Post #2105899

2026-04-30 13:51 UTC

@menos @esoterra What you're arguing for here is that "Copilot is down" is the same as "GitHub is down", and I don't think that's a good way to think about availability.

Replies (2)

@esoterra@hachyderm.io 2026-04-30 15:13

@yosh @menos I don't think it's "the one true way to talk about uptime" but "percent of time the system is not degraded" is something people care about. Let's call them AND-uptime and AVG-uptime. AVG-uptime is a better representation of how the component services doing, but AND-uptime is a better measure of how the system as a whole is doing. Especially, because many use cases will break if any part of the workflow does. If I'm trying to land a PR (git-ops) that runs actions where some steps use the API and load packages, I'm dependent on like 4+ of these services and the AND-uptime of them matters to me.

Open ##2105900
@menos@todon.eu 2026-04-30 16:15

@yosh That wasn't my intention; actually 'it would be a good idea to somehow distinguish between "one out of ten subsystems is down" and "the whole thing is completely hosed"' means the opposite. As @esoterra said, there's no one true way to look at it but more often than not your work will depend on several subsystems at once.

Open ##2105901