Fix hard-coded mystical datastore endpoint names

[Toshbrown]

DATABOX_DRIVER_TWITTER_STREAM_DATABOX_STORE_BLOB_ENDPOINT! 👎

This has implications for the Container manager as well.

[yousefamar]

Just referencing this here. All containers are already passed their own hostname, so we could just say that for every driver, the matching store's hostname is some function of the driver's hostname (like some-driver and some-driver-store or so), then directly refer to their stores by that hostname and let Docker DNS do the rest.

I think at some point something like that was inside the CM anyway, to make sure that stores have different container names (and thus hostnames) since you can't launch more than one container with the same name, which was bound to happen for stores being called by their image name as you have many databox-store-blobs. Was probably lost to the aether over the months though.

[Toshbrown]

That is how DATABOX_DRIVER_TWITTER_STREAM_DATABOX_STORE_BLOB_ENDPOINT is generated. But the env var is also called DATABOX_DRIVER_TWITTER_STREAM_DATABOX_STORE_BLOB_ENDPOINT which requires driver/app devs to know how the name is generated.

I think the env var should be renamed to DATABOX_STORE_<STORE_TYPE>_ENDPOINT for simplicity.

[yousefamar]

Ahh, I see. 👍

[haddadi]

this sounds like a good plan for longer-term, as we can't manually manage all the endpoint names once there are hundreds of drivers, each wth their own store serving different end points to different apps.