Repository navigation
Rig reads "degraded" while on-demand members are intentionally stopped #401
Description
Version: @openrig/cli 0.6.0
What happens: a rig whose on-demand members are stopped on purpose (rig seat stop <seat>, to be launched again when needed) reads degraded in rig ps and the TUI for as long as they stay stopped.
Why: deriveRigLifecycleState (daemon/dist/domain/ps-projection.js, around lines 30-52) returns degraded for any mix of running and non-running nodes, so an intentional stop and a crash look the same.
Expected: a way to mark a member as on demand, or an intentionally stopped state recorded by rig seat stop, that the rig-level rollup does not count as degraded.
Activity
mvschwarz commentedon Oct 2, 2026
Thanks for this. You're seeing a rig read
degradedinrig psand the TUI while its on-demand members are stopped on purpose withrig seat stop, the same as if they'd crashed. It's with the maintainers, along with your idea of an on-demand or intentionally stopped state. You reported it on 0.6.0. Could you tell us which OpenRig version you're seeing it on now? We'll update here.— comm-replies@v-openrig-build, on behalf of @mvschwarz
mvschwarz commentedon Oct 2, 2026
A follow-up now that we've looked: today
rig seat stopand a crash both read asdegraded, because the rig rollup counts any mix of running and stopped members that way. Marking members as on demand, or recording an intentional stop that the rollup doesn't count, is a fair feature request, and it's noted. The code is unchanged on current main, so there's no need to check your version after all.— comm-replies@v-openrig-build, on behalf of @mvschwarz
mvschwarz commentedon Oct 12, 2026
An update, @aanestesia: @Aschenbath has opened a pull request, #1192, which proposes recording an intentional
rig seat stopso that a rig no longer readsdegradedwhile its other seats are healthy. It's in review, and we'll update here.fix: distinguish intentional seat stops in rig health (#1192)
mvschwarz commentedon Oct 13, 2026
An update, @aanestesia: #1192 is merged. On main,
rig seat stopnow records an intended stop, so a rig whose other seats are healthy no longer readsdegradedinrig psor the TUI, and a crash in another seat still makes the rig readdegraded. It isn't in a release yet. Closing this as completed, and thanks again for the clear report. If it doesn't cover your case, say so here and we'll reopen it.