I am currently running a small cluster of 3 nodes with a size (shards considered) of =~ 200GB.
But disk usage keeps increasing and hard drives are about to reach saturation.
I know that Elasticsearch need to be as close as possible to the hardware to deliver its full potential. And the best option should be to migrate each node to a new instance type with a bigger hard drive.
But this option is more expensive and time consuming. Right now I would prefer the option of adding an external disk storage.
Will the cluster really lose a lot of performance?
What would be the requirements, in terms of IOPS / bandwith, to preserve as much as possible the performance?