Start with two links, not a synchronizer
IncidentRelay 2.4.0 does not ship a native Jira or Jira Service Management connector. The workflow below works by hand. The optional automation section is a design for a connector you operate, not a settings page in IR. There is no built-in Jira queue, retry dashboard or automatic status mapping to turn on.
Example: EU checkout times out at 14:02 UTC. Record impact and the operational assignee in the IR Incident; keep the source evidence and paging history in the linked AlertGroups. When the team identifies an unbounded database retry, create a Jira issue to cap it. Paste the Incident URL in the issue and put the Jira key and link in the shared incident log. Do not copy customer payloads or internal chat into a more widely visible project.
Give each question one owner
| Question | Authority in this workflow | What not to infer |
|---|---|---|
| Has someone acknowledged the alert? | IR AlertGroup | An assigned Jira issue does not acknowledge a page. |
| Is checkout working again? | IR Incident owner, using customer-path evidence | Moving a ticket to Done does not prove recovery. |
| Did the permanent fix ship? | Jira issue owner, with a deploy or test link | Resolving the Incident does not complete the engineering work. |
Write a ticket the next engineer can finish
Title: Bound checkout retries when the payment database is slow.
Origin: [IR Incident URL]; see linked groups for the alert history.
Owner: checkout team. Scope: timed-out writes, not all SQL calls.
Acceptance: exercise a slow write in staging and verify the retry bound.
Evidence: attach the test result and deployed revision.
Not a closure condition: the production Incident has been resolved.
If support also uses Jira Service Management, a customer-facing incident ticket and an engineering task can both be legitimate. Label their purposes and audiences. Keep service restoration updates in the former and acceptance criteria in the latter; do not broadcast every engineering comment to customers.
If you build a bridge, rehearse an ambiguous success
Suppose Jira created the issue but your client timed out before seeing its key. A blind retry can create a second issue. Give your bridge a durable mapping keyed by the IR Incident identifier, destination and purpose. Reconcile with the destination before retrying an ambiguous create. Do not assume a Jira create endpoint will deduplicate requests for you.
Test with a disposable project and a synthetic incident. Pause delivery, record two authoritative updates, then restore access and replay one event twice. Pass criteria: one intended issue, a visible failed or pending attempt, and no change to IR state merely because someone edited Jira. Also try a permission error: surface it to the operator instead of retrying forever. These checks belong to your bridge, not to IncidentRelay.
Keep a fallback that works during the outage
If Jira itself is down, keep responding in IR. Write “follow-up pending,” a named owner and a review time in the shared log. Once Jira recovers, that owner checks for an existing issue before creating one, adds the link, and verifies it from both sides. It is less glamorous than a synchronization diagram. It also works at 03:00.
For what the release actually exposes, see the Incident model and the integration catalog.
Want to make this less theoretical?
IncidentRelay gives you schedules, rotations, routing, escalations and alert actions in one open-source, self-hosted package. The pager may still be rude, but at least it will be organized.