Duplicate of T275555 T225191?

See comments from https://gerrit.wikimedia.org/r/c/operations/software/tools-webservice/+/920259 also:

A bit of a rambling, but I see the storage split in different groups/use-cases.

1. Read only storages for retrieval of public data (dumps, mediawiki), this would be parallel to the replicas as information sources.
2. Read/write persistent shared storage (current scratch), this would be for sharing between users/projects.
3. Read/write persistent private storage (current project home), this would be for your own tool, private
4. Read/write ephemeral private storage (current /tmp, home project), this would be for temporary big file processing and such, I think currently people just use the home directory.

Each could be implemented in a different way (probably the best, as they are quite different), but we might want to create different abstractions on top of them so we can evolve them differently.

That might eventually come to have something like (not saying to implement now though):

--data-services-storage=enable // to enable mounting dumps/mediawiki/..., #1
--shared-storage=enable // this would be #2
--project-data=enable  // to enable mounting the project data directory, #3
--ephemeral-storage=enable // #4

Then for each, we could have values like "enable/disable/preview" or similar to choose if you wont the current supported one, disable it, or use the "preview" version.

Also, from the list thread https://lists.wikimedia.org/hyperkitty/list/cloud-admin@lists.wikimedia.org/thread/CHGYMEUYSIFU3AIVPXFKEFIDT7UK6YV2/ (storage section).

I would try to keep it transparent for the user if it's using cinder volumes, nfs shares, ceph rdb images, or anything else.

I think we are all in the same page. Refreshed the task description with all the additional info.