Red Hat 7.2 hangs after 3 weeks repeatedly

[wbstrider]

I run a batch file every night. XCOPY overwriting files from a Novell server to the Linux server. Works flawlessly every night.

But, after 3-4 weeks the server mysteriously hangs. Must use reset button to reboot.

My reliability test for Linux is a flop if the system cannot stay up for more than that... I do not use the server for anything else at this time. Just testing it. My Novell server is still working. Up for years without being shut down. But I must make Linux work.

Thanks for any advice... One Code to rule them all, One code to bind them.

 

[fluid11]

Anything interesting in /var/log/messages?

 

[wbstrider]

I didn't know they existed, thanks.

I says
"Unable to handle kernal paging request at virtual address d5630e24"
printing eip

and lots more data follows. Don't know what to look for. Does the above give any indication? One Code to rule them all, One code to bind them.

 

[mrregan]

Hi,
Are you sure the server is hanging or is the console just locked up?

 

[marsd]

Okay,
You are not giving us any details and are being
netwarecentric about the problem.

You are also being very vague about your problem.
You have some mysterious connectivity between the linux
machine and the netware machine..what is it? What is
mounted? Are you using ncpfs, or what?

You do not give us a kernel version or distribution.
You do not give us hwinfo or a description of services
and processes running on the linux server at the time of
the crash.

For a problem like this, where it could be an unreliable kernel, memory leak, bad memory,
flaky hardware, you have given us no tools to work with.

I'd start with memtest86 if I were you.

 

[sleipnir214]

Take it easy, marsd. The guy didn't even know where the logs were kept. ______________________________________________________________________
TANSTAAFL!

 

[marsd]

I was a netware admin for three years and
migrated all my internet connected services to linux because netware tcp/ip,nat and
border manager sucked so bad.

Then you have some newbie come onto site
and start blaming linux for what is probably a misconfiguration or hardware.

I have zero patience with that stuff.

Give details and be objective or don't complain.

 

[Morsing]

It is NOT a hardware problem but a problem with the Linux VMM in connection with the driver.

Upgrade your kernel and the problem goes away.

P.S. As far as I know they are many very happy Netware users out there. Let people work with what they like. Linux isn't the answer to everything. Henrik Morsing
IBM Certified 4.3 Systems Administration

 

[wbstrider]

Ok, more information. Apologies for insulting Linux. I want more than you know to make this work.

Red Hat 7.2 (Enigma) release 2.4.7-10
AMD processor 550mhz 256mb ram. 3com 3c980c-txm NIC
Using Samba to allow windows client to connect to Linux box
I login using mapped drives from win98 client to both the Netware 3.12 machine and the Linux server. When the office is gone the win98 client runs a batch file to Xcopy the files from Novell to Linux.

I run no other programs on the Linux machine. It just sits there 24/7. I have turned off all power saving that I could find. The keyboard becomes locked and the server is inaccessible by the win98 client when it hangs.

I will order the Red Hat 8.0 upgrade today in hopes that the problem is then solved. One Code to rule them all, One code to bind them.

 

[Morsing]

Ok,
do "grep oops /var/log/messages".

Anything?? What happens is that either a driver or something else in the kernel requests a page in memory. These pages comes in different sizes but some sizes are more scarce than others. You can see the pages in a file in /proc but I can't remember which. mem-something??

Several things can have happened: The page requested is at an illegal address, this is a segmentation violation. The page is requested outside the process limits, this is a segmentation fault. If no pages of the rigth size is available, is will also fail.

If there is an Ooops-line in the log, it means that the process that made this illegal action was the kernel itself.
Normally the kernel terminates any process if it can't satisfy it's page request but it can't do that to itself, so it stops.

This is usually caused by a programming error in a driver and can only be solved by opgrading the kernel (or driver). You don't need to upgrade the whole OS.

Hope it helps. Ask if you don't know how to upgrade the kernel.

Cheers Henrik Morsing
IBM Certified AIX 4.3 Systems Administration

 

[wbstrider]

I downloaded a different driver (tar file) for the NIC and posted a request as to howto install the updated driver in this forum.

I gladly will research howto upgrade the kernal. Any Man pages you could suggest? I assume I can get another kernal from Red Hat website.

It's the simple commands that elude me. Howto actually use an RPM. Even though I set up the Linux server and Samba myself (with a little help from the Samba forum )

Thanks again!! One Code to rule them all, One code to bind them.

 

[marsd]

The page problem is not necessarily an issue with
virtual memory,. Is is also possible for a page fault
to be caused by a memory leak, bad memory, or flaky
hardware of several kinds.
I have had page faulting conditions occur, usually with
bad memory.

In any case upgrading the kernel is upgrading the "whole
OS" sans application and utility binaries, though it is basically trivial. If you will post part of your error
dump for us that would be helpful. Also any messages preceding it that look suspicious.

If you will do a google search against your error
message you will find a number of resources that can guide
you towards possible solutions.


Good Luck.