Skip to content

Live updater reports failure during healthy Gateway startup warmup #104365

Description

@steipete

Problem

The live updater performs one immediate gateway status --deep --require-rpc after restarting the managed Gateway. On clawmac, a healthy Gateway took longer than that first 10-second RPC timeout while startup warmup was active. The updater exited update_failed, but the identical probe passed eight seconds later without another restart.

Impact

Successful live updates can be reported as failures, leaving maintenance state noisy and encouraging unnecessary reruns even though the exact-head Gateway is healthy.

Evidence

  • Exact deployed head: 8e2ddd60ad8865a2c27624eec05951a234bcdf76
  • First deep RPC probe: timeout after 10 seconds; service running and health marked healthy.
  • Same probe eight seconds later: RPC/admin capability/version all healthy.
  • Current updater has no bounded post-restart readiness retry.

Desired behavior

Retry only the post-restart readiness proof for a bounded startup window, while preserving exact-build assertions, restart-window log auditing, and fail-closed behavior after the retry budget is exhausted.

Metadata

Metadata

Assignees

No one assigned

    Labels

    maintainerMaintainer-authored PR

    Type

    No type

    Fields

    Priority

    None yet

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions