Working with cross-region failover in AWS Elemental MediaPackage

MediaPackage supports workflows where a CDN transparently fails over between multiple MediaPackage Live v2 origins located in different AWS regions. The mechanism also works in a single region, but such a configuration would not protect against regional failures. The purpose of such failover architectures is to force CDN requests to be forwarded to one or more backup origins if the stream on the primary origin is not considered healthy. In order to be transparent, the failover needs all the MediaPackage origins to produce symmetric streams. This requirement can be satisfied through the combination of multiple requirements at different levels in the workflow:

• Encoding: The live encoders need to send aligned ingest streams to all involved MediaPackage channels. This is achieved through the use of epoch-locked CMAF ingest streamsets, where segmentation is using a regular cadence, and the indexing of the segments is based on the epoch time. This can be configured in MediaLive and AWS Elemental Live using the CMAF Ingest output group. For more details about the configuration and the input timecode requirements, see Creating CMAF Ingest Output Groups in the MediaLive user guide.

  The encoding settings and CMAF Ingest output group settings must be identical across multiple encoders. However, it’s possible to use heterogeneous MediaLive channel types in different regions, for example a standard MediaLive channel in region A and a single pipeline MediaLive channel in region B.

  Important

  Please note that your encoder configuration, or re-configuration, must adhere to some restrictions:

  • All video framerates in the CMAF output group need to be consistent. They should be either all fractional framerates, or all integer framerates, not a mix of the two. The combined use of framerates that are multiples of one another, like 25fps and 50fps or 29.97fps and 59.94fps, is allowed.

  • The transition from fractional framerates to integer framerates (or the other way around) across two encoding sessions is not allowed. The two framerates also need to be multiples of one another: 25fps to 50fps or 50fps to 25fps are allowed transitions, but 25fps to 30fps or 30fps to 25fps are not.

  • The output segments sequence numbers should not go back in the past across two encoding sessions. The segment duration should not increase and the reference Epoch time should not change after the first encoding session.

• Packaging/Origination: All involved MediaPackage Live v2 channels and endpoints need to be configured with strictly identical configuration settings, such as URL path, segmentation parameters, and encryption parameters. The channel input type needs to be CMAF. If the channel input type is HLS, the alignment of the output streams will not be guaranteed. Finally, the MediaPackage channel that will act as the primary origin in the CDN origin group needs to include a Force endpoint error configuration that will have MediaPackage return 404 responses to the CDN in some problematic situations, like missing or incomplete ingest, failed key rotations, or slate input (resulting from the upstream encoder having lost its contribution source). For more information, see Endpoint settings fields. We recommend to set the primary MediaPackage origin settings with aggressive Force endpoint error configuration settings (at least Stale manifests, incomplete manifest and Slate input). These settings should not be activated on the backup MediaPackage origin, in order to maximize chances of a successful origin failover at the CDN level.

• CDN: The CDN distribution should be configured to trigger origin failovers on network connection errors and 404 Not Found response codes returned by the origin. Other 4xx/5xx response codes can also be used as failover triggers, on top of the 404 code, to cover a wider range of problems that are not related to the MediaPackage endpoint error configuration. For more information, see Optimize high availability with CloudFront origin failover.

Provided that your workflow respects these requirements, origin failovers should be seamless for your viewers and not generate live-stream-playback disruptions.