4 ms·
You hate starting an AMI and having a simple script to connect to an ECS cluster?
by heisnotanalien 9y ago
You hate starting an AMI and having a simple script to connect to an ECS cluster?
- tbrock 9y agoNo, that’s the easy part. I strongly dislike that auto scaling groups do not communicate with ECS (service scaling) and that updating your nodes to run a new ECS optimized ami is tedious and error prone. Want to scale up/down when you have high or low load? Cool, configure it at the ECS service level AND at the auto scaling group level. Will the auto scaling group take instances away before they’ve been drained of connections or randomly? Who knows the systems don’t talk to one another to coordinate: Fail What about just removing a node from the cluster when it fails? Decrease service count min temporarily -> Drain connections elb -> remove from target group -> kill instance. This is a 5-10 minutes process to do correctly that should be 1 click. Same problem, stuff doesn’t talk. Want to update to a new ECS optimied AMI to get updates and patches but ensure a minimum of core services keep running while you scale down after you’ve scaled up? Good luck!
- rem7 9y agoAlso, last time I checked IAM roles sucked on ECS. Not every container in the cluster should have the same role. I wonder if fargate fixes this.