REPOST:Streams locks up. Need help real bad.

I tried a few days ago, got almost no response, and would like to try again with extra data:

I have an SCO 3.2.2 machine running TCP/IP 1.1.3.f, using Cabletron Ethernet card and driver, with Oracle, and 1776 disk array software. Every 4 to 7
days we get terrible response from Oracle and TCP/IP when attempting to perform remote access. From the point that the event begins, we can see that the 64 byte STREAMS blocks starts to ascend. I jacked the limit up to 512 but it doesn't seem to help. Before rebooting the machine I tried to kill off everything that might vaguely be contributory and came up with nothing: With the system reduced to sched, vhand, bdflush, init and some getty processes, the streams blocks do not go down UNTIL I do a kill -1 -1. I used to suspect OracleNet but now I believe it is something in the kernel because of the resources not being released when all of the other processes are killed off. Is there any way of saying what Streams blocks are allocated because of a particular process? Any ideas would be greatly appreciated.

This system is acting as a server for machines running in the US and Japan. It is crucial for us to be up and running with a very high degree of reliability. 4 to 7 days is unacceptable and we badly need to resolve this problem.

Please email to me and I will summarize as needed. Many thanks.

-- 


----------Time flies like the wind. Fruit flies like bananas.------------------


Steven W. Orr      steveo_at_world.std.com     uunet!world!steveo


----------Everybody repeat after me: "We are all individuals."-----------------