Implementation of one interview question one distributed lock every day

redis Implement distributed locks

Use setnx command , This command means if does not exist , Then set the value , If it exists, the value... Will not be set .

The problem is

When the expiration time is not set , If a thread dies after obtaining a lock, it becomes a deadlock
When setting the expiration time , If the thread takes too long to process the business after obtaining the lock , Cause the expiration time to expire , Will cause the lock to release , Other threads get the lock , When it finishes processing business , It will release the lock obtained by other threads , Can cause data problems .

resolvent

You can use a daemon thread to renew the expiration time of a lock , When processing business takes too long , When the expiration time is approaching , Reset the expiration time .

defects

redis Finally, there is the problem of deploying multiple nodes , There may still be lock data loss , Of course, this is a small probability event , If the business can ignore such small probability events , Then you can use it redis Do distributed locks .

Zookeeper Implement distributed locks

First ,Zookeeper A lock node will be created , Then the client requests that need to be locked zk When locking , A temporary node will be created under the lock node , If multiple clients request , Then the temporary nodes will be created in the order of the requests , Each temporary node will have a sequence number , This serial number is zk Internally generated ; When the client acquires the lock , It will determine whether the temporary node with the minimum sequence number under the lock node is created by itself , If it is , Get lock , If not , Then listen to your last temporary node .
zk To release a lock is to delete the temporary node you created
When the client goes down ,zk Sense , The temporary node created by it will be deleted .