We typically provision our elasticsearch clusters with a lot of indices, and one replica. The search traffic can vary based on the index.
Sometimes we get a burst of search traffic on one index (usually a node gets CPU bound and the search queue is backed up) and so we resolve it by scaling up the number of replicas.
Does elasticsearch have scaling signals for this type of scenario that could enable us to increase the number of replicas automatically?
Do you add more nodes or just increase the number of replicas?
Typically there are enough nodes in the cluster already, so we just increase the number of replicas as a mechanism to add hardware to the search traffic (e.g., with 1 primary, 1 replica, existing shards for the index under load are only on 2 nodes)
Yeah, if it was by adding new nodes, you could use the auto_expand_replicas settings, but to just upgrade the number of replicas there is nothing to automate it.
Would you be aware of any cluster-side signals that would indicate that there are not enough resources on an index to keep up with search traffic?
© 2020. All Rights Reserved - Elasticsearch
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.