How severe does this issue affect your experience of using Ray?
- High: It blocks me to complete my task.
Hi!
I have some data I’m trying to process on my worker nodes and my understanding is that I’d need to use Remote URIs because even when zipped it’s greater than 100 MB. I zipped it up and put it in S3 but I get the following error whenever Ray tries to create the working_dir:
Failed to create runtime_env for Ray client server, it is caused by:
botocore.exceptions.NoCredentialsError: Unable to locate credentials
I have set up AWS credentials, and the following runs without issue:
import io
import os
import uuid
import boto3
def list_bucket_contents(s3_resource):
bucket_name = f"my-tmp-bucket"
bucket = s3_resource.Bucket(bucket_name)
try:
for o in bucket.objects.all():
print(f"\t{o.key}")
except:
print("Error")
if __name__ == "__main__":
list_bucket_contents(boto3.resource("s3"))
I was wondering if anyone had any insight on the following:
- What’s the best way to send a large amount of local data to a remote Ray Cluster?
- Is there an alternative to uploading this data to S3 and then setting it as the working_dir (uploading it is kind of slow)?
- Am I missing something with the Remote URI setup to be able to use an S3 bucket as the the working_dir?
Additional Notes:
- I’m running Ray in Docker
- I’m running my code in Jupyter, which connects to my remote Ray Cluster