our company is trying to building a new Elasticsearch for service and We are studying ES
for a proper service .
And I have a question about replica shard. Should we set our replica shard more than 2?
Here is reason for asking this question
if there is too much shard, indexing and updating performance go bad.
many case in distributed system, making 2 replica (slave) is recommended for recovery.
here is our dillema, I think if we want to get safety, we should set our replica as 2, but if we do so, our indexing rate would go bad ( our index is almost 2000 )
But im not sure that we should set our replica as 2 (maybe 1 replica casue no problem )
The more replicas you have the more work is required in the cluster for indexing and updates, so increasing the number of replicas indeed tend to reduce indexing/update throughput.
This depends on the size of your cluster, how many availability zones you have and what level of resiliency you require. If you only need to be able to handle one node going down having a primary ad a single replica is often sufficient as Elasticsearch will reallocate a replica to another node if one is lost.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.