I'm preparing the production platform and I have some mistakes. Now I have available 3 servers working with 3 ES nodes, but I need a coordinating node due a load balancing. I prefer coordinating node than DNS round robin and I can't do in client side. Also I have checked HAproxy and I prefer the ES solution.
Can one server act as data node and coordinating node at the same server? is it fine?
I know the best scenario is start with 3 nodes as follows, this is the config I have now and coordination node in another but I haven't more hardware inmediatly. My app is going to Node2 but with no balancing now, if node2 fails the app doesn't work. This is the actual config:
Yes. I know the nodes doesn't do load balancing, is for this reason I'm thinking to put a ES node acting as coordinating mode (without data) to do load balancing.
Also I have checked to do with HAproxy or in our Citrix Netscaler but I prefere to avoid these solutions to not load other systems with additional work.
No. We are not using your library. We checked and for our app we prefer to create our own custom calls as I sent you. Now the app is working well and I need to go on production but I want high availability and load balancing in our nodes.
anything helping me about roadmap to follow? I need load balancing and high availability...
If you need high availability, you should either connect to more than one node and distribute load like our client libraries do or use a load balancer. Having a single coordinating node will make this a single point of failure.
it's true... I have an alert in our application who tell me if the coordinating mode fails then the app change automatically the IP so the app point to another node...
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.