Performance issues with redis cluster compared to redis sentinel

[ardadem]

Hello everyone, i am having an issue with redis cluster (using go-redis on the client side) deployment, and researching what the problem can be. So in the same machine, everything is exactly the same, the only difference is that i have switched to redis cluster (3 masters and 3 slaves in this deployment) from the redis sentinel (1 master and 2 replicas). In my tests, it is 3 to 5 times slower with overall go-redis calls (i am using only GET and SET commands). Does anybody have an idea what should i check, what could be the issues, how can i track down the bottleneck? It would be really helpful to having a hint about what should i focus on or hearing similar experiences.

[ardadem]

So in the end, the issue wasn't related to redis cluster, and the bottleneck caused by a configuration in the bitnami helm charts. So i am closing this conversation as resolved.

[ndyakov]

Hello @ardadem, Can you share more about what was wrong with the configuration and how you found the issue is not related to the client and/or Redis cluster in general? I do think this can be beneficial to other users trying to troubleshoot similar issues.

[ardadem]

Yeah, of course. I've noticed latency spikes in the output of these commands: "redis-cli --latency" and "redis-cli --intrinsic-latency 100" (see https://redis.io/docs/latest/operate/oss_and_stack/management/optimization/latency/). So it showed me that it is not really related to the client but instead, there is a bottleneck on the redis instances. So i've researched and learned that there is an option on the bitnami charts called "resourcePreset" and it is nano by default, which is limiting the resources of the redis containers a lot. Setting that option to "none" (or an option with the higher resources) have fixed my issue.