Train pipeline of any ML model is failing with error An error occurred (InvalidAccessKeyId) when calling the ListObjects operation: Unknown. How to resolve this?
Error:
An error occurred (InvalidAccessKeyId) when calling the ListObjects operation: Unknown. can be seen in pipeline logs of a newly started pipeline
Resolution: Follow steps below,
- Log into any AIC server node and execute below commands
- export KUBECONFIG=/etc/rancher/rke2/rke2.yaml && export PATH=$PATH:/var/lib/rancher/rke2/bin
- kubectl get pods -n uipath
- Copy the pod name starting with "ai-trainer-deployment-xxx"
- Execute below command to shell into ai-trainer pod. Replace with podname copied from above command
- kubectl -n uipath exec -it -c ai-trainer-deployment -- bash
- Execute below commands to get the accesskey and secretkey used by ai-trainer-deployment
- cat /secrets/storage.accesskey
- cat /secrets/storage.secretkey
- Now exit shell of the pod using below command
- exit
- Execute below command to get the accesskey and storagekey stored in the secret
- kubectl -n uipath get secrets training-storage-credentials -o jsonpath='{.data.*}' | base64 -d
- Compare the accesskey and secretkey in step 4 and step 6. If both are different, proceed to step 8. Else, reach out to UiPath Product Support
- Take backup of training-storage-credentials secret using below command
- kubectl -n uipath get secrets training-storage-credentials -o yaml > training-storage-credentials.yaml
- Delete the training-storage-credentials secret using below command
- kubectl delete secret -n uipath training-storage-credentials
- Verify that the secret is deleted by checking if the secret is not present by using below command
- kubectl get secrets -n uipath
- Once verified, login to AIC host tenant using admin credentials
- Provision AIC for one of the tenants for which AIC is not provisioned earlier
- Now verify that the secret is created again by checking if the secret is present by using below command
- kubectl get secrets -n uipath
- Once verified, cross check if secretkey and accesskey match by performing step 4 and step 6 again
- Once the credentials are matching, run pipeline without facing this issue.