Is there any benefit to enabling Elasticsearch master nodes if they will run on the same hosts as a data nodes?
My company ships a product with a hosted (AWS-based) and on-premises offering. For hosted, we use AWS Elasticsearch. For on-premises, we bundle Elasticsearch binaries.
We noticed that AWS Elasticsearch recommends enabling Dedicated Master Nodes and running 3 such nodes on compute-focused EC2 instances (that do not run ES data nodes). We follow this guideline for our hosted product. However, it is generally infeasible to do for our on-premises product, because we cannot expect our customers to allocate those dedicated ES nodes. For example, if they are only running a 3-node cluster for our product, we cannot exactly ask them to turn it into a 6-node.
In addition to the main question of this post, what are general guidelines for this scenario?