[ https://issues.jboss.org/browse/RTGOV-462?page=com.atlassian.jira.plugin.... ] ivan mckinley commented on RTGOV-462: ------------------------------------- I see cases that have to be addressed. Its a trickier topic than simply starting a ES node First the easy one. 1) Test and Integration : A lightweight local jvm ES is started using the local node feature from the ES java api. This would require the entire ES dependency stack for our module. As opposed to currently where we only have the client api 2) Embedded out of the box. - Should start a single ES node - Should be singleton in the JVM. I have not reviewed the ElasticsearchClient but we need to ensure that only one ES instance is started per vm. Is ElasticsearchClient "singleton" or is it shared amongst Keyvalue stores. Starting additional embedded nodes in 1 jvm should not be allowed (hence singleton or static as you mentioned ) - Should be capable of joining a ES cluster(or any). This would allow us to scale horizontally by adding additional FSW or ES nodes as data/load grows. - Allow unique configuration per instance. Once again would support scalability as perhaps certain jboss nodes with embedded ES would be designated as been responsible for a particular model. for example in multiple node we store all situations objects on dedicated nodes(sharding etc). -- This message was sent by Atlassian JIRA (v6.2.3#6260)