You should have a separate container for Kibana. You only need one Kibana for one cluster. So you don't need to include Kibana on any data/master node.
You need to have all of the plugins installed on every node.
You can base your own dockerfiles on these images:
Would it be ok to install kibana on one of the data nodes? This is because, if we chose elasticsearch support, we will save a bit of $$$ (elasticsearch seems to charge based on number of nodes). What are pros and cons of installing kibana on data node?
If the node hosting kibana goes down, wouldn't that mean single point of failure for kibana? Is there a standard practice or suggestion that would help us avoid this kind of scenario?
If you install the client node on the same server as Kibana. Then you can setup multiple Kibana servers, then use a load balancer in front of Kibana. If any data node goes down, then the client node gets cluster state updates and knows which nodes are up and where the shards are located.
Are you sure about the client node does not count? Let me know if there is any document that describes that. We are planning to buy support subscription and cost is fairly important. Sales representative informed us that the cost is based on the number of nodes (I will clarify that from my end).
Also, if we have multiple clusters, is there an easy way to monitor these clusters from one kibana dashboard?
Yes I am sure. I suggest you discuss this further with the sales rep. X-Pack Monitoring supports monitoring multiple clusters. I suggest you setup a separate single node cluster for monitoring. And then set the exporter for the production clusters to send the data to the single node monitoring cluster:
However if you're looking to have one Kibana be able to create Visualizations, Dashboards, etc using Index Patterns across multiple clusters, then Kibana 5.0 doesn't currently support this as this was previously accomplished with the tribe node:
In an upcoming version of Elasticsearch & Kibana will address this functionality.
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.