Since Tuesday we've been seeing our clustered NetWare POA (v8 HP1 - i.e. before SP1) being slow to respond to user requests. The server shows high CPU utilization and that's it for very obvious symptoms. No pending requests and no problem messages. The POA looks okay but runs slow.

The servers go from 4 to 20% util to maxed out. One cluster node is still on SP7 and the POA abended it (froze) so no abend log. Now the POA is on a SP8 node and its util has gone from peaks of 40% to peaks of 80% inside of a day where the only change was to put the POA on it. So SP8 seems to be better in that it's more resistant but it is still trending bad.

I've seen other threads where the solution was to go to SLES but we can't do that right now so is there any more news or advice?