Would you go home to sleep confident without knowing that your data is backed up? Would you go home to sleep confident without knowing that your data is fully restorable? You wouldn’t, because you can’t know if your data will be fully restored until you have already tested it. You might have an evidence of your backup, but this is not warranty you can already restore it 100% to the most updated version.
Evidence of control means you have to probe, with minimal or no doubt at all, that you can restore the data after a logic/physical failure or corruption –or disaster if you want to go further-. People confuse this “evidence” by only getting backup reports that show a successful finisher ofevery backup job. But, you won’t be sure about it until you do a real restore of your data or service.
Try to google information about these practices and you will find out there isn’t so much. I’m proposing some tips that can be useful:
Avoid being judge and jury: The evidence needs to be collected hopefully by an external team different than the backup operators. That avoids having only one team being judge and jury. It must be a team with some technical skills to understand what they are checking out and being confident that everything is working ok.
Processes don’t warranty results: You can define processes in paperwork to try to warranty that everything goes perfectly to test backup copies. However, processes need to turn out into real operation practices. You can figure out a weekly and monthly dashboard that shows you the overall progress of these practices to complement evidence reports.
Innovation make things easier: If you innovate in backup technologies like using snapshots to get fast copies of your data –I suggest you to read my note “Snapshots are breaking bad into data protection strategies”– is many times better to speed the restore processes than traditional technologies like tape or disk. You can restore a huge database of many terabytes just within few minutes. How much time do you think you need to do the same through tape?
Orchestration helps: There are many ways to orchestrate restores; you don’t need to do it all manually. Recent backup copies are easier to test orchestrated from disk or VTLs, and if you are using snapshots is much easier. You can get such evidence starting databases instances from restored copies and query them and send the results to your inbox every evening.
Define your restore zone adequately: You will have to create a dedicated infrastructure or restore zone separated from your other systems. This is important to get a clean restoration of every backup copy and as warranty that you will be full recovered against any event. This way you will see it is you who will support every system type at the minute you will really facing a problem like obsoletes pieces of hardware or applications without any support from vendors.
Define scope depending on complexity: The cost to build and maintain restore zones aswell as the people to manage it –remember, they have to be an external or a separated team from backup operators– could become in something not affordable for your company. Then, you can work through different options to reduce the investment and the operation cost over the time: prioritize the most critical applications, or schedule tests on monthly basis, or do tests randomly if you have so many instances.
Cloud could be a nice solution to reduce capital expenses and get a better control for these practices. There are many options to backup your on-premise or off-premise systems to the cloud and you can pay for resources as computing and disk as you go to execute these tests. Also, you can look for platforms (PaaS) that have these capabilities by themselves and justpaying for extra costs, but nothing compares to have to build it or mange it by yourself.