When certain data conditions exist, Responder may throw exceptions on call updates when under load.
This can occur when multiple predictions are attempting to process calls that happen to be on the same network. Prediction looks at calls and locks them for tracing when it's determining an outage. A prediction process will find a call and trace to determine the cause. This can cause some calls to be locked that may already be looked at by another prediction.
Part of this behavior is based on how Responder Message Queueing was designed and how it works with multiple processes. But part of the problem can be alleviated: an additional tag can be added to some fields that are commonly updated by Prediction to change the locking mechanism.
- Open the DatabaseSchemaConfig.xml file located in the Responder Server folder.
- Search for the 'Locate the COOKTIME_UTC' field.
- Add the highlighted tag to the end of the field:
- Repeat the steps to add the flag to the ANALYZING and the ANALYZING_EXPIRES_UTC fields. This will prevent some of the locks that are created by Responder and help speed up the system process calls.
- Save the file and restart the Responder Windows Service.