Spack in Airgapped Environments #44573
Replies: 2 comments 1 reply
-
|
Creating a mirror is the way to go, and then there is an extra step you need to do to put the bootstrap binaries in your mirror: For the three issues above:
|
Beta Was this translation helpful? Give feedback.
-
|
@teaguesterling lots of good points here. It would be nice if we could require packages that do (C) to place these fetchers in a member function so it can be overridden from an inherited package. One of the main ways I've seen projects handle these sorts of air-gap constraints is to either a) write their own package.py in a custom repo, or b) create a package in a custom repo that inherits from the builtin repo. The former is typically for teams who have their usage of a package dialed or run into issues with the builtin that can't be overridden via inheritance. This clearly won't handle all the issues, and I like your idea of a monitoring tool. It would be nice if we could add this to CI and audit/score packages based on their compatibility with air-gapped systems. I suspect many package maintainers don't think about these issues at all, but would if they were made aware at the PR level. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I'm curious about what approaches people are taking to use spack effectivity in airgapped environments. I've had decent success by creating a local mirror with simpler packages. However, I've run into lots of headaches with a couple of different kinds of packages and was curious what patterns people are using.
The sorts of packages that are problematic:
Is this a lost cause, is something fundamental I'm missing, or is the a roadmap that I could help contribute to in order to implement some of these needs?
For the record, py-torch is all of the above :-).
Beta Was this translation helpful? Give feedback.
All reactions