Is your feature request related to a problem? Please describe.

When opensips is handling calls using and external RTPEngine media proxy for SRTP/DTLS traffic (in my situation in front of FreeSWITCH media servers) it is important that the rtp proxy selected in the initial offer continues to be used throughout the call. In the event that a different media proxy is selected the DTLS renegotiation is triggered and this can disrupt and crash the back end media server (exact reasons unknown).

Describe the solution you'd like

When we disable an instance in the pool this instance should continue to be used for any existing call sessions and only new sessions should be ignoring disabled hosts. There maybe a reason for multiple disabled modes to allow flexibility for anyone wishing to immediately disable a node from all activity.

Implementation

• Component: rtpengine module
• Type: function select_rtpe_node

Describe alternatives you've considered

Capturing the RTP socket selected after the initial offer, storing to the dialog and forcing all subsequent requests to this node. This is not going to work correctly as the node will be disabled and will likely cause those calls to fail.

Additional context

The current implementation is based on a crude hashing algorithm which assumes no change in the underlying pool, when a change is made all calculations will be affected. Including all the sockets in the pool (enabled and disabled) for every calculation and then checking the disabled flag before final selection (moving on to first enabled socket) will address this issue and impact only those which would have selected the disabled node during initial offer. Adding nodes to the pool will still trigger the issue however.