Incident & On-Call Workflows

When a service breaks at 3 a.m., the responder does not want to hunt through three tools to answer two questions: who owns this, and what do I do about it. A developer portal that already knows every service, its owning team, and its documentation is the natural place to answer both — if you surface on-call and incident context directly on the entity page. This section covers wiring on-call schedules, incident status, and runbooks into the catalog so the portal becomes the first stop in a response, not an afterthought. It sits within Developer Experience & Self-Service Platforms, because reducing time-to-context during an incident is squarely a developer-experience problem.

Incident context on the entity page The catalog entity links to on-call schedule, live incident status, and runbooks, giving a responder one place to start. entity page service + owner on-call schedule live incident status linked runbooks
The entity page is the hub — on-call, incident status, and runbooks hang off the same service the developer already found.

Why the Portal Belongs in the Response Path

During an incident, minutes spent finding context are minutes the outage continues. The classic failure is a fragmented toolchain: paging lives in one product, incident timelines in another, runbooks in a wiki nobody bookmarked. A responder alt-tabs, copies service names between search boxes, and pieces together what the portal already knows in one graph. Because the catalog is the single place that maps a service to its team, its docs, and its dependencies, it is the cheapest place to also show who is on call and whether an incident is open — no new source of truth, just a new view over the one you have.

Fragmented toolchain versus portal hub Without the portal a responder alt-tabs between paging, incidents, and a wiki; with it, one entity page gathers all three. Fragmented paging incidents wiki responder alt-tabs, copies names Portal hub entity page on-call · incident · runbook one place, no hunting unify
The portal does not add a fourth tool — it collapses three lookups into the one page the responder already opened.

The Two Integration Directions

On-call and incident data flow in to the portal from your paging tool — the portal reads PagerDuty (or Opsgenie) and renders the current on-call and any active incidents on the entity page. Runbooks flow the other way: they are authored as content and linked out from the entity via annotations, so the runbook is one click from the service it rescues. Keeping these directions straight is what stops the integration from becoming a tangle — one is a read of live state, the other is a durable link.

Two integration directions Live on-call and incident state is read into the entity page; runbooks are linked out from the entity via annotations. entity page paging tool on-call + incidents runbook docs authored content read in link out
One direction reads live state that changes minute to minute; the other is a stable annotation that rarely changes — treat them differently.

Owner Accuracy Is the Foundation

Every one of these integrations resolves through the entity’s owner. If a service’s spec.owner points at a team that no longer exists or never took over, the portal will proudly show the wrong on-call and page nobody. Incident tooling amplifies whatever the catalog believes, so the value of these workflows rests on the ownership discipline established in Service Onboarding Automation — get ownership right and the rest is wiring; get it wrong and you have built a faster way to reach the wrong person.

Owner resolves to on-call A service resolves through its owning team to a paging schedule to the person currently on call. service owning team paging schedule on-call person
Every hop after "service" inherits the owner — a wrong owner silently misroutes the entire chain to someone who cannot help.

Frequently Asked Questions

Does this replace our paging tool?

No. PagerDuty or Opsgenie remains the system of record for schedules, escalation, and alerting; the portal is a read-only window that shows the current state next to the service. You still acknowledge and escalate in the paging tool — the portal just saves the responder from hunting for which service maps to which schedule.

What if we do not use PagerDuty?

The pattern is tool-agnostic. Any paging provider with an API for “who is on call for this schedule” and “list open incidents” can be surfaced the same way; only the client differs. The two how-tos below use PagerDuty concretely, but the entity-annotation approach and the read-live-state model carry over to Opsgenie or a homegrown system.