Feature Request: Public API Exposure for Sentinel Topology Events

[jszczesnyoa]

Summary:

Expose Sentinel events—such as SentinelTopologyRefreshEvent—in the public API, so that irrecoverable events during failover (e.g. "READONLY You can't write against a read only replica") can be captured and acted upon without relying on reflection or custom workarounds. This is currently due to the fact that the events are package protected.

Context:

We use Lettuce (version 6.3.0.RELEASE) in a Redis Sentinel setup. During master failover events, we encounter irrecoverable errors such as receiving READONLY You can't write against a read only replica errors on connections. We are testing some mitigations such as #2082 however as the connection issues are very hard to reproduce we'd like to increase the visibility in order to produce alerts on certain Sentinel events.

Alternatives Tried:

Reflection:
We attempted to access internal fields for topology refresh data.

Custom Pub/Sub Listener:
We implemented a listener that subscribes to Sentinel’s +switch-master channel. For example:

StatefulRedisPubSubConnection<String, String> pubSubConnection = redisClient.connectPubSub();
pubSubConnection.addListener(new RedisPubSubListener<String, String>() {
    @Override
    public void message(String channel, String message) {
        if ("+switch-master".equals(channel)) {
            // Capture irrecoverable events, e.g. cannot write to a read-only instance.
            LOG.warn("Sentinel detected a master failover: {}", message);
        }
    }
    @Override public void subscribed(String channel, long count) {
        LOG.info("Subscribed to Sentinel notifications on channel: {}", channel);
    }
    @Override public void unsubscribed(String channel, long count) {}
    @Override public void message(String pattern, String channel, String message) {}
});
pubSubConnection.sync().subscribe("+switch-master");

Benefits:

Enables direct integration with external monitoring or recovery logic.
Eliminates reliance on fragile internal APIs or reflection.
Provides a more robust mechanism for detecting irrecoverable failover events (like inability to write to a read-only instance) and triggering corrective actions.

Request:

Please consider exposing Sentinel topology events (e.g. SentinelTopologyRefreshEvent) in the public API or providing a documented mechanism for subscribing to these events. This would allow external components to react to failover events, capture irrecoverable errors, and implement custom recovery logic without resorting to workarounds.

[tishun]

Hey @jszczesnyoa ,

thanks for filing this request.

I would like to dig a bit deeper to understand better the problem you are having.

Specifically about the alternatives, I can understand why using reflection was not an option, but I am curious why does the pub/sub approach not work for you? Is it that you would like to not map the message type / payload yourself and would like to have a better abstraction?

Generally if there is a way to do this right now - even if it is not ideal - this would reduce the priority as we have some heavy backlog with missing features and bugs.

We would be happy to receive a PR with it though, if you or somebody else are considering a contribution.