Exceed signaling message payload size limit when trickle ICE is disabled.

[yuma-m]

When I choose STUN/TURN and disable Use trickle ICE on the KVS WebRTC Test Page, the message payload size exceeds the service limit (10K) and the signaling message will not reach to the master. Are there any workaround for this?

Environment

• Master (C SDK: kvsWebrtcClientMaster)
• Viewer (Browser: KVS WebRTC Test Page)

The actual message payload size of SDP_OFFER

• 13,014bytes (STUN/TURN is selected and trickle ICE is disabled)
• 8,974bytes (NAT traversal disabled and trickle ICE is disabled)
• 7998bytes (trickle ICE is enabled)

[lherman-cs]

@yuma-m you may want to modify this value to your need, https://github.com/awslabs/amazon-kinesis-video-streams-webrtc-sdk-c/blob/master/src/source/Sdp/Sdp.h#L76

[yuma-m]

Hi @lherman-cs, Thank you for your response.

I choose a bigger MAX_SDP_ATTRIBUTES_COUNT and confirmed the master received a SDP_OFFER bigger than 10k bytes. But after receiving the SDP_OFFER, an error below (lws_ws_handshake_client: client_rx_sm exited) is recorded and the master will not send SDP_ANSWER.

2020-09-14 12:47:13 DEBUG   lwsWssCallbackRoutine(): Client receive {"messagePayload":"eyJ0eXBlIjoib2ZmZXIiLCJzZHAiOiJ2PTBcclxubz0tIDYwNTM0ODQxNTUzNTgxNjEyODAgMiBJTiBJUDQ
... (omitted because the payload is very long)
TE2IHVscGZlYy85MDAwMFxyXG4ifQ==","messageType":"SDP_OFFER","senderClientId":"JOBANIPTZUH"}
2020-09-14 12:47:13 WARN    lwsWssCallbackRoutine(): Failed in LWS handling routine with 0x5d000028
[2020/09/14 21:47:13:4094] N: lws_ws_handshake_client: client_rx_sm exited, DROPPING 1

Can I modify other lines to avoid this error?

[MushMal]

The issue here is that the message limits are being hit. There is no easy workaround for this yet. The only thing that comes to mind is to reduce the size of the message by reducing the SDP size. It can be potentially done by disabling some of the network interfaces that are not going to be used.

[ycyang1229]

I hit this issue with aws WebRTC testpage(disabled trickle ice). I think aws WebRTC testpage should avoid generating the messages bigger than 10k since the limit is already known. Is it possible to fix the issue on aws WebRTC testpage?

[sirknightj]

We don't recommend disabling trickle ice unless absolutely necessary as it will increase connection time. To overcome this issue, you can reduce the number of ICE servers (specifically, KVS TURN servers) you're passing into the RTCPeerConnection configuration. After calling GetIceServerConfig, discard (filter) some of the TURN servers out.