Custom Action Server won't run after following the RasaX Tutorial

[Deprecated] Rasa X Community Edition
1

Hi there!

I tried to use the RasaX tutorials to set up my Deployment for user testing, but I can’t get the Custom Action Server to run, it seems.

Here are the things I’ve done:

  • Followed the Google Cloud Compute Engine VM Installation guide to set up the VM. I only spared enabling TLS by now.

  • Followed the Rasa Ephemeral Installer guide to install RasaX on the VM

  • Connected my GitHub repository with RasaX and was able to train a model on RasaX and talk to the bot there.

  • Used the Deploying Your Action Server tutorial with GitHub Actions to automatically build an image and push it to my DockerHub repository. Although the job on GitHub throws me a warning (“One or more build-args [DOCKER_IMAGE_NAME DOCKER_IMAGE_TAG GITHUB_REF GITHUB_SHA] were not consumed”), the jobs seems to be finished and new entries show up in my DockerHub repository.

  • Because I set my DockerHub repository to private, I used Creating a Secret with a Docker config by kubernetes and Adding Image Pull Secrets for Private Registries to create the secret and the values.yml file which refers to it.

  • Then I deleted my RasaX deployment and set up a new one with “rasactl start --values-file values.yml” to make sure RasaX can access the Image Pull Secret (according to the tutorial)

  • Connected the new RasaX deployment to my GitHub repository as before.

Now I can talk to my bot in RasaX, but everytime he should use a custom action (like a form) it seems to fizzle. It predicts the right custom action with a confidence of 1.0, but then “rewinds” and enters action_listen without performing the action. So i assume that the action server isn’t running at all.

As I am very unexperienced with all the tools and systems interacting here, I’d like to know where I messed up and how I could check if the image build by GitHub and pushed to DockerHub is valid and if RasaX can access it correctly?

2

Hello,

with kubectl you can interact with k8s (KIND) and debug the running rasa deployment.

check the rasa namespace for error events and if the pods are in a healthy state

kubectl -n rasax get events
kubectl -n rasax get pods

if you find that the custom server pod is not running or restarting you can check with

kubectl -n rasax describe pods <podname>

what container image is used and what error caused the pod to exit

kubectl cheat cheat to get you started

1 Like