Skip to content

Surface failed/unhealthy computing units instead of showing "Connecting" indefinitely #7669

Description

@eugenegujing

Feature Summary

Image

Caption: the computing unit dropdown. CU 2 is healthy (green dot, selected). CU 1 sits at a gold dot with the literal (Connecting) label — that is the only thing the UI ever says about it, whether it is two seconds into startup or has been dead for an hour. Hovering it reads "Computing unit is starting up".

When a computing unit (CU) becomes unhealthy — the pod crashes, gets OOM-killed, is evicted, fails to pull its image, or the local CU process dies — the UI gives the user no indication that anything is wrong. The run button just says "Connecting" with a spinner, forever, and the dropdown entry shows "(Connecting)" with a tooltip that reads "Computing unit is starting up". There is no timeout, no error state, and no way for the user to tell "still booting" apart from "dead and never coming back".

Proposed Solution or Design

1. Add terminal/unhealthy values to ComputingUnitState (Failed, Unknown, Terminating) so a failure has somewhere to live. Requires the matching widening of the frontend DTO type, which is status: "Running" | "Pending" today.

2. Map real pod state in ComputingUnitHelpers instead of collapsing it, so pod phase Failed/Unknown, a deleted or evicted pod, and container-level states (CrashLoopBackOff, ImagePullBackOff, OOMKilled) stop reporting as Pending. Carry the reason as a statusReason string on the CU DTO so the UI can say why.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions