3

Given one master node and multiple replica nodes with Sentinels enabled, Redisson configured to read from replica nodes only, and round robin load balancer and retries enabled. In case of a Redisson retry attempt, does it invoke the load balancer again to determine a new replica node or does it talk to the same replica that it had unsuccessfully communicated with before?

Our use case is such that we see are witnessing a 50% drop in commands processed per second when a single replica node restarts. We have retries enabled in Redisson but they don't seem to be helping so we are wondering if the retries also go to the same failing node.

Redisson version: 3.13.6 Redis version: 5.0.6

Deepak Bhimaraju
  • 71
  • 2
  • 4

1 Answers1

1

does it invoke the load balancer again to determine a new replica node or does it talk to the same replica that it had unsuccessfully communicated with before?

In each attempt it uses load balancer.

We have retries enabled in Redisson but they don't seem to be helping so we are wondering if the retries also go to the same failing node.

You need to increase attempts amount to survive slave node failover. Also you can reduce failedSlaveCheckInterval parameter to detect failed slave nodes faster without waiting Redis cluster topology update.

Nikita Koksharov
  • 10,283
  • 1
  • 62
  • 71
  • What's the downside if we make failedSlaveCheckInterval to a small value like 5 secs ? – Giri Aug 17 '21 at 19:01