Re: [Crash-utility] crashdc : automatic data collection for new vmcores in text format
Louis Bouchard schrieb:
Dave Anderson a écrit :
> Another question -- as I understand it, the RHEL kdump_post mechanism
> only runs from the initramfs -- which doesn't contain the crash utility
> and the vmlinux file (on purpose). How do you have it working?
The initramfs mounts the whole root FS so both crash & vmlinux are
available. This is a main difference with SLES where the root FS is
mounted under /root while we're in the kexec kernel so some path reorg
is needed in order to access the files.
I also need to tell SLES11's kdump to load crash, gdb & a few other
things to make it work. This is my current task.
If you have any questions regarding kdump on SLES11, you can also ask
here since I implemented it and I'm still reading the list (although I
don't maintain it any more since I don't work at SUSE nowadays).
The fact that the root file system is always mounted in SLES11 was not
the ultimate goal. The plan was handle also the case when mounting the
root file system was not successful (mostly for netdump), of course with
less features (without copying kernel and debugging information). But I
doubt that it really works. :-(
In any case: I really appreciate that automatic data collection. In
fact, LKCD had also a simple program that generated a crash report and I
always wanted to implement it also for kdump, but never had time to do it.
Regards,
Bernhard