kdump not writing any vmcore

[Neil Horman]

On Wed, Feb 10, 2010 at 05:31:47PM +0100, Gallus wrote:
> Hello
> 
> What can be any reasons why kdump isn't writing vmcore file in
> /var/crash/[date] directory? I have a box on which Kexec is property
> configured: /etc/init.d/kdump status shows: "Kdump is operational".
> There is no message like: "disabling kdump" in dmesg. Instead, there
> are messages: "kdump: kexec: loaded kdump kernel" and "kdump: started
> up".
> 
> After the random system-wide error occurs, kdump reacts correctly. New
> kernel is started, and there are messages like: "Memory for crash
> kernel (0x0 to 0x0) notwithin permissible range" (this is correct,
> secondary kernel starts with different kernel command line, without
> the crashkernel option, because it doesn't load any third kernel, as
> we know) and "disabling kdump". But, there is no message: "kdump:
> saved a vmcore to /var/crash/*" and no vmcore is saved.
> 
> When I tested exactly the same box with manually triggered system
> crash (echo c >/proc/sysrq....), everything was the same except that
> the "kdump: saved a vmcore to /var/crash/*" was emitted and the vmcore
> was saved.
> 
> Regards,
> Gallus
> 
To answer that fully, we'd really need the output of the console during the
kdump kernel boot/capture process.  This will most likely be an initramfs
problem (or possibly a kernel issue).  We'll be better able to tell with the
console log
Neil

> _______________________________________________
> kexec mailing list
> kexec at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/kexec