Elasticsearch in microservices architecture

I'm designing a system in which we have two different microservices to manage two different data types. Common approach is to have database per microservice, but what are your thoughts about if Elasticsearch takes place of a database? Should I have Elasticsearch instance (cluster) per microservice to claim my architecture nice and shiny, or one instance (cluster) with many indexes is ok? I expect answers like "it depends", but still I would love to know your thoughts about this.