A query error leads to an increasing size of the local database cache on the Media Recorders storing VOX/voice activity events which leads to performance degradation and affects media processing which can cause data loss in recorded conversations. The problem was introduced in v9.6 with the new feature which can avoid creating unnecessary CDRs for long open line calls without any media activity.
Am I affected?
All v9.6 trader voice recording deployments (BT ITS, BT IPTrade, IPC Unigy and Speakerbus) are affected where VOX/VAD triggered media recording is configured and Media Recorders and Recording Directors are deployed on separate servers. Note: v9.6 Unified Call Recorder service executable can be potentially deployed on top of previous installations too.
- The issue manifests after many weeks or months of deployment (depending on server load and server performance) and once the size of the cache reaches a critical size the issue becomes more and more evident - Parts of affected media records might become choppy/bad quality/some seconds missing
-Voice quality check reports packet loss and affected media records in UI/search/call details are marked with packet loss media processing error - In the case of BT ITS integration TTP packet loss alerts are triggered
It is recommended to manually delete the local cache database in a maintenance window, every 1-2 weeks depending on server load on Media Recorder servers (Recorder Servers acting as Media Recorder role for Unified Call Recorder service):
- Stop the Verba System Monitor and Unified Call Recorder services
- Delete verba_install_folder\work\unified rec\dbcache\voxhistory.db file
- Start the stopped services