Configuration of kibana with cross cluster search

I am switching my multi-cluster installation from using kibana running on a tribe node to using kibana with cross cluster search running on a single standalone elasticsearch node. From the documentation it appears to be recommended to use gateway nodes but there is no clear definition of what a gateway node is.

Old setup
DC1 cluster - 3 data nodes, 3 master nodes, 1 tribe/kibana node
DC2 cluster - 3 data nodes, 3 master nodes
DC3 cluster - 2 data nodes, 3 master nodes

New setup
DC1 cluster - 3 data nodes, 3 master nodes
DC2 cluster - 3 data nodes, 3 master nodes
DC3 cluster - 2 data nodes, 3 master nodes
Kibana - 1 data/master/kibana node

Question 1. Is a gateway node simply a client node ( i.e. node.master:false node.data:false node:ingest:false )

I have currently set up my cross cluster search seeds to be the 3 master nodes in each cluster

Question 2. Should I instead add at least one gateway node to each cluster and configure these as my seeds ?

It's the node(s) that the cross cluster configuration uses to talk to. It explains it here - Cross-cluster search | Elasticsearch Reference [6.1] | Elastic

Nope, see above.

Thanks Mark

Would you advise me to set the seed nodes in the cross-cluster configuration as the master nodes (as it is now) or the data nodes ? or all nodes ?

I'd use the data nodes, because they are the ones best used for queries (unless you had coordinating nodes) :slight_smile:

That was what I meant in my "question 2" above but I guess I have the wrong terminology.

So I should either use the existing data nodes in each cluster as seeds or add at least one coordinating node to each cluster and use these as the seeds

You've been a great help, thanks

Yep! :smiley:

Using Kibana in a Production Environment | Kibana User Guide [6.1] | Elastic

the ideas is to run coordinate only elasticsearch node and kibana would access your cluster through that node.

It seems that if I run Kibana on a coordinating node in one cluster that I am then unable to search across both the local indices in that cluster AND remote nodes

This would suggest that the only way to search across multiple clusters would be to run kibana on a completely separate elasticsearch node so that I can use a wildcard in my index pattern to read data from all clusters.

( I have to point out though that I am using version 5.6.2 as there are limitations in my configuration that may prevent me from moving to 6.x yet )

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.