-
|
Following examples from here we have build a logic around deploying a single workflow through llama deploy. We have 3 key elements all being invoked from a bash script with 10 seconds of sleep time between:
This was working fine up till recently, now on the first start it works fine, however if I decide to kill the workflow and redeploy it, then I get: |
Beta Was this translation helpful? Give feedback.
Answered by
masci
Jan 25, 2025
Replies: 1 comment 13 replies
-
|
@hristogg have you considered using the apiserver, or even deploy the llama_deploy official Docker image directly to Cloud Run? Example: Step 1: run the LlamaDeploy container, mounting the folder containing your workflow: At this point the control plane and the message queue are up and running inside the docker container Step 2: define your deployment in a See the quick start example for more details about this Step 3: deploy your workflow
At this point the workflow is deployed and waiting for requests, no need to add a custom Flask app in front of it Step 4: run your deployed workflow
Let me know if this is helpful for your use case and I can follow up with more details about how you could pack all of this into a Cloud Run instance, I'm working on something similar right now! |
Beta Was this translation helpful? Give feedback.
13 replies
-
|
Follow up making an async call - Executes the first run from the llamactl after deployment, the next one fails again however: |
Beta Was this translation helpful? Give feedback.
-
|
@hristogg looks like we have two bugs here, let's ignore the async/sync call causing the 500 for now, and focus on the previous message. Can you tell me exactly what's the error if you un-comment the LLM calls? If that's a 500, do you see anything in the server logs that might point us in the right direction? |
Beta Was this translation helpful? Give feedback.
-
|
Thanks @masci log_downloaded-logs-20250128-133745.json |
Beta Was this translation helpful? Give feedback.
-
|
@masci sent you an email as well with a full recent log. hope it helps. |
Beta Was this translation helpful? Give feedback.
-
|
@hristogg thanks, got the email, I'll go through the logs soon. In the meantime I think I've found the problem with the HTTP 409 error, currently working on proper end-to-end tests. I'll keep you posted. |
Beta Was this translation helpful? Give feedback.
Answer selected by
hristogg
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
@hristogg have you considered using the apiserver, or even deploy the llama_deploy official Docker image directly to Cloud Run?
Example:
Step 1: run the LlamaDeploy container, mounting the folder containing your workflow:
docker run -p 4501:4501 -v ./my-app:/opt/my-app -w /opt/my-app llamaindex/llama-deploy:mainAt this point the control plane and the message queue are up and running inside the docker container
Step 2: define your deployment in a
deployment.ymlfile, something like