Skip to content
This repository has been archived by the owner on Jul 21, 2023. It is now read-only.

Increase AWS MediaPackage HLS segment duration #338

Merged
merged 1 commit into from
Aug 17, 2021

Conversation

EdNutting
Copy link
Member

To reduce the probability of triggering the VideoJS segment buffering bug

What's improved

By increasing the segment duration from 1s to 5s, we hopefully reduce the probability of triggering the VideoJS client-side segment buffering bug by ~5x. Assuming probability works at scale the way I hope (and there are no other unintended side effects), this should dramatically reduce the impact on our end users while we wait for VideoJS to come up with a proper fix.

Details

Upgrading and Deployment

Instructions for how to deploy these changes to production. Please tick (i.e.
put an 'x' in) the boxes for the actions that are necessary.

  • Apply migrations to Hasura
  • Apply metadata to Hasura
  • Reload enum table/values in Hasura for [table names]
  • Re-deploy frontend
  • Re-deploy actions service
  • Re-deploy playout service
  • Re-deploy real-time service
  • Flush real-time service state (Redis/RabbitMQ)
  • Update environment variables for [name of component]

Any other steps necessary to re-deploy?

I honestly am not sure - I think any existing MediaLive endpoints probably have to be deleted so they can be recreated?

… segment buffering bug

By increasing the segment duration from 1s to 5s, we hopefully reduce the probability of triggering the VideoJS client-side segment buffering bug by ~5x. Assuming probability works at scale the way I hope (and there are no other unintended side effects), this should dramatically reduce the impact on our end users while we wait for VideoJS to come up with a proper fix.
@EdNutting EdNutting requested a review from rossng August 17, 2021 02:13
@EdNutting
Copy link
Member Author

I have tested this by manually making the change in one of our existing test streaming environments and it does appear to have prevented my original reproduction steps from retriggering the original issue.

@rossng If we aren't confident merging/deploying this, I have tested in AWS tonight and we can edit the MediaPackage endpoint directly to test it tomorrow.

@rossng
Copy link
Member

rossng commented Aug 17, 2021

Is it necessary to go to 5s? This will significantly increase latency.

@EdNutting
Copy link
Member Author

Yes. Unless we can fix the video player, all users receiving video reliably is far more preferable than 25% of the audience missing most of the talk because the player keeps freezing.

Copy link
Member

@rossng rossng left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That is a fair argument, and a small price to pay for better reliability right now.

IMPORTANT: you must tear down and re-create and existing channel stacks to benefit from this code change. Or you can manually edit the MediaPackage channel in AWS.

@EdNutting EdNutting merged commit 23f9109 into main Aug 17, 2021
@EdNutting EdNutting deleted the improvement/video-buffering branch August 17, 2021 20:00
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Development

Successfully merging this pull request may close these issues.

2 participants