One of the most common conversations I have about virtualisation is the "how do I make my virtual machines highly available?" one. Topics like Hyper-V Quick Migration are then discussed and off the techie goes to start testing. When I revisit the discussion I have noticed the false sense of security people get just because their virtual machines are highly available. HA options for VM's do not mean stop worrying about host clustering, network load balancing or traditional backups ... no matter how clever the technology. Highly available virtual machines no longer have the single point of failure at the (host) hardware level .. and that's about as far as it goes. You still have to mitigate against the same risks at the OS level (and above) regardless of whether or not its physical or virtual. Okay, some of you are saying duh! Bear with me though. It's not uncommon for people to think just because they have a Hyper-V cluster or VMWare HA that the majority of potential outages are accounted for.
Consider the following:
- Most HA options will require that the host(s) is/are functioning correctly. If you have problems with the host, expect problems with your HA solution.
- Replication of virtual hard disks will not protect you from data loss or corruption inside VM. The loss will be replicated.
- Replication of virtual hard disks will not protect against corruption of virtual hard disks or settings. The corruption will be replicated.
- Live migration options really only work for planned downtime. Unplanned downtime will result in your VM's being restarted with a (varying) loss of service.
With the above in mind, have a read of the following:
- Hyper-V Hardware considerations
- Backing Up Hyper-V Virtual Machines
- Hyper-V Clustering Options
- Design for a Failover Cluster in Which All Nodes Run Hyper-V
- Requirements and Recommendations for Failover Clusters in Which All Nodes Run Hyper-V
- Checklist: Failover Cluster in Which the Servers Run Hyper-V
Long story short, virtualisation has not changed anything when it comes to mitigation against system failure/outage. The same rules still apply. Virtualisation high availability solutions represent only the first layer of protection .. just don't forget the other options like:
- Traditional Backup and Recovery
- Host Clustering
- Network Load Balancing