Configure Kibana for multiple ES servers/nodes (in the same cluster)

[Ellesedil]

In our QA environment, we have two servers running Elasticsearch in the same cluster, which effectively means we have two nodes. Right now, we have Kibana set up on only one of the servers, and it's currently configured to use that sever's ES instance for it's connection to our data. When ES needs to be cycled or otherwise brought down on the box, Kibana breaks because it cannot connect.

After about 30 minutes of searching, this is the closest thing I could find my issue: #22. But the distinction is I want Kibana to still pay attention only to a single cluster, but have data served to it from multiple nodes.

Is it possible to configure Kibana to set multiple ES servers/nodes for it's connection to the ES cluster? If so, could the documentation be updated somewhere to reflect how to do this? Thanks.

[rashidkpc]

Simply setup an elasticsearch client-only on the Kibana host (see here: http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-node.html)

Then point kibana at localhost, the ES client node will do the work of making sure it knows which nodes are live and which aren't.

[Ellesedil]

That solves part of the issue, but Kibana still has a single node as a point of failure. If the client node has to be brought down for whatever reason, then Kibana has nothing to connect to. Configuring multiple client nodes for redundancy is a moot point. So, two things:

1. Is there any reference to this in the Kibana documentation so people can find how to do this in the future?
2. Is this the best that can be hoped for? At this point, it sounds like there should be a feature request to allow multiple nodes to be configured in Kibana, and it will cycle through them in order when there's something to do until it finds an active node or has checked them all.

[rashidkpc]

Configure the client node on localhost, have kibana talk to localhost. If that host goes down Kibana would be down anyway. Alternatively use something like HAProxy, or a dedicated load balancer in front of elasticsearch. We do not plan to implement cycling through nodes in the near future, elasticsearch client nodes accomplish the goal and do a great job of it.

[ccrivelli]

I have two questions about this topic.

1. If the elasticsearch cluster (server nodes) is already configured with shield and ssl certificates, should shield be installed and configured (with all the certificates in the keystore) also in the elasticsearch client node (where kibana is running)? In other words, should an elasticsearch client node be 'secured' like the server nodes (creating a certificate request, signing the the certificate request by the root ca, then installing the signed cert) to become a member of the existing cluster and to be able to connect to the server nodes?

2. The same elasticsearch cluster (server nodes) is configured to authenticate against an external Active Directory over ldaps. This has NOT to be implemented/configured on the elasticsearch client node. Correct?

Thanks.

[chamarts]

@ccrivelli old thread. but just wondering if you figure out the answer

[chilcano]

The same here!
@ccrivelli +1

[ankitjhil]

Same here!
@ccrivelli +1

[eamocanu]

+1

[Dom-nik]

@ccrivelli +1

@rashidkpc
Your answer is useful, but this is not an available option for everyone, it seems. For the users who are considering X-Pack it means paying extra for another ES license (for the client/load balancer node) to get HA in Kibana.

[mchenx]

is this what you expected? https://discuss.elastic.co/t/configure-kibana-for-multiple-es-servers-nodes/2431/3