Add contact_support on update status page

[david-crespo]

New boolean contact_support field on update status added in oxidecomputer/omicron#10271.

I tried it inside the properties table as Contact support: Yes and it felt terrible.

Robot notes on the API logic behind contact_support

omicron#10271 adds a contact_support: bool field to the system/update/status API. It is the last piece of a minimal system health check tied to update status, intended as a stopgap until the fault management subsystem lands (RFD 612).

What it means

When contact_support is true, Nexus has detected one or more known conditions in the latest inventory collection (plus a few additional checks) that require Oxide support to resolve. The field collapses several sub-checks into a single boolean because none of the individual conditions are actionable by the customer — the only action is to call support. The detailed breakdown is logged server-side and lands in support bundles.

The intended usage maps to two cases:

• Before an update: if contact_support is true, the customer should not start an update — resolve the issue with support first.
• After an update: if contact_support is true, something went wrong; the customer should call support immediately.

Conditions that trigger contact_support: true

• Unhealthy zpools — any zpool not in online state (e.g., degraded).
• Enabled SMF services not online — services that should be running but are in maintenance, offline, or degraded.
• Stuck sagas — sagas that have been running longer than ~15 minutes. (A sample of 10,000 done sagas on dogfood showed only 3 exceeded 15 minutes from creation to completion.)
• Stale inventory collection — no recent inventory collection (~15 min threshold), meaning Nexus has lost visibility into rack state.
• Stalled update — an update is supposed to be in progress but the planner hasn't taken a step in ~30 minutes.

The list is explicitly minimal and not exhaustive — contact_support: false does not guarantee the system is fully healthy.

Suppression during an active update

Health checks often fail transiently during an update, so the API suppresses contact_support: true while an update is genuinely in progress. The field only surfaces a true value when either (1) there is no update in progress, or (2) an in-progress update has stalled past the threshold (matching the 10–15 minute guidance) in the Reconfigurator Ops Guide for when support considers an update stuck).

In practice this means the field always presents in one of two contexts: the system is idle (pre-update or post-update), or the update has stalled long enough that the result is no longer a transient artifact.

Issues to resolve

• Explain the situation without overdoing it
• Tooltip looks terrible in message box, what if we link to docs instead
• Should probably link to a way to actually contact support, probably the support email that goes to Zendesk

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
console	Ready	Preview	May 22, 2026 6:06pm