Hmm my response (via email) was truncated for some reason ... trying again:
I think (not certain) for logstash it's the number of workers you specify for the Elasticsearch output? I think the default is 1: https://www.elastic.co/guide/en/logstash/current/plugins-outputs-elasticsearch.html
Or in your custom Go client, you would control how many threads (goroutines?) are sending bulk indexing requests concurrently.
You can ask for node stats, then look under thread_pool -> bulk -> active to see how many bulk threads are currently working, but this isn't a precise measure since when you check, it could be your client side threads had just received results and are working to prepare the next request and so they are not counted.
You want to have enough client side concurrency so the server side resources are saturated (either CPU or IO or, if you're lucky, both!).