Socket error 10054: Connection reset by peer
When performing a load test against the following system:
WAPT <-> IBM HTTP SERVER (APACHE) <-> IBM WEBSPHERE (2 NODES) <-> DB2
We got Socket error 10054: Connection reset by peer, logged by WAPT, when the nominal load reached about 350 virtual users.
What we observed was that both IBM WEBSPHERE nodes where running hot. Much, much more CPU utilization than any other node. Much time seamed to be spent in garbage collection. (We enabled the verbosegc flag, and noticed a lot of GC one every ~5 second, and each GC took about 0.3 -> 3 seconds.)
Is it possible that the Socket error 10054 we saw in the WAPT node, is a result of 'bad behavior' from the WebSphere nodes? Or, is it only a matter between the WAPT and the APACHE?
All nodes are running on separate multi-core windows machines, with a lot of RAM, 20 GBytes.
Between the WAPT - APACHE - WEBSPHERE, the HTTPS protocol is used.
1) WAPT is WEB Application Test tool from http://www.loadtestingtool.com/ running on 8-core Windows 2003 machine
2) IBM HTTP SERVER (APACHE) is running on 16-core Windows 2003 machine
3) The WebSphere is a 2-node config, running each on a 8-core Windows 2003 machine.
4) DB2 is running on 16-core Windows 2003 machine
Re: Socket error 10054: Connection reset by peer