That suggests an idea for the possible cause, to me.

Symptom: inbound ssh connections start hanging, and eventually, after a long time, recover. Each new ssh connection requires some high-quality random bits, unguessable absent knowledge that can't be externally observed. Suppose the ssh server is drawing them directly from an entropy pool that is only slowly, gradually refilled by the OS observing things like the rotational latency of spinning drives as that's slightly disturbed by turbulence?

I don't know the OS that underlies netapps, so I don't know if that pool can be manually fed from commandline, nor if the ssh server can be redirected to a PRNG itself just seeded by the entropy pool. Or if so how to do it.

-Bennett On Jun 29, 2011 1:31 AM, "Stefan Funke" bundy@usage.de wrote: