We need to provide an index template that Elasticsearch will use on information from Couchbase. The one we use below is an example provided as part of the Couchbase Plug-in for Elasticsearch:

shell>	curl -XPUT http://localhost:9200/_template/couchbase \
-d @plugins/transport-couchbase/couchbase_template.json

You can also provide your own index template in the future. If you provide more than one template, you can maintain multiple indexes that can be individually updated. Upon success, Elasticsearch returns:

{"ok":true,"acknowledged":true}

For each Couchbase data bucket that we want to search, we create an empty index in Elasticsearch:

shell>	curl -XPUT http://localhost:9200/beer-sample

In this case we name our index beer-sample. Upon success Elasticsearch returns the following:

{"ok":true,"acknowledged":true}

Provide this setting to Elasticsearch to change the number of concurrent requests it will handle:

shell>echo "couchbase.maxConcurrentRequests: 1024" >> config/elasticsearch.yml

Stop and restart Elasticsearch for your changes to take effect.

Change the number of concurrent replicators in Couchbase Server from 32 to 8:

shell> curl -X POST -u Administrator:password1  \
http://10.4.2.4:8091/internalSettings  \
-d xdcrMaxConcurrentReps=8

When Couchbase Server successfully updates this setting, it will send a response as follows:

HTTP/1.1 200 OK
Server: Couchbase Server 2.0.0-1941-rel-community
Pragma: no-cache
Date: Wed, 28 Nov 2012 18:20:22 GMT
Content-Type: application/json
Content-Length: 188
Cache-Control: no-cache

One of the issues that can occur when an Elasticsearch node is overwhelmed by replication from Couchbase is that the node can fail. If this does occur, you may also experience errors from remaining nodes.

For more information about this XDCR parameter, see Couchbase Server Manual, Changing Internal XDCR Settings.