Version

XT 7.0A-0407

Date

3 Nov 2004

Summary

When the Argent services stop and restart, some servers randomly cannot close the TCP port cleanly.

As result, the restarted Argent engine cannot receive any packets from scheduling engine.

Technical Background

In the history screen of the product, many ‘retry’ tasks are visible.

Inspect the service log and note no incoming requests.

The only lines there are ‘… monitoring engine functions normally’.

Inspect the service log of scheduling engine — TCP connection errors are repeated.

Check the registry entry:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\TcpTimedWaitDelay

This value does not exist by default in Windows 2000.

If it is not set, the value is 240 seconds by default.

If this value is much larger than 30 seconds, the binding to port 3190 in the restarted monitoring engine may not succeed, as previous instances have not timed out yet.

Resolution

To address the issue, we need to manually edit the value to be 30 and if the production schedule allows, reboot the machine.