I deployed my prod app and it went very smoothly. Then made a few changes and ran the suggested: ag ws patch --env prd --infra aws --name service
Logs are below. Everything normal. But new app is not deployed. Load balancer reports 5 clear health checks. I’m baffled (but I’m also not that experienced!). Any ideas? What is the best way to push changes to prod?
Update: I had to do some manual push of the image because it built from a previous one. Now I have it deployed. Will see what the next patch behavior does.
@Monali we still have major issues with this. Each time we use the ag ws patch approach, it does not create the correct image to deploy. Then we have to go in and manually create a new image and push it. But we are running into issues with that now. Why is the patch process not creating the correct image and directing the deployment to use that image? Are we missing something? Can you help with the proper commands to push changes to prod?
We’re using ag ws patch -e prd -i aws to deploy updates. The command successfully updates AWS resources like the ECS Task Definition and Service for our API (agent-app-prd-api-service).
However, it seems the built-in image build/push step isn’t working correctly – the deployed Task Definition ends up referencing an old image tag/digest in ECR, or sometimes even an incorrect image URI (like pointing to Docker Hub instead of our ECR).
We expect ag ws patch to:
Detect code changes relevant to the service.
Build the corresponding Docker image (e.g., using agent-app/Dockerfile.api for the API).
Push that new image to our ECR repository (e.g., updating the :latest tag).
Update the ECS Task Definition/Service to use that newly pushed image.
Our manual workaround (manual docker buildx push + manual ECS update) works, but we want the ag ws patch command to handle the image update correctly.
Question: How can we configure or troubleshoot ag ws patch to ensure it correctly builds (using Dockerfile.api), pushes, and deploys the latest image for the agent-app-prd-api-service? Is there a specific configuration or naming convention required to link the service definition within Agno to its corresponding Dockerfile for the build step?