Publish and subscribe to messages with Spring
Send and receive Redis pub/sub messages in a Spring application.
Spring Data Redis supports Redis pub/sub with two parts: the template publishes messages, and a message listener container receives them. See the pub/sub use case for when to choose pub/sub.
Subscribe to channels
Define a RedisMessageListenerContainer bean. Add one listener for each channel or
pattern you want to receive. This example subscribes to the news channel and to every
channel that matches alerts:*:
@Configuration
public class PubSubConfig {
@Bean
RedisMessageListenerContainer listenerContainer(RedisConnectionFactory connectionFactory) {
MessageListener listener = (message, pattern) -> {
String channel = new String(message.getChannel(), StandardCharsets.UTF_8);
String body = new String(message.getBody(), StandardCharsets.UTF_8);
System.out.println(channel + ": " + body);
};
RedisMessageListenerContainer container = new RedisMessageListenerContainer();
container.setConnectionFactory(connectionFactory);
container.addMessageListener(listener, new ChannelTopic("news"));
container.addMessageListener(listener, new PatternTopic("alerts:*"));
return container;
}
}
The container subscribes when the application starts and delivers each message to your listener on a separate thread.
If you use a MessageListenerAdapter to call a method on your own class, the
second String argument of the handler method is the pattern that matched, not the
channel. For a pattern subscription such as alerts:*, the adapter passes
alerts:* rather than alerts:disk. Read the channel from message.getChannel()
in a MessageListener, as in the example.
Publish messages
Call convertAndSend on the template:
template.convertAndSend("news", "hello");
template.convertAndSend("alerts:disk", "full");
With the listener container above, the listener receives news: hello and
alerts:disk: full.
Delivery guarantees
Redis pub/sub is fire-and-forget. A message published while your application isn't subscribed is lost. If you need to keep messages until a consumer processes them, use a stream instead.