Does anybody know what the replica concurrent setting is?
I just set the number_of_replicas from 0 to 1 in a production environment with the below settings:
cluster.routing.allocation.cluster_concurrent_rebalance": "10"
cluster.routing.allocation.node_concurrent_recoveries": "2"
I expected there to be a max 10 replica shards allocation, but instead, there were almost 100 shard allocation tasks triggered which caused severe
impact on the online search, and I want to know the exact settings for replica allocation.
Increasing this setting may cause the cluster to use additional resources moving shards between nodes, so we generally do not recommend adjusting this setting from its default of 2 .
I'm not sure, but cluster_concurrent_rebalance only limits the number of rebalancing shard movements in the cluster, there are lots of other reasons for shard allocations. Could you share the output from GET _recovery from the time of the problem?
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.