I have performed many swaps or switches for replication software for the IBM i and a question I hear often is “What is the difference between a swap/switches and a Virtual one?”
The feature of the virtual role switch/swap is a very useful tool. This allows you to switch your environment and allows validation testing on your target environment without impacting any of your users on your source system. There is no downtime and you can validate your data without having to update your network.
You may have applications you cannot test during a virtual swap but there’s still a good way to test your data and find issues. During a virtual swap, the only replication is ended. Mimix/ iTera keeps track of all changes that were made on the target during the virtual swap and once you swap back, you will undo any changes made to the target.
Journals will still be capturing the source system activity during this time and will be applied to the target once the virtual swap completes. The swap back time will depend on the amount of changes that were made during the testing during the virtual swap. This will be something you want to factor into your plan for any virtual swap. This is the best way to do destructive testing because during a virtual swap no data will be replicated back to your source system.
Replication role swap simulates the event of a true disaster in the event that your source system is unavailable. This will have the users utilizing the target and any changes will be synced back to your source system. This demonstrates that if you lost your system users, you would be able to work on the target. Once you got a source available again, the changes can be synced back with limited downtime. You can circumvent the sync back to production by saving your data before the swap on both source and target, restoring them back after testing is completed. You will need to regenerate your journals and restart the data groups from the new journals after the restore is completed on both sides. This would only be for destructive testing and is best to be done during a virtual swap.
I would recommend a planned swap four times a year, one full swap and three virtual swaps. Three times a year, you can validate your data and most applications depending on your environment. This will give you the confidence that you are replicating all your data and would be ready in a disaster situation. This reduces downtime, and if you’re a replication customer, downtime is probably in short supply.
Once a year, it is important to do a full swap to flush out any issues. This will need a short outage to swap the environments. It is much better to find out something needs attention during a test than during a disaster.