Distributed lock

1. Distributed lock
   1. redis Distributed lock
   2. zookeeper Distributed lock
   3. Database distributed lock
   4. summary

Distributed lock

Distributed locking is an important topic in distributed projects , When we use the traditional java When the shared resources cannot be locked , We need to consider distributed locks . So what are the technical solutions of distributed locks ？ We can go through redis Distributed locks for , such as redis Of setxnx command , You can also use redisson, The second technical solution is to use zookeeper, What we use is zookeeper Temporary order node for , The third technical solution is to use the primary key or unique index of the database to unlock the lock

redis Distributed lock

Lock command ：setnx key value nx ex 10s

key You can set it according to your business ,value It can be a random value , Make sure the whole situation is unique ,nx Express this key Return success when it does not exist , Otherwise, the execution fails , This ensures that the lock will not be locked again when it exists ,ex Used to set key The expiration time of , Prevent the deadlock caused by the failure to release the lock after the program hangs during operation , With the expiration time , Even if the program does not actively release the lock , It's time to expire ,redis Will also let this key Fail to release the lock . If the current business is locked 10s I haven't finished my business ,key It's due , Other threads may lock successfully , It's time to be right key Renew , What you use is watch dog, When deleting a lock, make sure you delete your own lock , It's also in the process of deleting locks key Of value Compare , It is generally used lua Script to release the lock , Guaranteed atomicity

To make a long story short , Both locking and unlocking should ensure atomicity

Use redis Generally speaking, distributed locks redis Both are read-write separated architectures , Data replication is asynchronous , Some data may be lost during master-slave switching , because redis What is guaranteed is the final consistency

zookeeper Distributed lock

zookeeper Distributed locks are used zookeeper Characteristics of temporary sequential nodes , The client executes business logic before zookeeper Create a temporary sequence node in , And then to get zookeeper All the nodes in , If the current node is the smallest, locking succeeds , Otherwise, we have to wait . Releasing the lock is to delete the temporary order node you created , The reason why temporary contact is needed , Yes, ensure connection zookeeper After the service of hangs up, delete the node in time to release the lock .

zookeeper Distributed locks use temporary sequential nodes , If the client and zookeeper A network condition may cause the wrong release of the lock , And if there are too many nodes ,zookeeper Data synchronization between clusters also takes time

Database distributed lock

The distributed locking of database is to use primary key or unique index , Ensure that after the current business adds a piece of data, other businesses cannot add the same data, so as to ensure the success of locking , To release a lock is to delete data

summary

This article introduces the implementation of distributed locks , It's based on redis Distributed lock 、 be based on zookeeper And the distributed lock based on database , It is rarely used in production , If it is used, it will use the framework encapsulated based on these technologies, such as Curator、Seate wait .