Tag Archives: quorum

Sharded MongoDB config on Nutanix (2) : High Availability

One of the prime availability considerations for any horizontal scale out application, like a MongoDB cluster, is how that cluster behaves under a failure event. We have seen (in the MongoDB case) how replica sets are configured with additional secondary instances to handle the failure of a primary instance in a replica set. We also create a mini quorum of configuration database servers and query routers to give redundancy to the cluster “infrastructure”. However, the Nutanix XCP environment provides further protection through certain features of the Acropolis management interface. Your key VMs need to be enabled to run under high availability. This is so that when the underlying hypervisor host fails for any reason, these VM’s failover to another host with sufficient CPU and RAM resources. The screenshot below shows how this (Tech Preview) feature can be enabled (pre NOS 4.5) on a per-VM basis :

enable-HA

The underlying migration functionality is also used for the manual placement of key VMs. As an example, let’s consider the following layout, where two of the configdb VMs in a MongoDB cluster are co-located on the same AHV host:

mongodb-colocated-vms

Notice in the screen capture above, there are two configdb VMs on host “D”. This means that ideally we want to migrate a MongoDB Config DB to another AHV host. Let’s move the VM mongo-configdb02 to AHV host “C”…

mongodb-migrate-VM

Note that the migration process could have automatically chosen an appropriate AHV host to receive the VM. In the above case however, we have instead specified the desired host ourselves.

We can monitor the progress and duration of any migration via the VM tasks frame in Prism:

mongodb-vm-tasks-migration

As always, this workflow can also be done manually (or scripted) through the acli interface. In this example I am migrating the VM running a query router (mongos process)….

<acropolis> vm.migrate mongos01 host=10.68.64.41 
mongos01: complete 
<acropolis>

As of the time of writing this post. Acropolis Base Software (NOS) 4.5 has been released and this feature has become part of general availability (GA). It can now be enabled cluster wide:

ha-enable-menu-4

enable-ha-4

Nutanix customers are strongly recommended to enable this feature when they require HA functionality for their VMs.

In my next post I will be completing this short blog series on sharded MongoDB configs on Nutanix. I intend to cover how Nutanix Acropolis managed snapshots and cloning are employed to create backups and then use them to perform rapid build out of potential dev/QA type environments. Stay tuned.