Resolve the error message "Loadtest Controller - 3223: Agent connection lost".
The error message "Loadtest Controller - 3223: Agent connection lost" is reported when the connection between the SilkPerformer Controller and a SilkPerformer Agent has been broken or if the underlying communication unit of the SilkPerformer loadtest controller (perfOrb..) reports any connection issues.
Outlined below in detail are the most common potential causes and solutions for troubleshooting this communication error between the SilkPerformer Controller and Agent software:
The use of a Proxy Server or firewall in the network environment between the SilkPerformer Controller and Agent software is resulting in the connection being aborted.
Temporarily remove the Proxy Server or Firewall from the test environment before running a loadtest once again to verify if the third party software/hardware is responsible for the error message.
A high number of errors occur on the SilkPerformer Agent resulting in a degradation of the available CPU and IO resources (consumed in the processing of the errors) and impacting the Agent's capability to respond to the Controller's request to update the monitor data. This therefore means that if the SilkPerformer Controller tries to request the monitor data from the Agent and no response is forthcoming after a certain number of attempts the Controller will close the connection and report the error message "Loadtest Controller -3223: Agent connection lost".
In this type of scenario the single best approach is to reduce the load on the Agent machine by reducing the number of virtual users which it executes.
The Agent machine stopped responding to the SilkPerformer Controller. Possible reasons for this include, the Agent Software hanging, the Operating System host being rebooted, or the remote host using a "hard close".
The best approach to verify the above behaviour is to first locate the log files named orbClient_.log, orbServer_.log which reside in the SilkPerformer directory (C:\Program Files\Borland\SilkPerformer 2006 R2\Logs); these files report any communication errors between the MMC and Agent such as the common "WinSock: 10054 - Connection reset by peer" and can be opened in a text editor (i.e notepad). It is recommended that you review the time stamp of the most recent error message within this file and check the NT Application and System Event Logs (for both the SilkPerformer Controller and Agent machines) in order to correlate if the error corresponds with any issues being reported by the host machines. For example in this scenario such issues may include an error indicating an application (SilkPerformer) crash or a warning informing that the system was rebooted after a windows update was successfully completed.
The SilkPerformer Controller had not been configured to minimize risk of data loss between the SilkPerformer Controller and Agents.
Implement the settings described in the resolution below which will ensure that the communication with the SP Controller and Agent software is kept to a minimum.
How can I configure SilkPerformer so that I minimize risk of data loss due to connectivity issues with remote (WAN) agents?
Finally if the error still persists then please open a new support incident and provide the following files:
- orbClient_.log (SilkPerformer Agent)
- orbServer_.log (SilkPerformer Controller)
- NT Application and System Event Logs (For both Controller and Agent Operating Systems)