Elasticsearch rivers have been deprecated, which makes one think about alternatives. The recommended method is using Logstash input plugins, but that feels a bit too heavy for my needs.
There's a node.js based solution which allows to perform one-off imports from CouchDB and continuous sync as well. The little tool is available as couch2elastic4sync and in its minimal usage only needs the source database url and the target index.
If that's enough for your needs, just head over to it. If you want to run it in a Docker container,
feel free to use this repo, or directly run the image gesellix/river. A mapper has been added
to make the load and sync actions ignore CouchDB design documents. See below for other options.
docker run -e couch2elastic4sync_database=http://couchdb:5984/db-name -e couch2elastic4sync_elasticsearch=http://elastic:9200/index/type gesellix/river
docker run -e couch2elastic4sync_database=http://couchdb:5984/db-name -e couch2elastic4sync_elasticsearch=http://elastic:9200/index/type -e couch2elastic4sync_end_on_catchup=false gesellix/river
docker run -e couch2elastic4sync_database=http://couchdb:5984/db-name -e couch2elastic4sync_elasticsearch=http://elastic:9200/index/type gesellix/river load
Caution: the last example also indexes the design documents. To make them being ignored, just add the provided mapper like this:
docker run -e couch2elastic4sync_database=http://couchdb:5984/db-name -e couch2elastic4sync_elasticsearch=http://elastic:9200/index/type -e couch2elastic4sync_mapper=drop_design_docs gesellix/river load