Unauthorized changes: who decides retain vs. rollback, and who clears the flag?

RohitJ432388693
Tera Contributor

Hi all,

We're designing our process for handling unauthorized changes detected on CIs, and I'd like the Community's view on how responsibilities should be split. I'm not asking how to clear the Unauthorized checkbox. I'm asking who should own each decision, and in what order.

Our setup

  • When an unauthorized change is detected on a CI, a change record is created with Unauthorized = true (Emergency type).
  • We populate the Support Group field on cmdb_ci. Our intention is for that CI's support group (mostly technical/operations staff) to investigate and decide whether the change should be retained or rolled back.
  • These staff don't normally hold the change_manager role.
  • I recently learned that a user with change_manager can convert an unauthorized change into an emergency change model.

Diagram 1 – Original flow

The Support Group received a task that followed our existing Emergency CHG approval workflow, then approved (keep) or rejected (roll back).

1_original_flow.pngDiagram 1 – Original flow

Diagram 2 – Current flow

We replaced the approval task with a direct "accept / retain or roll back" decision by the Support Group. Rollback tasks go to the person's own Support Group (if a person caused it) or to the Change Implementer's Support Group (if automation caused it), with a fallback Unauthorized Change Support Group. Retained changes close the CHG, and rolled-back changes cancel it.

2_current_flow.pngDiagram 2 – Current flow

Diagram 3 – Proposed revision

The main change is that accepting a change that has already been made and clearing its Unauthorized flag become two separate decisions. The yellow boxes are the points I'm unsure about.

Diagram 3 – Proposed revision (yellow = open questions)Diagram 3 – Proposed revision (yellow = open questions)

My questions

  1. Should the support group's decision be only a technical recommendation, with a separate Change Manager giving formal approval and handling any reclassification?
  2. When should an unauthorized change be converted to an emergency change? Is deciding to keep it enough reason to convert, or should it stay recorded as unauthorized even after it's accepted?
  3. Would you grant change_manager to technical support group members for this, or keep that role separate? What permission or governance issues would granting it create?
  4. If the roles stay separate, where should the handoff to Change Management happen? Who owns the change record, and who owns the investigation and rollback tasks?
  5. For rollback, should we use the original record or a linked remediation change? And what closure outcome should replace the automatic "Cancel CHG" in our flow?
  6. How do you keep unauthorized-change reporting accurate after a record is reclassified?

Example: the support group investigates and says, "This change is technically acceptable; keep it." What should happen next? Who approves it, who changes any flags or the model, and which state should the record move to?

I'd appreciate a recommended sequence of steps and owners for both the retain path and the rollback path, or a pointer to how you've handled this in your own instance.

Thanks,
Rohit

1 REPLY 1

Shahjay
Mega Guru

Hi @RohitJ432388693 ,

Yes, keep the support group's call as a recommendation and let Change Management own the decision. That's also how the OOB feature is built. The unauthorized change is created with Assignment group = Change Management, and the Authorize approval is the after-the-fact acceptance. Once it's approved, the record moves to Review and a Post Implementation Review task is created for Change Management. The CI's Managed by, Owned by and Assigned to only get a notification. So adding an investigation task for the support group in front of that approval fits the model rather than fighting it.

On converting to emergency, I wouldn't treat "we're keeping it" as a reason. Clearing the flag is a reclassification: it says this shouldn't count as unauthorized at all. Valid reasons are a false positive, work that was covered by an approved change that just didn't match your detection query, or break-fix done during an outage, which is what the OOB outage route is meant for. A change someone made without approval that you've decided to live with is still unauthorized. It's just approved after the fact, so leave the flag set.

I'd keep change_manager away from the support groups, for two reasons. First, the person who made the change is often in the CI's support group, so they could end up approving their own unauthorized change. Second, the role covers every change in the instance, not just this decision. It's also one of the roles that can open Change > Administration > Unauthorized Change Properties, so someone could switch event processing off or narrow the CI class inclusion.

For ownership and handoff, this is what I'd do:

  1. The change record stays assigned to Change Management for its whole life.
  2. Your flow creates a change task for the CI's support group (cmdb_ci.support_group) to investigate. Give the task a choice field for the recommendation (retain / roll back) rather than relying on work notes. The flow can then branch on it, and you can report recommendation against decision.
  3. Closing that task is the handoff. It triggers the Authorize approval for Change Management.
  4. Retain path: Change Management approves, the record goes to Review with the PIR task, then Closed as Successful. Flag and model stay as they are.
  5. Rollback path: Change Management rejects, and a linked remediation change (Parent = the original) is raised and routed with your existing logic. Once the rollback is confirmed, through the next Discovery run or the CI owner checking the values, close the original as Unsuccessful with close notes pointing to the remediation change.

I'd use a separate change for the rollback rather than the original record, because the rollback is itself a change to the CI and detection will see it. Detection only treats a CI as covered when there's an open change matching the Change request query on the properties page (the docs example is active changes in Implement or Review). A rejected unauthorized change sitting in New won't match a query like that. So a rollback done under the original either raises another unauthorized change, or doesn't get flagged at all if it falls inside the notification ignore period (1 day by default).

On the Cancel step: OOB, a rejected Normal or Emergency change goes back to New, so the auto-cancel is coming from your flow. I'd drop it. Cancelled reads as "this never happened" and a lot of change reporting filters cancelled records out, but this change did happen. Getting from a rejection to Review and Closed will need a tweak to your flow or state model, so test that path.

For reporting, keep in mind the flag isn't only cleared by change managers. The docs also let an itil user clear it by creating an outage (type Outage) from the change. So counting records with Unauthorized = true will undercount over time. I'd add a read-only true/false field, something like u_detected_unauthorized, and have a business rule set it whenever Unauthorized becomes true and never clear it. Report "detected" on that field and "still unauthorized" on the OOB one; the difference is your reclassified count.