yesterday i tried to setup an mp4 movie as a live-stream following these instructions: https://aws.amazon.com/blogs/media/metfc-live-stream-pre-recorded-video-using-aws-elemental-medialive/ The input was a 4.4 GB mp4 file and when testing, everything worked fine and i didn't find any issues... When it came to the stream really starting with about ~400 viewers, nearly all users had problems with the stream not running constantly... always having to reload the page where the video-player was embedded (as explained in the link)... so the medialive stream input/output was constantly running and there were no errors/alerts on aws side...
i really don't have any idea what could have been wrong here and want to ask what i could have missed? any special settings other than proposed in the link?
Would be very thankful for any hint :)
Kind regards Simon
Update:
I tested a bit around, and what i found out is, when i use the "Workflow Wizard" which is available in the AWS Elemental MediaLive Service, then the HLS Settings are different in the output group. CDN Settings are set to "HLS webdav". In the instructions i posted above there is mentioned setting this to "HLS media store". If i compare the files available in the mediastore, then with the instructions i only have some ".m3u8" files available there. When using the wizard, then i have these .m3u8 files (one for every resolution) and additionally tons of ".ts" files. Does this maybe have any impact related to my issue?
Thanks for the note. This is a good question! When we get to multi-stage workflows like the one you referenced, there are several places where issues or errors might occur which could result in playback stalls. To determine where the issue(s) occurred, we need to visit each stage of the workflow and check for errors or slowdowns.
Each step of the workflow within AWS will have both metrics (rate per second charts) and error logs captured in AWS CloudWatch. CloudWatch logs can answer many questions, such as: How many customers got 4xx errors from the CloudFront CDN during the live stream? Was the CDN waiting for segments to appear at origin? Did the MediaLive Channel ever see a loss of input?
In my experience the vast majority of playback issues are in fact input issues where the streaming input coming into MediaLive experiences some transient networking slowdown resulting in a frozen output from MediaLive, which then propagates all the way to the viewer.
I recommend starting by examining the relevant metrics and logs in MediaLive to see if there are large spikes in input errors or output errors. This log inspection process should then be applied to each step of the workflow until the issue is located.
This blog post has links to scripts for quickly creating the relevant CloudWatch dashboard charts for MediaLive:
https://aws-blogs-prod.amazon.com/media/cs-quick-and-easy-media-services-dashboards/
Once you have a Dashboard created, you can adjust the time backwards to see the metrics captured during your event.
If you have a Support account with AWS, I suggest opening a support case so that a support engineer can pull the logs and metrics for the specific channel, Container and Distribution used during your event.
I hope this provides you with a plan for moving forward.