The cloud can be a great way to protect your business critical applications in the event of a disaster. It can offer flexible services to help get your business back on track quickly. But just because your cloud provider offers a great RTO (recovery time objective) it doesn’t mean that you can ACTUALLY recover from a disaster.
Let’s say a disaster strikes and the power is off for an extended period. The southern Alberta floods are an example of a disaster few companies were prepared for. Your main systems are now unavailable after the batteries in the UPS ran out. What do you do? You enact your Disaster Recovery plan and start turning on the systems that you replicated to the cloud, right? But wait a minute; your Disaster Recovery plan document was stored on the file server that is now offline so you don’t know which servers have to come up first. Someone had a copy in the office but the office tower is inaccessible and has no power. The point here is that the first step to recovery is to be prepared.
Even if you are prepared when disaster strikes you still need to make sure you can recover. Let’s say your critical NetApp® CIFS data is replicated to the cloud via Snapmirror®. All the data is there but when users try to access it they are getting an access denied error. It turns out that the share permissions have not been copied to the cloud and they now have to be recreated before users can access their data. The point here is to make sure all data including metadata is copied to the cloud and all the dependencies that rely on that data are understood.
Dependency maps become extremely important. If your production applications are replicated to the cloud but their required database servers are not – there is no chance of recovery for that application. A good cloud provider will work with the customer to assist with dependency mapping and have the expertise to identify potential issues during a Disaster Recovery scenario.
There are a number of things that they can help identify:
• Does the client have enough bandwidth to replicate everything to satisfy the RPO?
• What order do servers need to be powered on? (AD -> SQL -> App -> Web)
• Does routing or DNS need to change for external access?
• Do VMs have virtual disks on datastores that are not replicated?
• How do you fail-back to production afterwards?
With all the services that cloud providers can offer there is no substitute for good planning, periodic testing, and user training. Talk with your cloud provider about disaster recovery and make sure you are both prepared when the worst happens.