our plan is to store logging information in three datacenters. The data should be high available - also in case of a complete outage of one datacenter.
A tribe node is not an option as in case of a datacenter outage parts of the data won't be available.
Our current architechture is: Logstash (Shipper) -> Redis (Broker - no cluster) -> Logstash (Indexer) -> ES.
Our idea is to copy the data to all datacenters on Broker level. We think about using the cluster functionality of Redis. Maybe Kafka or RabbitMQ are also suitable instead of Redis.
Does anybody have experience with such an issue?