Service cluster

Abstract

The invention provides a service cluster. When a plurality of master name nodes exist, standby name nodes can learn whether faults occur in the master name nodes or not by sending test information to the master name nodes and send master selection applications to the target master name nodes when the faults occur, wherein the master selection applications are used for applying to become the master name nodes. The target master name nodes can add the standby name nodes to master name node lists. When only one master name node exists, the standby name node can learn whether the fault occurs in the master name node or not by sending the test information to the master name node and add the standby name nodes to the master name node list master name nodes when the fault occurs. According to the service cluster, the fault information of the name nodes can be automatically discovered; when the fault occurs to certain master name node, the standby name node is added to the master name node list as the master name node; the dependence on a Zookeeper cluster is removed; the system architecture is simplified; and operation and maintenance costs are greatly reduced.

Description

technical field [0001] The invention relates to the technical field of big data, in particular to a service cluster. Background technique [0002] Nowadays, Hadoop HDFS has become the most mainstream distributed file storage system in the field of big data, and the name node (NameNode) is the most important management node in HDFS. Its role is mainly reflected in the following aspects: responsible for maintaining the entire file The file directory tree of the system; storing metadata information of the file system; responsible for receiving user operation requests. It can be seen that the stable operation of the name node is very important for HDFS. [0003] In Hadoop 1.X, HDFS has the single point of failure (SPOF, Single Points Of Failure) problem of the name node. paralysis. In the Hadoop2.x version, Hadoop proposed a Zookeeper-based HDFSHA solution. In this solution, a primary name node and a backup name node were used, and a Zookeeper cluster and a master-standby swi...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present invention provides a service cluster to solve the problem in the prior art that Zookeeper clusters need to be introduced in the Zookeeper HDFS HA ​​solution, which is very redundant for scenarios where only HDFS is used alone. And increase the problem of operation and maintenance costs, the technical solution is as follows:

Images