems.

I also updated and re-ran chkrootkit just in case.

This morning at 4am sharp, it still locked up.
The cron jobs had not started... but accounting had, just a bit before
the last entry on my xconsole line.

I am able to reboot, fsck, and resynch my mirrors cleanly, at the cost
of 15 minutes reboot time + 75 minutes resynch slowing my computer to a
crawl.

Just about out of ideas on this one.
Going to leave SETI at home off tonight and see what happens, and work on
accounting now...

-- 
Exile In Paradise, Linux User #258896, RHCE #809003961007973
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20031204/a6c5bb7d/attachment-0001.sig>


More information about the fedora-list mailing list