Hey all,
I'm running RHEL4 and am trying to get a kernel core dump on crash using netdump.
By all accounts netdump looks like it's working, the service is started on both client and server. The ip-timestamp folders appear correctly when the client comes up and the server appears to begin taking a dump when the client crashes. The only problem is the client does not actually dump core, it gets to the Oops and does not continue.
On the client 'service netdump propagate' and 'servce netdump start' have been run without error.
Parameters set:
sysctl -a | grep panic
kernel.panic_on_unrecovered_nmi = 0
kernel.unknown_nmi_panic = 0
kernel.panic_on_oops = 1
kernel.panic = 0
echo "h" > /proc/sysrq-trigger creates a new log on the server (As I think it's supposed to?).
I've tried crashing using a null pointer dereference and simply doing 'echo "c" > /proc/sysrq-trigger', both get the same (lack of) results.
From the console:
mynode # insmod panic.ko
Unable to handle kernel NULL pointer dereference at 0000000000000000 RIP:
<ffffffffa042e014>{
anic:initKernelPanic+20}
PML4 2e373067 PGD 31af9067 PMD 0
Oops: 0000 [1] SMP
From the netdump logfile:
[...network console startup...]
warning: many lost ticks.
Your time source seems to be instable or some driver is hogging interupts
rip __do_softirq+0x54/0xf0
Unable to handle kernel NULL pointer dereference at 0000000000000000 RIP:
<ffffffffa042e014>{
anic:initKernelPanic+20}
PML4 2e373067 PGD 31af9067 PMD 0
Oops: 0000 [1] SMP
Note: This is on a virtual machine through VMWare Server 2. My own initial suspicion was VMWare was disconnecting the NIC when the system crashed, but this is not the case. VMWare still reads the NIC as connected.
Any help here would be greatly appreciated, thanks!
Andrew