Understanding #vROps High Availability
While there is a reasonable amount of information about how HA works in vROps I have found there is still some confusion as to how HA actually works with vROps or rather what are the benefits and perhaps more importantly the cost for enabling the feature.
HA is a great feature and in my opinion should be considered seriously with any deployment (where possible).
Not only does HA protect your Master node (which essentially behaves as an index for your vROps cluster and if lost will render your cluster dead unless you have a working backup of your cluster) it will also allow your cluster to tolerate a data node failure. So in short what is not to like!
The important thing to remember here is the cost in terms of object/metric capacity for your vROps cluster and in turn the additional nodes and or hardware you may need to facilitate/provide for having a HA enabled vROps Cluster.
If you have deployed vROps before then I would hope you are already familiar with the VMware provided vROps Sizing Spreadsheets When you input your environment figures you will be informed of the size of nodes and number of nodes required to monitor your environment. Furthermore you can select how many months you plan to retain data and whether you will enable HA or not.
When you Enable HA in the spreadsheet you will see the number of required nodes will double. This is because the nodes are for want of a better expression split in half, therefore halving their capacity. This is to “mirror” the data and create “shards”. Please reference the diagram below.
So if your existing cluster node had a capacity of 10000 objects when you enable HA it’s capacity is effectively halved to 5000. The other half being set aside for sharded/mirrored data.
Now that you understand what HA effectively does in terms of capacity you will now appreciate that there may well be an impact for you in terms of the number of required nodes to monitor your environment.
If you needed a 3 node cluster before you enabled HA you will need a 6 node cluster afterwards. This of course then introduces the dilemma of whether to enable or not especially if your management cluster is “tight” for resources.
Ultimately the choice and design decisions are yours to suit the best needs for your environment.
If you are looking for more information on how to architect your vROps clusters why not check out part 5 of a webinar series I Co-host with Sunny Dua – Design and Deployment considerations.