New Test Instance + S3 Bucket for latest version of Bandersnatch

[cooperlees]

Hi all,

pypi/warehouse#8569 has sat dormant for awhile and I was wondering if I could have an instance and a s3 bucker for the sprints to proof of concept using S3 vs. the large posix storage volume for the internal PyPI bandersnatch mirror. I believe this could save PSF $$$s/tokens.

Bandersnatch s3 documentation: https://bandersnatch.readthedocs.io/en/latest/storage_options.html#amazon-s3

• 1 instance that's only going to sync
  • Syncing would run in a Docker container - I've made our bandersnatch docker publish a s3 container
  • Service would move to s3 static sites
• A s3 bucket
  • Would then need an access_key_id + secret_accss_key with write privledges

I am happy to ansible/salt infra by code all this however wanted. The deployment will be in docker. I would need help / skills with the S3 static sites as I've never done that.

Once this is up and POC'd (I can even do it with a subset of packages if we wanted) we can then look at the options to move PyPI from current bandersnatch fallback to s3 fallback and decommision the old system once happy.

In an ideal world, I would then love to get a second bucket where I could test new versions of bandersnatch (from same instance) with a subset of packages before upgrading in the future to do safer docker pulls etc.

[cooperlees]

We're now all python3.11 and PEP691 ready ... Is there any chance of this ever happening?