How do I select the right sized hosts for deployment?

I've been using AWS's hosted version of Elasticsearch for a bit now and they have a max size per node. I'm in the process of moving off of AWS and into running it myself.

I'm looking for suggestions on how to pick the right size master and data nodes as well as how much space should I allocate to the data nodes? Is there an amount that is too much? Amazon allows up to 1.5TB per data node, how much further can you go beyond that? 5TB? 10TB?

Also what size nodes are recommended per master, data, etc?

The answer tends to be "it depends," but to figure out the answer that's right for you, you should watch https://www.elastic.co/elasticon/conf/2016/sf/quantitative-cluster-sizing. There's no hard limit of "too much data per node," but once you do start getting into very large sizes, you do have to consider "what would happen if the node was lost and I had to replace it?" If you have 10TB on a single node, the answer is that Elasticsearch would need to recover 10TB onto the new replacement, which would involve copying that much data over the network, which could take a very long time depending on your network and disk speeds.

1 Like

Thanks for the info, I will watch that video and adjust accordingly.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.