We are currently looking into a way to provide the ability to do parallel installs of a snap as part of a Juju charm (see the Grafana Agent charm on Charmhub). The feature still being described as an experimental feature is however making us hesitate a bit, even if it was introduced 5+ years ago, as we don’t want to rely on something that might quickly go away. Regardless of how unlikely that might be in reality.
What is the status of this feature, and is there a possibility we could bring it out of it’s experimental stage to allow for wider and more confident adoption?
i guess @mvo, for instance, might have an opinion or thought here?
There were some known issue with parallel installs and aliases that we never quite had the time to fix. However recently we managed to spent some time on this and snapd in edge should behave correctly now for aliases.
We also need user feedback on how well it works - we are fairly confident in the feature and have good integration tests for this but we all know that real-world feedback is the best here. So we plan a call for testing here soon. Once all of this is done and we got good feedback we can move the feature out of experimental.
I use parallel installs since years without any trouble on basically all my development machines (ubuntu,fedora,debian). But I probably have a very simple use case: i use it to have up to ~7 different versions of the single snap I maintain (different tracks, different channels, even branch channels). Since recently I also use the hold feature on some of the versions (mostly on branch channels).
I also advise others to use parallel installs for our snap, and yet there were no problems reported.
I never tried something more complex like parallel versions of content snaps or similar.
