Setting maxWaitMillis from 50ms to 25ms causes a large number of cluster slot commands to be executed

[bozai111]

Redis cluster and redis client version information：
jedis version :2.9
Redis cluster version : 4.0 Shard deployment：100 master 100 slave qps：2w per second for a single shard

scene description
Setting maxWaitMillis and soTimeout parameter from 50ms to 25ms causes a large number of cluster slot commands to be executed,
The redis cluster load is very high and the cluster service is unavailable,Is there a best practice for jedis parameter setting?

[sazzad16]

First of all, Jedis 2.9 an ancient version. Please use a latest/updated release.
There is no defined 'best' practice but it'll be a good idea to increase the time related parameters. I'm assuming your maxTotal is not high enough, so increase that as well.

[sazzad16]

Is there a recommended value for maxTotal?

No. commons pool uses 8 as default and it seems good enough for most applications. But as it seems it's not good enough for you.

[bozai111]

Current development is used to expanding the number of application servers to accommodate more business requests, and the number of connections to redis cluster is also increasing. Do we need to restrict this behavior?

[sazzad16]

Is Jedis(Cluster) running in application servers?

[bozai111]

yes，There is jedis on each application server, and the jedis client will be initialized when the application service is started

[sazzad16]

Current development is used to expanding the number of application servers to accommodate more business requests, and the number of connections to redis cluster is also increasing. Do we need to restrict this behavior?

There's no simple answer. If you find Redis servers (shards) are getting very busy, you may consider the restriction.