I wanted to take a moment to talk about some of the ongoing playback issues that we’re seeing in production and provide some updates on what we’re doing internally to address these issues. We’re actively working on two issues that are causing the majority of failures we’re seeing today.
Audio Streams abruptly stop.
This happens when Alexa aggressively buffers the HLS segments and then goes silent for awhile. A standard Plex client would send heartbeats to PMS to keep the session alive. We’re unable to query the Alexa player to get the required information. This means PMS eventually kills the transcode session because it times out. Internally we’re experimenting with sending adhoc heartbeats to your PMS while media is streaming to keep the transcode session alive. We’re anticipating this will be the first fix and will likely ship soon.
This is a challenging issue because of how Alexa Voice Service interacts with your media server. In short your echo device only ever directly hits your server when it attempts to load a media stream. This is why all of the other features of the skill generally work when media playback fails. We don’t have enough information to know if your PMS is local to your echo device so today we just send the remote access address to your device. Internally we’re investigating ways to infer whether or not your echo device shares a network with your media server so we can intelligently support local and remote access connections. We don’t currently have an ETA on when this might be shipped but we’re seeing promising results internally with the experiments we’ve conducted.
We shipped an update that will now send timelines on a regular schedule to the PMS playing media. This should result in tracks no longer ending prematurely.