What is the single point of failure
is usually distributed system uses a master-slave mode, multiple processing nodes connected to a host, the master node is responsible for the distribution business, and the child node handles traffic when the primary node fails, it will cause the entire system fails, we such failure is known as a single point of failure.
Single point of failure to solve the conventional
1. Usually a backup master node will pay more, backup master node will send a ping packet to the master node, the master node will spare the master node receives a ping packet corresponding ack response byte packet illegally , whether a backup master node is received by the ack-byte packet to determine whether the normal operation of the master node, that the master node receives no fault has been replaced and the position of the master node.
2. However, when a loss occurs when the network shocks will lead to the master node receives the data part of the data, that is, small package, will lead to spare the child node considers the master node has failed away replace the master node which led to the phenomenon of multi-master, then you to join the concept of a distributed lock.
3. Distributed lock is to control multiple processes to access the same resource, the resource wants to be orderly access, lock the main control process, unlock, lock timeout. Let more into the car and orderly access to the server.
4. distributed coordination can solve the synchronization control multiple processes, the main core is to achieve a distributed lock
5.zookeeper is distributed coordination services, in order to achieve a distributed lock structure is a tree structure, the child node is mainly composed of four parts, a reference node, data rights meta data.
Reprinted: https: //blog.csdn.net/weixin_43739465/article/details/84328554
