Rasa X buggy: able to deploy on EC2 but not functional when uploading models

Rasa Open Source
1

Hi Rasa team,

I’m having some issues getting Rasa X to work properly in EC2 although I’ve managed to deploy it using the One-Line deploy script on https://rasa.com/docs/rasa-x/0.26.4/installation-and-setup/one-line-deploy-script/ curl -s get-rasa-x.rasa.com | sudo bash

My EC2 is clean brand new instance (t3a.large) running on CentOS 7 x86_64 HVM EBS ENA 2002 with 100 GB storage, 2 vCPU, 8 GB RAM

image
image1100×284 42.4 KB

I was able to get it running the first time, changed my password, logged in, uploaded a model… but after that, it was really not functional…

I checked that I met the minimum requirements for Rasa X but somehow it seems like it’s still buggy? Or am I missing something here…

Pls advise. Thanks for your help in advance!

2

Hi @wchoo, can you describe the issues you’re encountering?

3

Hello,

Thank you for getting back to me.

One of the issues I observed was that when I uploaded a new model, it seemed like it uploaded, but then when I tried to check the intents, it was still ‘caching’ the older intents from a previous model (that I had deleted). I also had some problems uploading new stories YML as it always gave a 500 error or something.

For now, I have re-deployed using the docker compose method and it seems to be much more stable and I could upload models etc. So I’m not sure if it’s due to the overheads with kubernetes that might be causing the earlier issues I observed. At this pt, I am happy with the docker approach, but just wondering if there’s something that really I missed out, or are there inherent configuration differences between both methods.