Support for connections to multiple servers in cluster

[dmkoch]

Both the perl API: http://search.cpan.org/~drtech/ElasticSearch-0.37/lib/ElasticSearch.pm#new() and the Ruby rubberband API: https://github.com/grantr/rubberband/blob/master/lib/elasticsearch/client/retrying_client.rb support connections to multiple elasticsearch nodes in a cluster, including round-robin load balancing, auto retrying, and auto discovery of nodes after initial connection. Would you considering adding similar features to Elastica? It would eliminate a single point of failure in a clustered environment in the event the node that the Elastica client is hitting goes down.

Thank you!

[ruflin]

I think this is a feature that is really needed. As far as I remember, someone started to work on that some time ago, or at least we were discussing it. But I haven't received any pull request for this yet.

Probably this refactoring could be done in combination with the refactoring for adding different transportations (see issue #33) as it needs some additions to the client.

[ruflin]

I added support for multiple servers in my last commit (see link above). Currently just a round robin is made. You can use it by just passing an array of servers with hostname and port to the client:

$client = new Elastica_Client(array('servers' => array(array('host' => 'localhost', 'port' => 9200))));

[dmkoch]

Thanks so much for a quick turnaround!

[ruflin]

Can you do some testing on it? Probably there are still some issues I haven't thought of yet.

[ruflin]

There is now also a blog entry about it. I hope to improve it in the near future. I will close this issue now.

http://ruflin.com/en/elastica/165-using-elastica-with-multiple-elasticsearch-nodes