[plug] re: server freezing

[Craig Ringer]

> [root at gfpmsql root]# /sbin/lspci 
> 00:00.0 Host bridge: ServerWorks: Unknown device 0012 (rev 13)

Try updating your pci.ids database from pciids.sourceforge.net . It 
doesn't really change much, but it makes lspci output much easier to 
work with and identify.

>>>[root at gfpmsql root]# cat /proc/interrupts 
> 
>            CPU0       CPU1       CPU2       CPU3       
>   0:   16506895   16507517   16507378   16507241    IO-APIC-edge  timer
>   1:          1          1          1          1    IO-APIC-edge 
> keyboard
>   2:          0          0          0          0          XT-PIC 
> cascade
>   8:          1          0          0          0    IO-APIC-edge  rtc
>  11:          0          0          0          0   IO-APIC-level 
> usb-ohci
>  12:         12          8          8         13    IO-APIC-edge  PS/2
> Mouse
>  14:          1          0          0          1    IO-APIC-edge  ide0
>  18:       8502       8519       8504       8532   IO-APIC-level  eth0
>  20:          4          4          4          4   IO-APIC-level 
> aic7xxx
>  22:       5211       5230       5251       5209   IO-APIC-level  ips
>  29:     332445     332628     332689     332663   IO-APIC-level  eth1
> NMI:          0          0          0          0 
> LOC:   66029161   66029191   66029191   66029191 
> ERR:          0
> MIS:          0

All looks reasonable. I can tell right off that you don't use your 
onboard SCSI or ATA interfaces :-) More importantly, nothing is trying 
to share an interrupt or anything silly like that.

>>>[root at gfpmsql root]# uname -a
> 
> Linux gfpmsql 2.4.18-14smp #1 SMP Wed Sep 4 12:34:47 EDT 2002 i686 i686
> i386 GNU/Linux

Try updating to 2.4.20, especially as you have a fairly recent chipset 
(newer than your kernel and distro at least). Sometimes chipset-specific 
tweaks and fixes get in. It shouldn't be needed, but sometimes it is.

> While gathering this information the server froze, no errors nor
> messages.

That's just damnn strange. As I said before, you should probably look at 
talking to its IPMI interface to ask it if if thinks there's anything 
wrong. Unfortunately, I think this requires kernel patches to 2.4.x .

The only other thing I can suggest now is to try running on a 
uniprocessor kernel - just to see. It's remotely possible that you're 
encountering a problem with a driver that has locking problems or 
something similar. Depends, really, on if the server can take the 
workload on only one processor with HyperThreading disabled.

Craig Ringer