No Route to Host - Graylog not Starting

After restart graylog nodes, they cannot communicate with the elasticsearch database. Graylog version is 2.2.3, ElasticSearch is 2.4.4. Graylog logs report the following:

2017-04-19T09:24:30.906-04:00 INFO  [zen] [graylog-c27e0f18-5abd-453d-93f3-57be2ffa5b11] failed to send join request to master [{001-elasrchdb-02.example.com}{iT4HNsk3Rza4ifz9CfHxIg}{172.16.1.154}{172.16.1.154:9300}{attr.zone=prod, attr.rack_id=even}], reason [RemoteTransportException[[001-elasrchdb-02.example.com][172.16.1.154:9300][internal:discovery/zen/join]]; nested: ConnectTransportException[[graylog-c27e0f18-5abd-453d-93f3-57be2ffa5b11][172.16.1.61:9350] connect_timeout[30s]]; nested: NotSerializableExceptionWrapper[no_route_to_host_exception: No route to host]; ]

If I do a netstat on the elasticsearch nodes, I see connections and doing a tcpdump -A reports similar to the following:

."=..$...client.true.data.false.master.false..|...........001,graylog-c27e0f18-5abd-453d-93f3-57be2ffa5b11.bmujzCKZSNS9rarcTWFGeA.172.16.1.61.172.16.1.61...

so it is making to the host.

All Elasticsearch nodes must be able to connect back to the embedded ES node running inside Graylog.

In your case, the Elasticsearch node must be able to connect to 172.16.1.61:9350.