[volunteers] SVLUG server - any ideas?

[Daniel Gimpelevich]

On Thu, 06 Dec 2007 16:33:31 -0800, Paul Reiber wrote:

> Hiya, Joe!    The server booted back up just fine - all's well for the
> moment.
> 
> Looking at the messages that were on the console right as/when it hung
> - I noticed nscd
> had crashed - but for some reason we have it limited to core dump size
> of 0, so there's
> no way to tell why it crashed...  The logs are there, though... so
> maybe some looking
> around later will yield new clues.

The daemon is configured to have 16 threads, which may be a major factor.
The debug-level is zero, so there are no logs.

> The reason there _may_ very well be one or two new clue this time is
> that I was able to sync
> the disks with that magic-sysrq voodoo stuff that (for some reason,
> which I'm fixing right now)
> no-body seems to talk about much.
> 
> Evidently the "best" thing to do, failing anything else, with a hung
> linux kernel is to try:
>        "Alt-sysrq-s, then Alt-sysrq-u, then Alt-sysrq-b"

This assumes that magic-sysrq had been left enabled, which is becoming
less and less common.

> I took a look at the registers and locks and stuff too, just to see
> what they did.
> They... dumped out densely packed lines of meaningless hex data, for
> the most part.

It might not be meaningless to someone armed with the nscd source code,
but that may not matter.

> Well... it's up again... could  stay that way for another 10 minutes
> or 10 years, hard to say.

"I can say 'rat turd,' but I would not want to eat one."
	--Joel Grey's character in "Remo Williams: The Adventure Begins"