Hi all,

I have a couple of BM 5.1 + SP7 + latest TCPIP + PROXY.NLM which exhibit the following symptoms:

Normal server utilization about 15-20% under load. Something happens to PROXY or so it seems, utilization increases to a new baseline of about 30-40%. Then there will be some further disturbance of PROXY and the baseline CPU utilization rises again, say to 50-60%. This process of stepwise increases continues untill the server is at 100% chronically and must be rebooted.

After unloading PROXY, utilization still stays high. Unloading virtually every NLM you can does not seems to lower utilization. The utilisation is in 1 or 2 "SERVER xxx processes" or "SERVER 00:02".

After rebooting, utilization drops back down to 10-15% and will stay there for days, weeks untill one of these short 1-2 minute disturbances happens.

If you look at the PROXY status / activity screen the only visible symptom is that the connection count, which looks like the following normally:

Connections in use: 1205 Connections awaiting tear down: 1194

will show something like 5000 in use and 2000 in teardown. This seems to increase during these step-ups in CPU utilization. When unloading the proxy, while it is clearing open cache files, the in use count might be something like 3000. Normally this is 0 when the NLM is exiting. Otherwise every other resource looks in great shape. It is as if some work-to-do gets stuck in a loop and never finishes and then you end up with death by a thousand cuts.

What I don't understand is why does this happen in steps, and why does unloading PROXY not bring the server back to 0% utilization...

Is there a known mechanism that might cause this? Suggestions?

-- Bob

- - - - - - - - -
Traffic Shaping Engine for NetWare
- - - - - - - - -