Logging panic messages

[Tony Arcieri]

I'm running a 5.3 system with a kernel built from the latest RELENG_5 sources
as of 15 days ago which has just experienced another sporadic reboot in a
long sequence of sporadic reboots which have occured since the system was
originally installed.  It was up some 30 days without trouble before I
colocated that, and after that it was up approximately another 30 days
before it began exhibiting problems.  Since then it has crashed approximately
every 15-25 days or so.

I originally configured a dumpdev only to discover that the size of swap must
exceed physical memory by 1MB in order for it to work, which it does not,
unfortunately.  So dumpdev won't work to collect crash data.

What I'm really in need of is some way to determine the cause of reboots,
and panic messages aren't being logged.  I've heard of patches to the
kernel which would provide ways to do this such as network console support
(my colo provider doesn't provide serial consoles).

Does anyone have any suggestions as to how I could obtain a log of the
kernel panics and any other pertainent information I would need to
debug the problem?  Are there any patches to write debugging data
to the swap partition when the system panics in lieu of a complete kernel 
core, such as the panic message or even better, a KTR dump?  As is I have 
absolutely no information to go on in debugging this problem.

Tony Arcieri