We use Elasticsearch as log management solution. In daily operation, we often encounter this problem: "mapping update timeout occurred when logs of shard were recovering, so index speed was reduced ". When this happened, we have to set
cluster.routing.allocation.node_concurrent_recoveries to 0, then mapping update succeed.
I am wandering "what is the relationship between mapping update and shard recovering" ?
Both of these operations involve the master node. From your description it sounds like your master node is overloaded. What does
GET _cluster/health report?
The cluster is in green or yellow status. Next time I will list the _cluster/health output.
why the master node is involved when shard recovering ? The global cluster status is updated during recovering?
Yes, the cluster state is adjusted in order to start the recovery, and then again to indicate that recovery has completed and the shard is ready to serve searches.
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.