Clustered Redis: periodic reconnects after idle; how to keep all channels active

[patrykwierzbilo]

I see reconnect INFO logs on idle against a managed, clustered Redis.
TCP keep-alive is enabled. No timeouts. Traffic is fine.
I need a supported pattern to keep all channels active when the app is idle.

Environment:

Lettuce:

• lettuce-core:6.4.2.RELEASE
• spring-cloud-azure-starter-data-redis-lettuce:5.22.0

Spring Boot: 3.4.8
spring-data-redis:3.4.8

Java: 21

Runtime: Kubernetes (Linux)/ AKS

Server: Azure Cache for Redis (TLS). Also reproduced locally with grokzen/redis-cluster (--timeout 45 --tcp-keepalive 0).

What happens

• On idle, every ~10 min (or 45s in local repro) I see:

INFO ConnectionWatchdog: Reconnecting, last destination was <cache-host>:15000
INFO ReconnectionHandler: Reconnected to <cache-host>:15002

• CLIENT LIST shows our client idle grows to the server timeout before the reconnect.

What I tried

• Enabled TCP keep-alive (TCP keep-alive helps detect dead links but does not reset Redis’ idle counter).

• No pool (single shared connection) + scheduled PING per shard like (tested locally)

  private final RedisConnectionFactory factory;

  @Scheduled(fixedRate = 20_000)
  public void pingAllShards() {

    try (RedisConnection raw = factory.getConnection()) {

      if (!(raw instanceof RedisClusterConnection cluster)) {
        log.warn("Heartbeat skipped – not a cluster connection");
        return;
      }

      for (RedisClusterNode node : cluster.clusterGetNodes()) {
        try {
          String pong = cluster.ping(node);
          log.debug("Heartbeat {}:{} → {}", node.getHost(), node.getPort(), pong);
        } catch (Exception ex) {
          log.warn("Heartbeat failed for {}", node, ex);
        }
      }
    }
  }

I used Lettuce config from documentation: link.

Why I’m asking

• Operationally this is not breaking service, but the reconnect logs add noise.

• I’d like a supported solution (pattern or API) that works with Spring Data and Lettuce, keeps clustered channels warm during idle, and avoids reconnects.

[tishun]

Hey @patrykwierzbilo ,

Azure have their own idle connection management system, briefly mentioned here.

To avoid that it is not enough to set a keep-alive, however it is necessary to send a command to avoid having to reconnect with no reason. A possible solution is described in the description of #2078

Since this is cloud provider-specific it would most likey continue to be a custom solution outside the driver itself.

Does this help?

[patrykwierzbilo]

Yes, it fixed my problem. Thanks for your help!