Why would mountd die?

[Chuck Swiger]

Tuc at T-B-O-H wrote:
[ ... ]
>>DJB's daemontools come to mind, also Big Brother (www.bb4.org) can monitor a 
>>system, keeping track of the running processes, and yell (ie, email you, page 
>>you, etc) if something bad happens.
>>
> 
> 	Ok, thanks.
> 
>>The first step is to look for log messages (see mountd's "-d" flag),
>>
> 
> 	:-/ Logs to sysout, so can't be run from the /etc/rc.d/mountd. Was
> hoping to have it log to syslog. I guess I could wrap a shell around it.

Or use -d with daemontools, that's pretty much the intended purpose.

The thing is, watch out for binding with portmap or whatever it's called 
nowadays (rpc.portmapd? rpcbind?), you can feed mountd the -p flag, but you 
might have to adjust the clients then, too.

>> and to unlimit coredumpsize (see /etc/login.conf and the shell
>> environment) and see whether you can get a coredump for gdb.
> 
> 	If its run as root, isn't coredumpsize already unlimited?

Maybe.  Kinda depends on what the default is and whether it's been changed 
somewhere down the line.  It's something to check.  :-)

>>It'll help to rebuild mountd with -g for debugging purposes...
>>
> 
> 	If I can get it to do the core, might be the best idea.
> 
> 	I decided to run it in a screen to see if I can see maybe something
> that way. If it gives a clue, maybe won't have to increase coredumpsize
> and compile with debug for gdb.

OK.

-- 
-Chuck