[rhelv6-list] Odd load average spikes

[Chris Adams]

Once upon a time, Stephen John Smoogen <smooge at gmail.com> said:
> It could be a number of processes hitting device wait at the same time.
> Every process in device wait seems to increase the load average by 1. I
> have seen servers with load averages of 100's running along fine but 100
> processes stuck in device wait for a network connection or something and
> not lowering the average. I would check to see when this happens what is in
> D state.

Yeah, I've seen that before (logged into a server last month with a load
average of 463!), but it doesn't appear to be the case here.  When this
happens, everything is sleeping except my ps/top/etc.

> The next two areas would be looking at iostat and sar during those times to
> see what might be causing it to pop up.

Neither of them show anything odd either.  Since the only services
running (keepalived, dnsmasq just doing DNS, snmpd) don't hit disk under
normal conditions, it doesn't seem like it could be disk I/O.  It could
be network I/O related, but that's all in-kernel I guess (and I don't
see any unusal network traffic spikes at those times).

-- 
Chris Adams <linux at cmadams.net>