[olug] sendmail help again (ugh)

[Ryan Stille]

Dave Thacker wrote:
> On Wednesday 12 March 2008 13:08:50 Ryan Stille wrote:
>   
>> Whoo! I figured this out.  Sendmail has options to directly queue
>> messages when the CPU load is over a certain limit.  Evidently this
>> applies not only messages as they come in, but it affects whether
>> sendmail will attempt to process anything int the queue, too.
>>
>> So when I ran sendmail -q and it did NOTHING, it was because the CPU
>> load was over one of its limits and it therefore ignored the messages.
>> I changed those limits to something reasonable, then ran sendmail -q -v,
>> and its processing all the messages.
>>     
>
> Darn, I should have remembered that.  I've been bitten by that one too. 
>   
>> BUT I have no idea why my load average looks like this: "load average:
>> 22.00, 22.00, 22.00".  Seems kind of odd, doesn't it?  And when I run
>> top, I see it pretty much staying at 99% idle.  There's not a lot going
>> on on this server...
>>
>>     
>
> That's strange.  I'd reboot ASAP and watch top with one eye and basketball 
> with the other.  
>
> Dave

I found out why my load numbers were funny.  Actually they weren't - 
there *were* 22 processes 'waiting' (actually hung) for the CPU, so the 
normal load average of 0 + 22 = 22.  I figured it had something to do 
with that when I checked today and load was up to 27.  Thats 5 points 
higher in 5 days, so I figured it was something incrementing each day.  
The /etc/cron.daily/00-logwatch process evidently starting hanging every 
night 27 days ago.  I found something about this here: 
http://lists.linuxcoding.com/rhl/2004/msg08855.html but I don't really 
get it.  Something to do with the symbolic link and an SMP kernel (which 
we have).  I'm just going to remove the symlink, as we don't use logwatch.

-Ryan