Hello, I am a bit at loss: I have created a Workfl...
# ask-metaflow
b
Hello, I am a bit at loss: I have created a Workflow template using the command
python flow.py --with retry argo-workflows create
But as I try to submit the workflow in Argo, I get the following error message
Copy code
fatal error: Unable to locate credentials
Failed to download code package from <s3://bucket-name/metaflow/FlowName>
Looking into the manifest, I see that my secret for accessing my MinIO storage gets loaded through
envFrom
so I'm unsure why the credentials cannot be found. Any inputs? thank you šŸ¤— (Note: I have already validated my artifact repository set-up with a classic hello world test)
I also tried adding the kubernetes decorator with the secret and service account specified on each step. How can i check the credentials necessary for MinIO get loaded / are in the manifest ?
Hello team @ancient-application-36103, @straight-shampoo-11124, not sure if I should have made a new post or ping you: I went with the later. If you would rather I make a new post, let me know. And thank you immensely for your help 😊 It's been very nice getting to use Metaflow and this is the last step in the flow šŸ’Ŗ Here is an update for this issue. On Argo Workflow side: Looking into container settings when I submit the workflowtemplate on argo, there is no
METAFLOW_KUBERNETES_SECRETS
in the Container Env (see screenshot). Maybe I am not on the right track with this, you tell me šŸ˜‰ Locally on VS-Code, trying to debug: • I also tried to use the debug tool and run the flow with just the
--with kubernetes
option. (I also tried specifying all configurations secrets, serviceaccount, namespace) The pod always fails with a
Datastore error:
message. I do not see the
METAFLOW_KUBERNETES_SECRETS
when doing
describe
on it • When I run
aws s3 ls --endpoint-url my_endpoint
(my_endpoint being a representative name of the http), the connection works, so the issue is not coming from a misconfigured endpoint I have also read over all the posts on this slack for similar issues. The following solutions were proposed: • a kubernetes secret such as described here for Airflow-Metaflow example: ā—¦ I have a secret configured, which I reference with the metaflow variable (Following the example mentioned, I do not have the metaflow helm chart installed, but this should not be necessary, right?) • checking service accounts: ā—¦ there I am not sure. I have given all argo-related service accounts admin rights (not best practice but this can be fixed later) to manage objects. however, there is no service account value set in the Env of the container on Argo (see screenshot), so could this be tied to the issue? • let me know if i missed anything else Apologies for the long message but I have tried several things and have not yet figured out why the credentials for my s3 store aren't available to the pod being launched. I was going to book an office hour to discuss it but it seems you've suspended that for the moment Thank you so much šŸ™ ⭐