Hi team,
We have two data center and hosted three Elasticsearch (master + data) node in each of these two data centers (2 node + 1 node). The intension is to achieve high availability during diester recovery. All these three ES nodes are clustered.
The issue we are facing is when two of our ES nodes goes down (one of the data centers), the remaining one ES node (in other data center) do not survive with ERROR: Master node not found which is expected. How to deal with this kind of issue?
FYI - We also tried setting up 4 ES nodes, two nodes in each of the data center but this set up also not helping.
Regards | Bibhu
That is expected and explained here. Elasticsearch always require a strict majority (2-out-of-3, 3-out-of-4) in order to be fully operational. There is therefore no way around this apart from ensuring you do not lose a majority of nodes.
It is not recommended to deploy Elasticsearch across data centers unless network throughput is good and latency is very low.
If you want to be able to handle a full data center going down you will need to deploy across 3 data centers in order to assure a majority of nodes is always available. Note that the third data center only need to hold a small dedicated master node (possibly voting only) that can break any tie.
1 Like
@Christian_Dahlqvist, Thank you
Appreciate your quick response.