Distributed lock

Catalog

One 、 database

Two 、Zookeeper Distributed lock

3、 ... and 、Redis Distributed lock

---

Why do I need distributed locks ？

Distributed projects , Each project is independent . The lock needs to be independent of each service , Not in service .

One 、 database

1. Using primary key conflict control, only one thread can obtain the lock at a time , Non blocking 、 Do not reenter 、 Single point 、 Failure time, etc .

Two 、Zookeeper Distributed lock

zk Through temporary nodes , Solved the deadlock problem , Once the client gets the lock, it suddenly hangs up （Session Connection is broken ）, Then the temporary node will be deleted automatically , Other clients automatically acquire locks , Temporary sequential nodes to solve the group shock effect （ In multiple processes / When multiple threads wait for the same resource , There will also be surprises . That is, when a resource is available , Multiple processes / The thread will wake up , Compete for resources . This is the surprise group in the operating system ）.

3、 ... and 、Redis Distributed lock

settnx, Single threaded processing of network requests , There is no need to consider concurrency security , All service nodes have the same settings key, return 0, Lock acquisition fails .

setnx problem ：

1. Earlier versions did not have a timeout parameter , It needs to be set separately , There is a deadlock problem （ Halfway down ）.
2. Later versions provide atomic operations of locking and setting time , But there is a task timeout , Lock release , Causes concurrency problems , Locking and releasing locks are not the same thread problem .

Delete lock ： Determine the unique ID of the thread , And then delete .

Reentry and lock renewal are not realized , adopt Redisson solve （ similar AQS The implementation of the ）

redlock：redis Ensure high availability through sentinel mode , If this master For some reason, the node has switched from master to slave , Then the lock will be lost （redis Synchronization may cause data loss ）.redlock Apply for locks from multiple nodes , When more than half of the nodes succeed , Lock is success ,Redisson There is a corresponding implementation .