[ECS] [request]: Add an ECS Service Event when using an impossible Deployment Configuration
- Dominant language
- Shell
- Stars
- 5.4k
- Forks
- 334
- PR merge metrics
- No merged PRs in 30d
Description
**Tell us about your request**
ECS should fire an Event exposed in the Events tab of an ECS Service, when a Deployment Configuration leaves you in a deadlocked situation with a certain Desired Count value. Right now this occurs silently with zero insight into what's broken.
**Which service(s) is this request for?**
ECS, probably impacts Fargate also?
**Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?**
Let's say you deploy an ECS Service with a desired count of 2, and use a sane deployment configuration of min 50 (1 task) max 150 (3 tasks). All fine, works as expected.
Then you adjust the desired count to 1 for some reason (code bugs in distributed architecture, whatever it may be), but forget to update the min/max Deployment Configuration. ***This leaves you in a state where ECS can't drop to 50% of 1, and can't increase to 150% of 1, effectively leaves your service deadlocked on the previous ECS task definition revision.***
What needed to happen to resolve it is change the min/max values to either 0/100 or 100/200 (depending on desired effect).
The impact of not solving this can lead to a developer spending endless time wondering "why did I deploy and update my task definition revision, yet ECS is not promoting my ECS tasks?"
**Are you currently working around this issue?**
In-house expertise to identify it when it occurs. In-house tooling to audit and validate the Desired Count against the min/max healthy Deployment Configuration values.
**Additional context**
N/A
**Attachments**
N/A
Contributor guide
Research direction
The issue names no repository files, tests, or implementation entry point. Start by reviewing ECS service deployment configuration and event behavior, then define the validation and event requirements for impossible desired-count and min/max combinations; done means the deadlock is surfaced in the ECS Service Events tab.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- aws
- Domain
- cloud
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100