Elasticsearch disaster recovery setup

[flalar]

We’re running our ES cluster on a local VMware setup, which works fine. Now that this solution is becoming increasingly business critical. We’re looking to also include this in our disaster recovery as a service setup. The DR setup basically replicates our most critical VM’s to a service provider and lets us start them up on a secondary site.

Given this setup it would be great if we could include a subset of our ES cluster, eg one node that had a copy of all the data, for replication to the secondary site.

Any inputs on how to achieve this? Have been looking into shard allocation awareness, but not sure that is the most applicable solution to the problem.

[spinscale (Alexander Reelsen)]

First, cloning the VMs is no guarantee that everything works properly when starting up. Second, running the same cluster across different data centers is not recommended. You could use forced awareness, if running in two racks or two availability zones in a single data center, see https://www.elastic.co/guide/en/elasticsearch/reference/6.4/allocation-awareness.html#forced-awareness

In the future there will be a feature called Cross Cluster Replication, that will allow to do this (but it is going to be a commercial feature).

You could try to use snapshot/restore if you are willing to have some additional waiting time and non real-time updates, or index into two clusters. None of those solutions are ideal, thus the current work on CCR.