Before the change: define what success means
Write the operational objective in testable terms. “Update a policy” describes an action; “the approved source can reach the required service while the denied source remains blocked” describes an outcome. Identify who can authorize the change and who decides whether the result is acceptable.
Collect the relevant baseline while service is in its expected state. Record the device, software version, routing or security context and time. Choose checks that can be repeated afterward with comparable conditions. A baseline assembled from different incidents may not be suitable for evaluating this change.
- Confirm target devices, affected services, maintenance window and change owner.
- Validate proposed inputs and commands against the applicable device documentation and approved process.
- Define expected changes and conditions that must remain unchanged.
- Prepare the recovery method and confirm that it is usable for the actual platform.
- Set stop conditions, communication contacts and the required approval before execution.
Use a small validation matrix
| Check | Before | After |
|---|---|---|
| Management access | Confirm authorized access and recovery access. | Verify that the team can still reach and manage the device. |
| Operational state | Record relevant interfaces, adjacencies and routes. | Compare against expected changes and investigate unexpected differences. |
| Service behavior | Run the representative application transaction. | Repeat from the relevant locations and record the result. |
| Security intent | Record permitted and denied test cases. | Confirm both the required access and the intended restriction. |
| Monitoring | Capture applicable alerts and performance indicators. | Look for new faults and define an observation period. |
During execution: keep the scope reviewable
Use the approved execution sequence and record what actually ran. If a step fails or returns an ambiguous result, pause according to the agreed stop condition. Continuing through the remaining commands can make it harder to identify which state the device reached.
An AI-generated proposal should be reviewed in context: correct device, correct mode, correct scope and appropriate sequencing. Approval should cover the proposal being executed, not a vaguely described goal. Separate read-only verification from write operations in the workflow.
After the change: verify the service, not just the configuration
Repeat the selected baseline checks and compare the results with the expected outcome. Confirm application behavior from the relevant source locations. Investigate unexpected routing, policy or health changes even when the intended configuration appears correct.
If the acceptance criteria are not met, use the authorized recovery decision. Reverting a configuration does not automatically prove service recovery; repeat verification afterward. Preserve the pre-change state, commands, approvals, observed results and final decision in the change record.
Example: after changing a route preference, the intended next hop may appear in the table while the application still fails because the return path is different. A route check and an application transaction answer complementary questions. Neither should silently substitute for the other.
Build a reviewable workflow with NetPrompt AI
NetPrompt AI can help draft configurable workflows from an objective, gather evidence and support remediation with human approval. Use those capabilities to organize the checks your team has defined. Validate every required command, connector and device profile before using the workflow on production infrastructure.
This checklist is an operating procedure, not a claim of built-in digital-twin simulation, automatic rollback or universal pre-change prediction. Recovery mechanics remain platform-specific. During evaluation, demonstrate the actual failure and stop paths as well as the successful sequence.
Common questions
- What is the difference between pre-change and post-change validation?
- Pre-change validation checks the plan, inputs, baseline and readiness before execution. Post-change validation checks the resulting configuration, operational state and service behavior against the agreed objective.
- Does a successful CLI command prove the change worked?
- No. It indicates that the command was accepted or processed, depending on the platform. Verify the intended operational and application outcomes separately.
- Does this checklist require autonomous remediation?
- No. It can be used with manually executed changes or an approved automation workflow. The review, authorization and verification requirements remain explicit.
Continue exploring
- Watch approved remediation
- Build a workflow from an objective
- Review security and governance
- Record findings in an RCA
- Investigate configuration drift
- Network blast radius analysis: scope the risk before a change
- Plan infrastructure end-of-support reviews with clear evidence
- Review AI-generated network commands before execution
Further technical reading
External references explain general practices; they are not endorsements of NetPrompt AI.