[aws batch] Gracefully handle network-level errors during uploading about cli HOT 6 CLOSED

👍 will provide next time this happens

from cli.

My intuition tells me this is most likely a network- or host-level problem that the CLI can't solve. We might be able to provide better messaging by catching the error, though I'm not sure yet what that message would be.

from cli.

Based on the OS-level error message (socket.gaierror: [Errno 8] nodename nor servname provided, or not known), my guess is that this is a DNS resolution issue, i.e. DNS is timing out or returning NXDOMAIN.

@jameshadfield Could you provide (in private if you wish) the redacted endpoint URL in Could not connect to the endpoint URL: <redacted> above? This should include the domain name which didn't resolve for you. I expect this is an AWS S3 endpoint which probably includes a bucket name.

from cli.

It's related to uploading the large (400+Mb) repo which is needed to run SARS-CoV-2 builds currently.

Hmm, do you have the sequences data downloaded locally when you launch the build? If you can start from a clean ncov repo (rather than a partial build), it should be much smaller. Mine is 57 MiB uncompressed. The build will automatically download the latest sequences, though you'll need your AWS credentials in the environment.

from cli.

Hmm, do you have the sequences data downloaded locally when you launch the build?

Yes (and this is necessary sometimes, e.g. testing changes to the data that are not on the s3 version). Does the localrules not apply on AWS?

from cli.

Understood that it's sometimes necessary.

"localrules" does not apply here. The workflow runs on the Batch node directly (e.g. it is the "local" host), so Snakemake doesn't consider any remote job execution to be happening.

from cli.