Sentinel Performance

[basdeve...@gmail.com]

I have a 3 node sentinel setup. Node 1 is a master with Nutcraker and Twemproxy agent. Rest two nodes are slave.

I am having a application running on server which is having a approx. 6000 records input of data per minute during some defined time period.

Server configuration for all three server are 16 VCPU, 64 GB RAM , 100GB of space.


We have Master with two slaves and redis-sentinel configuration on all the redis instance (configuration involves changes made in redis-sentinel.conf by changing my master parameter)

Twemproxy Agent is used with sentinel so that automatic updation of new master is done in nutcracker.yml file.

The problem I am facing; sentinal works fine but the application is delaying the data storage into the redis. Before sential it was approx. 0.3 sec which increase to 10 sec after sentinal.

Do anyone have any comments about performance or solution in the above issue i am facing.

[The Real Bill]

Since Sentinel is only ever consulted when needing to determine where to connect to, and you are apparently not using sentinel in in-band or client code, then Sentinel "performance" is unlikely to be the cause. 

You'll need to examine the sentinel logs to see if you changing the config is resulting in sentinel trying to fix it for you and possibly causing twemproxy restarts (since it does not yet support hot reload) which could in theory be causing delays. But you've not posted enough details on your setup and conditions to do much more than guess at this point.

[basdeve...@gmail.com]

Thanks for the message.

I am attaching logs...please check and advice on the problem.