Serverless Docker AWS ECR failed connections still charges requests until timeout
In one of my endpoints that uses an ECR docker image (pulls it on new workers) and using the contaner registry auth from settings, couldn't load the image when runpod restarted my workers (because the token was probably expired).
But instead of not receiving requests, it does and run them until the timeout, resulting in a lot of money spend (by over 1000% the amount I am paying daily on a single day!)
How can I make sure this won't happen again?
3 Replies