Understanding Liquid Web’s High Availability cPanel/KVM Servers

Why didn’t the HA-cPanel/KVM kick-in when Apache was down for 20 minutes?

Apache operates on the service level of the HA-cPanel/KVM infrastructure. Service levels issues are not monitored. This is because they are contained within the virtual machine running on within the virtualization layer. Quite simply, the underlaying parent servers have no idea what is running on the guest VM they are protecting. Consider the following simplified chart of a typical HA-cPanel/KVM setup:

As we can see the LAMP Stack where Apache resided is further up past the Fail-safe level. There guest server runs in its own virtual environment. It’s not even aware that it is a guest in the environment and behaves like a standard Virtual Machine.

The HA-cPanel/KVM hardware layer protection is constantly at work. Each parent is perpetually talking to each other over a dedicated cross-over cable using a heartbeat daemon.  The moment the PRIMARY server fails to identify itself to the SECONDARY server, the SECONDARY server springs to life, promoting itself to PRIMARY and spinning up the virtualization platform to launch the guest-vm.

Fail-over is not completely seamless in this configuration. The guest-vm must be booted up on the new PRIMARY parent. However, due to the constant synchronization of data using DRDB  there should be minimal lost data.

Why not monitor the service level?

That’s a tricky questions. Fail-over in this system is designed to protect against imminent hardware failure and not a system built to maintain service uptime. Service level issues are generally resolved without needing to reboot a server. Since fail-over is effectively an abrupt power off of the guest-vm, it’s not a clean shutdown and there is always a minimal risk of some data loss or corruption.

There are better suited configuration for handling service level downtime. A highly trafficked site can take advantage of Load Balancing techniques to ensure responsive and stable uptime of their Apache LAMP stack. HA-cPanel/KVM can also work as a single node in an LB setup offering both the service level stability of a Load Balancer as well as the hardware level protection of HA-cPanel/KVM. You can have the best of both worlds.

If you are looking for a reliable system that protects against hardware level damage, give the folks at Liquid Web a few moments of your time to consider a solution like this for your valued data. I promise they won’t bit and the support is truly heroic.