From: John Byrne <john.byrne@hp...> - 2003-10-28 21:14:10
-------- Original Message --------
Subject: Re: [SSI-devel] Kernel panic during loadleveling on highmem 4GB
Date: Tue, 28 Oct 2003 13:12:01 -0800
From: John Byrne <john.l.byrne@...>
To: Bruce Hunsaker <hunsakerbn@...>
Bruce Hunsaker wrote:
> Openssi 9.95 under redhat 9 -
> I am running a 2 node cluster. on 2 identical IBM Eserver machines each with 4GB ram and 2 pentium xeon processors.
> When I attempt to run the demo-proclb loadbalancing demo from node 1, I get a kernel panic on node 2.
> The demo works fine on a kernel without highmem
> configured, but then the nodes only have 1GB of ram each available instead of 4GB.
> Here is a stack trace from the kernel debugger:
> __ics__dgram_parse + 0x8d
> ics_dgram +0xa4
> tcp_rcv_established +0x6b6
> tcp_v4_do_rcv + 0x11f
> tcp_v4_rcv + 0x4b5
> The only kernel config parameter I changed was
> to set CONFIG_HIGHMEM4G=y (via make menuconfig)
> The demo starts , and does all of the forks. The 2nd node always panics after a few (2-4) processes have been migrated to it.
This should be fixed in the CVS repository.