In article 6817@novell.com, dseeman@novell.com (Daniel Seeman) writes:
[some stuff deleted]
>A couple of things come to mind:
>
>1. Be sure to SET MINIMUIM PACKET RECIEVE BUFFERS = 150 in your STARTUP.NCF.
>
>If you are already at 150, try 200 just to be sure the allocation is not the
>problem.
Heck, I had mine all the way up to 2000 at one point, and it used 'em all up!
>2. Make sure the NE3200's driver version is 3.18 (I see you listed the driver's
>date, but I cannot remember what date 3.18 is).
>
>Earlier NE3200 driver versions (previous to 3.18) had ECB "leakage" problems.
>
>Finally if neither of these yields a solution, please verify where the "leak" is
>by using MONITOR.NLM. In MONITOR.NLM, select RESOURCE UTILIZATION/LSL PACKET
>RECIEVE BUFFERS. In this area, you will see ECB usage statistics for a few
>"devices (LAN drivers, protocol stacks, etc.)." You will know where the leak
>is when you look here. These statistics should all be fairly steady. If you
>should find the ECB usage for any one of these devices is continually increasing
>and never goes down, you have found your leak. Let me know what happens...
I should have thought of that. We have three NE3200 (see earlier post), and all three suffer from these problems at varying intervals. Before I gave up, I had one running v3.18, one running 3.15, and the other running 3.11. In all three cases, the server
died after a period of time. I think that it was the appletlk NLM that was "eating" the ecbs, because when I unloaded it I got the famous "Resources not released" message, and the resource was ECBs. Maybe I'll try the NE3200 again for a while.
