Using Unicast discovery with logstash & elasticsearch

I want to use unicast discovery in one of my logstash deployments as mentioned here in the best practices. I have gone through this link.

My ES Cluster looks like below:-

  • 3 Master-Data Nodes - es-masterdata1, es-masterdata2, es-masterdata3.
  • 1 Dedicated Client Node for now - es-client1

I have added unicast host names in elasticsearch.yaml (as below) in all 3 Master data nodes:-

discovery.zen.ping.unicast.hosts: ["es-masterdata1:9300", "es-masterdata2:9300", "es-masterdata3:9300"]

I am having the following doubts:-

Doubt 1
Do I have to include es-client1 in the list of unicast hosts?

discovery.zen.ping.unicast.hosts: ["es-masterdata1:9300", "es-masterdata2:9300", "es-masterdata3:9300", "es-client1:9300"]

Doubt 2
Do I have to include discovery.zen.ping.unicast.hosts in the elasticsearch.yaml of the es-client1 as well?

Doubt3
How the logstash config should look like? Do I have to include es-client1 as the host name or I have to give the IP of one of the master / data nodes? Which is the best practice? Is the below logstash output config fine?

output {
  elasticsearch {
    cluster => "elasticsearch"
    host => "es-client1"
    port => "9300"
    protocol => "transport"
  }
}

As per my understanding after reading the docs the answer to Doubt 1 & Doubt 2 is No. I wanted to confirm that.

But I could find any proper guideline regarding my 3rd doubt.

Do I have to include es-client1 in the list of unicast hosts?

No, not if es-client1 is able to connect to the cluster.

Do I have to include discovery.zen.ping.unicast.hosts in the elasticsearch.yaml of the es-client1 as well?

Yes. How would it otherwise find the rest of the cluster members with multicast disabled?

Thanks @magnusbaeck.