Objects Storage Scaleout : Multicluster

Anirudha | Wed, 08/19/2020 - 09:18

You may or may not always have additional nodes to expand the underlying AOS cluster. But you may have other Nutanix clusters which are under utilized (w.r.t Storage). And you want to use that storage for Objects workload.

Objects 2.0 was released with one of the amazing features called “Multi-cluster”, which helps you to scale out your Objects cluster’s storage layer beyond AOS cluster boundaries. 

Now you stretch your Objects cluster across multiple AOS clusters. And not only this, it's also platform independent, which means you can add AHV or ESXi clusters , and Objects will consume storage irrespective of underlying hypervisor type.

Lets go to Objects UI -> Clusters Tab -> “Add Cluster” -> Select one of the available cluster from list -> Save

Here I am selecting the ESXi cluster and going with the default 25% usage limit.  You can also input the custom value from 1%-100% or choose any predefined percent, depending on how much storage you want to allocate to Objects cluster.

This takes about a minute or two to add the cluster, and Objects can now use storage from both the clusters based on the limit configured.

Objects is now reporting 36.67TiB storage available.

You can add up to 4 clusters. And Objects will balance all the workload equally across all the AOS clusters, to avoid any potential skew.

If you want to limit the storage consumption on any of the cluster while adding or post addition, then the same UI also gives you that option. Refer to the “Limiting or Updating Storage” post to understand this in more detail.

Objects will use all the newly added clusters for new writes .

Things to Note here:

  • Once the underlying AOS cluster reaches the storage capacity Or reaches the limit specified in Objects cluster, then writes may start failing but read will continue. 
  • You can add up to 4 AOS clusters. Cluster can be AHV or ESXi, and based on available storage or defined limits, Objects will start serving writes.
  • Objects will spread new writes across all the AOS clusters in round robin , so each cluster is used equally.
  • Hyper-V is not supported.