The principle of the migration day: repeatability and evidence
The migration of the structure is repeatable: each run builds the structure from scratch, the save runs in a single transaction with a trial mode, meaning a rollback, and it is preceded by a copy of the structure to an SQL file. A mistake is therefore not final, and the coordinator has something to return to.
The script does not save any data if one of the consistency gates is different from zero. Each gate has a red test: an injected old error turns the gate on. The checklist makes sure these mechanisms are used in the right order.
Before the save: backup and trial
First a copy of the structure is made to an SQL file, and the coordinator checks that the file exists and is larger than zero. Then the script runs in trial mode: it performs the whole save and reverts it, and the coordinator reads the result of the gates.
If any gate is different from zero, the save is not performed. The coordinator records which gate turned on and passes it to the team for a correction of the script or the data.
Order of steps on the migration day (step number)
Save and independent check
After a green trial the save runs in a single transaction. Right after it, a second, independent check runs: it reads the result through the editor's API and compares the state by content: name, area, level and geometry. The coordinator does not close the day until the results of the check are recorded.
Roles and division of responsibility on the day of the save
The implementation coordinator runs the list and takes the decision to save. The script's author runs the trial and the save, but does not assess the result of the independent check. That assessment is made by a second person, preferably an internal auditor or a representative of the OHS department who knows the plant's structure.
Such a division protects against a self-inflicted error: the author sees what they intended to save, and the checker sees what actually ended up in the system. When both people agree, the coordinator closes the day, and the result goes into the deviation report.
It is worth planning the migration day outside the hours of greatest traffic in the hall, so that the message to users reaches them in advance and a possible rollback does not coincide with the shift in which the most people work.
Message to users and rollback plan
Users learn in advance when the save takes place, what will change on the map and whom to contact with comments. The message contains the time, the scope of changes and the coordinator's contact. Paper registers keep working, because the plant is in the parallel run period.
The rollback plan names the copy of the structure, the person authorised to restore it and the decision criterion. A rollback is justified when the independent check shows deviations that cannot be corrected in the next run on the same day.
Checklist
Video
Frequently asked questions
What do I do when a gate is different from zero?
The save is not performed. The coordinator records which gate turned on and passes it to the team for a correction of the script or the data.
Can the migration be repeated?
Yes. The migration is repeatable, because each run builds the structure from scratch.
Do I have to switch off paper on the migration day?
No. On the migration day the paper registers keep working, because the plant is in the parallel run period.
Let us go through the migration day together
We will prepare a checklist for your plant and rehearse the save in trial mode before the actual day.
Let's talk