On Tue, May 17, 2005 at 12:51:23AM -0300, Marc G. Fournier wrote: > On Mon, 16 May 2005, Kris Kennaway wrote: > > >On Mon, May 16, 2005 at 05:57:33PM -0700, Jorge Aldana wrote: > >>Well now I don't feel alone. > >> > >>5.2.1, 5.3, 5.4 I've gotten to lock up a workstation, with Nvidia-driver > >>as > >>well though, htt-on, htt-off, only SHED_4BSD . Now with a new NetApp > >>server > >>amd locks up as well but not the whole workstation, mounts to other > >>servers > >>ok, nfs issue. Our FreeBSD count is getting low now due to these issues. > >> > >>All I need is a howto on getting the dump for the NFS issue so I can send > >>something but on the lockups are total lockups, nothing in the logs its > >>like somebody shot a freeze gun on the system, a hard reboot then its back > >>(after a fsck on the system disk) but nothing in the log as if time > >>stopped.... > > > >See the developers handbook, which has a chapter on kernel debugging. > > Just curious here ... there used to be a fantastic kernel module that > allowed you to do a 'coredump' across the network, which I used way back > when I didn't have enough swap space to do it ... > > Which brings to mind ... has anyone every tried to do a 'console over > tcp/ip' that would send everything it could to a remote monitoring > machine? There are several cases where, for some reason, you can no > longer do logging to syslog, which can be critical for debugging purposes > if you have no other way of knowing 'cause' of a crash/reboot ... > > I realize that you can do alot of this at the hardware level, but for a > remote server that hasn't been setup for this, suddenly causing problems, > this might be a means of getting some sort of information ... even a 'gdb > over tcp' would be cool, if that was possible ...
Robert Watson did something like that once, I don't know what came of it. Kris
pgp5JhB3s2WHy.pgp
Description: PGP signature
