Throughout the years, I watched how infrastructure migration projects succeeded, partially failed or created chaos right after a cutover action. And the accuracy of predicting a failure increased with a decrease in clarity just before the cutover. I’ve collected the following items to increase the success rate of any migration work, assuming that previous project phases have been dealt with:

  • Involve packet capture and analysis early in the migration project.
  • Establish a communication matrix of the current infrastructure.
  • Evaluate, discuss and agree the fidelity level of the communication matrix, i.e. the level of accuracy with which it represents the current infrastructure. The more accurate and comprehensive the matrix is, the less headaches and cognitive exhaustion the project team will have during the cutover.
  • Define the possible outcomes. This should include description of the expected technical infrastructure as well as the client-facing services that should resume working after the cutover. To minimize decision hesitation among stakeholders, define a Minimum Viable Outcome (MVO) document.
  • Elaborate a rollback plan that works. The validation of the rollback plan occurs in stages, where each stage starts with the finalization of each design module.
  • Define the order of execution of the migration steps using precise wording and include micro validation checks. An incomplete example is “implement security rules, implement OSPF routing then do a show ip route ospf”. A better example can be “Implement ruleset RS1. Verify reception of OSPF routes X1, X2 and X3.
  • Involve all possible stakeholders, including power users that have adequate access to the business applications. Ensure their work shift matches the project change window.
  • Establish active horizontal communication: This is the most trivial communication mode, where members of the project inner team exchange thoughts, agree on procedures and execute steps.
  • Establish active vertical communication: Keep the Team Lead / Director / CTO / CIO always informed on the recent developments of the project, the current issues and alignment with the project deadlines.
  • Establish Change Windows and Change Freeze Windows. Communicate them to the stakeholders.
  • For every traffic pattern of the current infra, there should be a corresponding traffic pattern in the target infra. Each traffic pattern of the target infra is ideally tested, documented in an isolated environment and discussed with the project team long before the cutover.
  • Capture the infra state: This involves for instance documenting the location of the IP default gateways, the router MAC addresses, NAT rule locations, DNS usage, etc.
  • Whenever suitable, plan redundant human resources that can execute assigned cutover actions in lieu of the primary resource. Physical sickness does not have an outlook calendar.
  • Avoid unecessary tensions. Even with good preparation, sensitivity to stress and potential for misinterpretations increase on a migration day.