plume vs aws cli about coreos-assembler HOT 5 CLOSED

Fair enough! Honestly I am just unfamiliar with the ore/plume codebases. I now notice that I basically reimplemented ore aws copy-region...will try to use that instead, and in general use ore more.

from coreos-assembler.

Historical context: the CL AWS release pipeline originally used shell scripts around an AWS command-line client. We moved to plume to get away from all of that. 😃

plume's idempotency is a really nice feature when the release process fails for some reason. Also, over time, plume has had to gain a number of workarounds for AWS bugs and eventual consistency issues, and I'd hate to have to replicate those in shell scripts. plume also supports Azure and GCE, and @sayanchowdhury is working on moving the Fedora AWS upload pipeline over to it.

What problems do you anticipate if we stay with plume?

from coreos-assembler.

I think that context for the action we're trying to solve for is important for this discussion, can you expand upon what specifically you're wanting to use them for (developer image upload, release publishing, etc)?

plume is specifically designed to handle releasing production images, it's entire focus is on idempotent release staging (for AWS this includes the image upload, replication to different regions, launch permissions, cross-partition handling).

from coreos-assembler.

plume also supports Azure and GCE

This is probably the biggest reason in my mind to go with plume (for production release use cases) and maybe ore for the developer/hacking use case (as @arithx mentioned above). i.e. we get one entrypoint and the details of handling talking to clouds and such is handled in that tool.

from coreos-assembler.

Closing, unless someone has a reason to keep this open.

from coreos-assembler.