java - Lock on a Resource queue to ensure only one container access the Resource - Stack Overflow

admin管理员组

文章数量:1341455

I am planning to develop a Java application and deploy it in K8s containers where the minimum containers are 3.

My requirement is to maintain a common queue for the 3+ containers and run a scheduler all the containers. When the scheduler runs on all the 3 containers, only one container can access the queue and perform Read Write operations on the Queue.

The other containers should wait for the lock to be released and only on acquiring the lock perform Read Write Operations on the queue.

Redis Cache is one option am exploring. Please suggest me if any other possible solutions.

I am planning to develop a Java application and deploy it in K8s containers where the minimum containers are 3.

My requirement is to maintain a common queue for the 3+ containers and run a scheduler all the containers. When the scheduler runs on all the 3 containers, only one container can access the queue and perform Read Write operations on the Queue.

The other containers should wait for the lock to be released and only on acquiring the lock perform Read Write Operations on the queue.

Redis Cache is one option am exploring. Please suggest me if any other possible solutions.

• java
• apache-kafka
• redis
• queue

Share Follow asked Feb 24 at 6:13 Sangram AnandSangram Anand 10.7k2424 gold badges7474 silver badges105105 bronze badges

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 1 +200

Redis has a cool feature named Consumer Groups.

This feature is based on:

• Redis stream introduction
• Redis Consumer Groups
• Redis stream insight

The application can have:

• 1 stream where the jobs to do are stored
• 1 group connected to the stream
• X consumer connected to the group, where ONLY ONE CONSUMER AT A TIME can read a new message

Example:

docker run -d --name local-redis redis
docker exec -it local-redis redis-cli

Then run these commands:

-- create a consumer group
XGROUP CREATE my-stream my-group $ MKSTREAM

-- create one consumer for each node
XREADGROUP GROUP my-group node-1 COUNT 1 STREAMS my-stream >
XREADGROUP GROUP my-group node-2 COUNT 1 STREAMS my-stream >
XREADGROUP GROUP my-group node-3 COUNT 1 STREAMS my-stream >

-- add a job and create the stream
XADD my-stream * foo bar

-- now, if you run XREADGROUP again, only the first node that will run it,
-- will read the message, the other nodes will read `nil`

-- the node that reads the job, must tell it to redis
XACK my-stream my-group "1741454766159-0"

-- or the job will be pending forever
XPENDING my-stream my-group

-- read the stream content: the jobs are still there
XRANGE my-stream - +

-- remember to run the XTRIM to clean the processed jobs
XTRIM my-stream MINID "1741454766159-0"

This is a simple implementation, it is needed to take care:

• when the my-stream should be trimmed? (there a are many techniques such as MAXLEN or XTRIM)
• do not leave unused group (so XGROUP DESTROY must be called)
• enable the auto-ackwledge feature or, call XACK + XPENDING to avoid unfinished jobs

本文标签： javaLock on a Resource queue to ensure only one container access the ResourceStack Overflow