Peeps

Sorry for the long post but I'm having a complete nightmare with the latest distros of linux, and wanted to give the full story.

I have an AMD Athlon 1Ghz, 512Gb RAM, ATI 3D Rage II+, D-Link network card, onboard sound.

I have been happily running Mandrake 8 for some time. It has run very well with very few crashes. I mainly use the PC as an Oracle database server so not a great use of KDE etc, but when needed it was fine.

I need use Oracle 9i so tried (and failed) 5-6 times to install on MDK8. Probably library versions etc. ... over my head a bit

Decided to upgrade to MDK9.1 and this was not smooth. Eventually reformatted and reinstalled 9.1 from scratch. Then the window manager just started locking randomly with things like Kolf (golf game) or screensaver setup. NO ctrl-alt-del, NO ctl-alt-bksp, NO virtual screens, NO ssh sessions (locked too). It does not lock immediately just after n seconds/mins of using these.

Of course restart was a major issue with disk check finding lots of problems.

Tried running Oracle 9i install and it crashed PC several times also as before.

Gave up and installed RedHat 9 instead. Gnome seemed better and allowed full Oracle install. Also had some lockups after this with Java apps. Virtually the same thing with KDE.

Tried disabling USB in BIOS (read this somewhere??), onboard sound, floppy drive (occasional errors on boot).

Then tried even TWM and that crashed using Kolf too.

Oh by the way there are no (useful) messages in XFree log or messges files (not sure where else to look).

I'm guessing it may be graphics card but everywhere I've searched says the card is compatible (XFree, KDE, Mandrake etc.).
Also it does not make much sense that it runs Gnome well on RH9 and previously on MDK8.

Help please, oh wise ones

Thanks

Zollies

First thing I'd do check the memory with
MemTest86 utility
see if memory is acting up, if all is good, I move on to the different components, unfortunatelly I don't know any utility to check different hardware components except decifering log files (syslogd allows you to increase verbosity of the messages, read up on it from man pages and documentation available from the net), if you have spare hardware components like a video card, NIC card, try replacing it one by one to narrow down the troblesome component, also you muight want to play with BIOS settings if you know what you are doing.

Slap in a second (?) hard drive and try again. From your description "with disk check finding lots of problems" there seems to lurk the trouble. (I don't know disk check, though).

thanks for the replies

I have now run a test and found that running the box with only text or remote ssh logins does not crash it. The server is running Oracle 9i with a test database so box is occasionally being used, although far from stressed.

Problem definitely seems to be with graphical environment which makes me think video card, but KDE/Gnome worked fine on MDK8 ???

Unfortuately I don't have a suitable spare card and am reluctant to spend money without knowing the problem.

neo77777 - I am currently running memtest86. Seems OK so far. As far as BIOS, I mentioned that I have disabled USB, Floppy and onboard sound but no effect. Any other suggestions ?

JZL240I-U - sorry but I think you misunderstood. The disk errors are a result of the crash and subsequent volume checking routines. They are all fixed but the boot takes approx. 5-10 mins after a crash.

Also the other (final !!) option would be to put MDK8 back on but, even if that proved to work well, I'm still screwed with regard to Oracle 9i ... aaargggh.

I'll post updates here as I go.

Certainly sounds like a vidcard issue to me. I think it would be hard to know for sure without trying a different one. One thing you could try would be to alter the XF86config for a VGA setting (640 x 400 x 8) or somthing similar. If the issue is related to your vid card overheating (which happens quite often when cards start to age) then changing the settings lower should allow the system to run longer because the card is not being as heavily taxed. Just a though, worth a shot.

slight

neo77777 :

memtest86 locks up on test 6 - lots of garbage on screen ??? Bad news I assume !

Does this definitely mean bad RAM module ? bummer if it is cos I've only got 1 x 512Mb chip in the machine

Are there any other tests or can the RAM be fixed (e.g. ignore bad bits - thought I read this somewhere).

Thanks

I see. You should consider using a journaling file system then. I use ReiserFS (which is blindingly fast to repair [<1 second] e.g. after accidental power off), but there are ext3, XFS, ... as well. They should be included in your Mandrake distribution, don't know about RedHat.

[Edit]: You could also try to alter the timing of your RAM. For Windows there are tools to read the SPD (non volatile specification of the RAM-module), which should include the timing parameters. Compare those to your actual settings in your BIOS.

Hey gang ... me again.

Now I've done a load of tests since I last posted:

a) Tried RAM from a different PC in the Athlon box and memtest86 still failed.
b) Tried RAM from Athlon in P3-500 and memtest86 passed fine.

So from the above I thought, aha ... it must be to do with the motherboard.

Now I then decided to try some other tests:

a) Used Micro-scope 8 (micro2000.com) which is a very well known diag tool. The memory tests all went through fine on both PCs and on all memory ??

b) Used DocMemory (simmtester.com) and again the RAM tested fine on the Athlon.

Both the above are launched from a bootable floppy, so no O/S interferance.

I am very confused. How reliable is memtest86 ? I have read of some issues with Athlons but not quite the same. Mine crashed out on test 6 (not 5 or 8 as in docs).

Next step is to install Linux on the P3 with different memory but I am still baffled.

Thanks

Tried everything including re-install of previously working MDK8 with different RAM and NO PCI cards at all (onboard VGA) ...

... and it BOMBED ... aaaaarrrrggggh !

Oh well it MUST be either processor or motherboard then.

So I gave up and bought (well had installed for me ) a new P4 m/b and 2.4Ghz processor.

Istalled MDK and all is rosy

What a pain in the butt !!

Still a bit confused about the RAM thing tho ... only memtest86 gave any error and only on that AMD m/b ... hmmm ?

Thanks anyway for your help.

AMD and AGP conflicts can cause crashes unless you pass mem=nopentium to the kernel, but I would be concerned about the mem test failing. Did you try another memory slot? I have a mobo where one slot gives errors and the other 2 are OK. I would guess that my mobo has a broken/bad solder connection on that slot, but it runs well using the other 2 slots.