1. The quickest way to free up connections would be to reboot the Harvester, this will free up all connections and allow you to run reports again. You can also try recycling the "Netqos NQMysql" Service.
2. It is important to see what was causing the connections to fill up by examining the MysqlProcesslist.txt created from the command above in order to prevent this from happening again. When looking at the file if you see connections that have been open for more then a few seconds with a user and db of 'nsas', these connections are likely coming from Anomaly Detector queries.
If you are not actually using Anomaly Detector data we would recommend disabling Anomaly Detector and it should prevent the problem from coming back.
3. If you do need Anomaly Detector running, it may be best to limit the number of active sensors at any given time.
Also make sure that any Anti-virus software excludes both the \CA\NFA directory and the C:\Windows\Temp directory.
You can also reduce the frequency in which Anomaly Detector runs these queries by following the steps in this document TEC1825431.
4. You can also try moving the Netqos NQMysql Temp directory to the same drive as the application so that the temp tables do not fill up the C: drive when large queries are run, and the temp tables won't be interfered with by anti-virus software. See: How do I change the MySql tmp directory in NFA to a different directory?
Moving the Netqos NQMysql temp directory can also have positive benefits when running Flow Forensic reports which can generate very large temp files.
5. If you are running Symantec Endpoint on your Harvesters it could also cause Anomaly Detector queries to be hung up, see: Symantec Endpoint can cause major performance issues with CA Network Flow Analysis.
6. You should also ensure that your Harvesters are not overloaded and meet the current System Recommendations and Requirements. Anomaly Detector is very resource intensive and works best on low volume Harvesters.