Elasticsearch clusters doesn't provide virtual ip support, instead data can be pushed and accessed to/from every node of the cluster (master or not).

In case of failure, zabbix should be able to switch from one ES node to a other (that's what logstash and many others do).
That's why the possiblity to set multiple URLs in "$HISTORY['url'] (frontend) and HistoryStorageURL (server) would be nice.

Otherwise, to ensure full availability, a reverse proxy (like nginx or F5, etc) would have to be put between zabbix and ES cluster.