Question posted in
The official documentation can be found here.

When do Redis PUB/SUB events fire-and-forget?

Redis events are fire-and-forget. If no one is listening, they will just remain undetected. There’s no history. But nowhere on the internet did I find anything that answers the following question (I’m using redis-py and aioredis for python):

Does the code actually need to LISTEN to the events to receive it or will the redis connection cache events for me?

Let’s suppose I have this loop:

redis_db = Redis(...)

while True:
    print("Hey, I'm listening now!")
    msg = redis_db.blocking_subscribe("some_topic") # pseudo-code, because I'm using various libraries
    if msg == "calc":
        do_a_heavy_5_seconds_calculation()
        print("YAY, I was busy and am back up now!")

So will I miss all events that occur in the 5s calculation or will my redis_db cache them for me?

In other words… does fire-and-forget apply if I don’t have a connection or if I have a connection, but am not listening ?

Tags:

3

Answers


  1. 0

    As you mentioned, if you are using Redis Pub/Sub, it is a fire and forget so your application has to be up and running and listen to the event.

    I have to say that I have not tested your use case.

    For your logic, to have more control, I would suggest you look at Redis Streams.

    • First of all, this is not a fire-and-forget publish-subscribe, ut a more resilient messaging system.
    • Your application will control which messages to read, have to be read and send an ack when read.

    Your pseudo-code will be very easy to implement with Streams, with better control.

    Login or Signup to reply.
  2. 0

    “Fire and forget” applies to the server’s contract with the client. It means that the message will be sent only once, assuming that the client is connected. If the message is lost in the network and never gets to the client, then that’s that. If the message gets to the client and the client process/server crashes before reading it, then that’s that.

    In your case, the client (redis-py) listens to the network socket that makes the connection to the server. The socket itself is handled by the OS and buffers the incoming messages. As long as you keep draining the socket by reading from it messages aren’t lost.

    Login or Signup to reply.
  3. 0

    It’s a bit of both. You need a connection AND that connection must be in a “subscription” state. As soon as execute the “blocking_subscribe” method, you have effectively changed that connections state to the “subscription state” so redis will continue to buffer these messages for you regardless if your python thread is busy processing other messages at the same time. Now of course, if it crashes for any reason, then that connections server side buffer will be lost.

    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search