Hi,

I have some problems with the weekly "Content" maintenance of one post
office. (The post office is quite large, but this is a problem that I can't
address immediately...)

GW 14.2.2 -128215

I have moved this post office from a more than ten years old blade hardware
server (Xeon E5420 2.5 GHz, 4 Cores, 8 GB RAM) to a ESXi VM on new hardware
(Xeon E5 2630 2.2 GHz, 10 Cores, 32 GB RAM assigned); the post office runs
with NCS on a NSS SAN volume; the SAN volume is the same, it's RDM mapped to
the VM, the VM was added to the same NCS cluster as the blades.

The post office works very satisfying on the new system, it "feels"
substantially faster than before from (non caching) GroupWise Clients.

But the weekly content maintenance needed about 20 hours on the old
hardware, but needs 35 hours on the new hardware; but the numbers are not
comparable because there is another problem: I had to reduce GWCheck Worker
Threads from 4 to 1. If this is set to 4 I have observed the following: the
maintenance job runs about 20 hours with one thread; then it changes to
using 4 threads - and the server is (almost) not responsive any more. With
Ganglia or top I see high values for "wait IO", between 30 and 60 in
Ganglia, quite often 99 or 100 for a single core in top.

My guess is that there is a problem with disk io, but I have no idea how to
deal with this. Why is the faster machine with the same SAN device slower?
Is the CPU to fast for the SAN? Does the RDP mapping possibly cause such
problems? Any ideas, any practical experience?

Thanks,
Mirko