containers-roadmap
containers-roadmap copied to clipboard
aws
[ECS] ALB slow start on ECS deployments
Community Note
- Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
- Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
- If you are interested in working on this issue or have submitted a pull request, please leave a comment
Tell us about your request In 2018, AWS announced slow start load balancing algorithm for ALB. Unfortunately, this feature is ignored by ECS.
The problem is that on a new deployment, when the new (and cold) tasks enter the "healthy" state on the target group, the old ones are shut down automatically, forcing the new tasks to exit "slow start" mode and receive 100% of requests all of sudden, leading to undesirable latency spikes.
ECS needs to shut down older tasks only after the slow start period.
Which service(s) is this request for? ECS.
Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard? I have low latency requirements, and my application takes about ten requests to warm up and respond with full performance. ALB's slow start feature would fit like a glove, but it's incompatible with ECS.
Are you currently working around this issue? No, but I'm open to suggestions.
Attachments
We are also looking for this feature and it seem to be a major gap in the original intent of the slow start feature.
This is really bad for production because the service stays with a low capacity. Please fix.
