Is there a way to reduce the time for a service to get recognised as being stable? I have a deployment pipeline set up via Jenkins. The build time is ~8 minutes only due to this last step
aws ecs wait services-stable --cluster $C --service $S
which roughly takes 5-6 minutes. The target group is set up to be healthy after 2 checks. The check is done every 20 seconds, so this should get flagged as healthy after 40 seconds if I'm not mistaken.
I found the issue was related to my load balancer's target group Deregistration delay.
(From the gui)
This was set to 300 seconds.
I also found the answer in here: https://forums.aws.amazon.com/thread.jspa?threadID=257357
To quote one of the relevant answers.
Are you using target groups? My first thought is it's draining and waiting for a state change from that. Default deregistration_delay.timeout_seconds timeout is 5 minutes.
There is a fantastic article about Speeding up Amazon ECS container deployments that includes answer to this question and offers other ways to speedup deployments.
Regarding this settings, if you have a classical REST API app, then you can reduce this settings deregistration_delay.timeout_seconds to 5 sec.
https://nathanpeck.com/speeding-up-amazon-ecs-container-deployments/