I recently deployed a new version of my application on Amazon ECS on October 24, 2025, and while the deployment showed as successful, I noticed that after a while, the app started acting like it was running an older version of the code. It's strange because there were no rollbacks or errors reported. Additionally, I couldn't find any logs in CloudWatch during that time, which added to the confusion. After pushing a new change and redeploying, everything returned to normal, but I'm curious if anyone else has faced a similar issue, especially in light of recent AWS problems.
3 Answers
Have you checked the ECR for the image tag you were using? Sometimes, an older version might get pushed accidentally, or you might have selected an older tag by mistake. It's worth double-checking that.
It sounds like there might have been an ECS control plane desync. AWS can sometimes reuse an old task definition or a cached container layer, even when the deployment shows as successful. I recommend forcing a new task definition revision and stopping all running tasks before redeploying. Also, make sure your CloudWatch log group and agent bindings are correct. There have been some recent backend replication delays that led to similar issues for others.
I rebuilt it after a new commit, and that seemed to solve the issue!
You might want to check the version consistency setting. Sometimes this can cause issues like what you’re describing. Make sure to verify if it’s enabled or disabled for the container you're using.
I checked the ECR and the tag wasn't the problem. The changes were tested in production, and the issue happened even without any new changes. Just wanted to see if anyone else had experienced this with the latest AWS problems.