Elasticsearch master node, data node, client node differences and their respective characteristics

What is the meaning of master node, data node, and client node in the ES cluster, and what are the characteristics of each?

Master Node

The main function is to maintain the metadata, manage the state of each node of the cluster, data import and query will not go to the master node, so the master node pressure is relatively small, so the master node memory allocation can be relatively less; but the master node is the most important, If the master node hangs or has a brain fissure, your metadata will be confused, so that all the data in your cluster may be lost, so be sure to keep the master node stable.

Data node

is responsible for the data query and import, it will be more pressure, it needs to allocate more memory, select the server when the best choice to configure the higher machine (large memory, dual CPU,SSD ... Local tyrants ~); If data node is broken, it may lose a small piece of data.

Client node

The

is used as a task distribution, and it also contains metadata, but it does not make any changes to the metadata. The advantage of Client node is that it can share some of the pressure of Data node, why Client node can share part of the pressure of Data node? Because the query of ES is the result of two layers aggregation, the first layer is to make the query result aggregation on Data node, then send the result to Client node,client node receive data  Node sends the result after the second aggregation, and then the final query results back to the user; so we see, Client node help to the second layer of aggregation work, naturally share the pressure of Data node.
Here, we can give an example, when you have a Big Data Query task (such as hundreds of millions of query task volume) dropped to the ES cluster, if there is no client node, then the pressure is directly lost to Data node, if data  Node machine configuration is not enough to accept such a large query, then it is likely to hang up, once hung, Data node will be re-recover, re-reblance, this is an abnormal recovery process, the result of this process is to cause ES Cluster service stop ...   But if you have Client node, the task will first be lost to Client node,client node if not, the most is Client node stop, will not affect the data The  node,es cluster also does not take an abnormal recovery.

for the es  cluster to design the nodes of these three roles, but also from the hierarchical logic to consider, only the relevant functions and roles are clearly divided, each node to do their own responsibility, in order to play a distributed cluster effect.

                                &N Bsp                  

For more elasticsearch knowledge, see the following figure:

650) this.width=650; "Src=" Http://s2.51cto.com/wyfs02/M00/85/DC/wKiom1esi1eB4ojLAAihBMsKgF0640.jpg-wh_500x0-wm_3 -wmp_4-s_3407292064.jpg "title=" es javanew.jpg "alt=" Wkiom1esi1eb4ojlaaihbmskgf0640.jpg-wh_50 "/>

Elasticsearch master node, data node, client node differences and their respective characteristics