Hands-On Software Engineering with Golang

Scaling across an Elasticsearch indexer implementation

A caveat of the in-memory bleve-based indexer implementation is that we are more or less limited to running our index on a single node. This not only introduces a single point of failure to our overall system design but it also places a hard limit on the amount of search traffic that our service can handle.

We could definitely argue that we could try to scale our implementation horizontally. At the time of writing, bleve does not provide any built-in mechanism for running in distributed mode; we would need to roll out a custom solution from scratch. One approach would be to create a multi-master setup. The idea here would be to spin up multiple instances of our index service and place ...