[feature] Remote cache without moonbase account

[prabirshrestha]

Is it possible to have remote cache similar to how Rush does it as mentioned here.

• Directly depend on Azure blobs or AWS S3 then use native expiration it supports. S3 compatibility can be useful for self-hosted scenarios too.
• Allow CI to write to cache but devs in local can only read from cache.

[milesj]

The goal is to support different cloud providers through moonbase, but it's low priority at the moment.

[akshat99812]

@milesj @prabirshrestha I am new to open source I want to contribute to this repo ,would be a great help if you can guide how do I start

[doron-cohen]

This is required for the organization I work for. I would not be allowed to store cached artifacts outside our cloud accounts. Even if they are encrypted the process with infosec will take months. NX has this built in.

[milesj]

We're planning on supporting bazel remote APIs, which would enable bazel-remote and other services. We will retire moonbase at some point.

[doron-cohen]

@milesj could you please share when supporting an open source / S3 cache is planned to? This is a blocker to start using moon in our repo

[milesj]

Unfortunately there is no timeframe.

[MrLuxyfer]

Would it be possible to symlink the cache folder to a network mounted drive in order to share the cache with the team?

[milesj]

@DialLuesebrink Interesting approach, never tried it or accounted for it, so not entirely sure if it would work.

[MrLuxyfer]

I would assume that this might require minor changes on how the files are written though in order to work properly. Eg if the file is copied with the proper name already while someone elese tries to unpack the file. Thus the files should be written with a different name like xxxxx.tar.gz.tmp until they are fully written and then renamed to xxxxx.tar.gz . Thus all reads have the finished file.