Update: PowerEdge6400 PCI Interrupt Bug

Charlie Cook cook at ncbi.nlm.nih.gov
Wed Mar 27 19:22:01 CST 2002


At 05:00 PM 3/27/2002 -0500, Curtis Rubel wrote:
>         At any rate, we currently have all 4 cpus reinstalled, running
>the same 2.4.18 kernel we had previously, our PCI cards are
>installed in slots 1,2,5 and the system is running ok.  We will leave
>it cycling overnight to verify things.
>
>         So either we have a bad PCI slot on this motherboard or
>there is some other issue.

         This problem I'm about to describe was on 8450s, but may somewhat 
apply.  A PERC 2/DC was in slot 1 and a network card in slot 2 which put 
both cards on PCI bus 1.  Hitting both the disks and the network hard would 
crash the systems reliably.  Just one or the other usually wouldn't.  After 
going round and round with the problem, we discovered that by simply moving 
the network card to a different PCI bus the problem went away.  The theory 
was that we were overloading that particular PCI bus.  The upshot of the 
incident was that the factory was supposed to stop shipping 8450s with both 
cards on PCI bus 1.  We do have a 6350 that likes locking up during heavy 
network and disk loads, but in the past, we have liked to blame 
Solaris/Intel for that although we have wondered about it since the 8450 
discovery.

Charlie




More information about the Linux-PowerEdge mailing list