Recently we received a request to design a high availability backup solution for two physical servers running a handful of Virtual Machines on VMWare. The prime goal is to ensure that in case of a failure of one physical server all virtual machines would be brought back online in a timely manner. In the world of corporations and big budgets this can be achieved by deploying a cluster + SAN/NAS. Unfortunately associated fees and hardware costs are not realistic for a typical small / mid-size business. Here we discuss our solution that only required VmWare Essentials license and two licenses for Unitrends Backup Essentials ( <$1000 total )

The idea involves setting up twin-servers splitting the load of Virtual Machines 50/50 (let's say: 3 VM's on Host A and 2 VM's on Host B) while acting as Unitrends backups for one another. In case of one of the servers catching fire or flood damage or getting stolen the other one could take advantage of instant recovery whereas internal backup storage is used to bring missing virtual machines back online. Without vCenter it's still possible to do with Untirends "audit mode" instant recovery. In our tests it took less than 20 minutes to go through the steps of igniting "lost" VM's from a backup.

dual backup hosts
At this point we must admit a few drawbacks and points for caution:
  1. There must be enough spare RAM, CPU power and FAST backup storage for the ability to launch a quick recovery with proper performance. In our tests the servers had at least 50% slack in CPU power and RAM provisioned. Prime backup storage was run on a dedicated SSD RAID10. This approach delivered good performance results, certainly enough to cover a week of production time till a replacement is ready.
  2. The tricky part is moving "instant recovery" VM(s) to the replacement server. This would require downtime to move data in the absence of vCentre. However it's much easier to do when one got time to plan ahead.
Now, that we mentioned vCenter: VMWare essentials does come with it. So, one could deploy it on a standalone machine to orchestrate two servers. This would provide a natural way of Instant Recovery (vs Audit Mode) and going back to the production without much downtime. We haven't tested this route as our goal was to come up with a minimalist approach when it comes to budget. Besides, vCenter machine would introduce a single point of failure which would contradict the main objective of having as much redundancy as possible on a reasonable budget: Error Correction Code memory, RAID drives, redundant power supplies, dual UPS units. 

It's worth noting that with plenty of FAST internal storage one could opt out for a traditional recovery vs "Instant" one. Depending on the size of the backup and particular RAID / Motherboard specs it could be 0.5-1-5 hours of downtime. For example: at 1AM it'd be a preferred method for a law-firm, however at 1PM an instant route would be desired quite a bit more due to time-pressing issues.

Overall, our customer was satisfied with the achieved reliability / cost ratio. Besides, it plays along the classic approach of 3 backups philosophy quite nicely for Unitrends provide a smooth Backup copy solution.
© 2024 - Allora Consulting