Problem with get slave via sentinel.

[valentyn-hrynevich]

How to get a slave instance via sentinel.
For example https://github.com/andymccurdy/redis-py have method "slave_for" that return client object with slave host.
How to get something same using this library.
I don't found anything about in a documentation.
Thanks for help.

[vmihailenco]

NewFailoverClient returns plain redis *Client so I am not sure where to add slave_for method. And now we even don't retrieve/store information about slaves which is another problem.

I guess that you want to use slave(s) for read only commands. If that is the case I would recommend you to use Redis Cluster and enable ClusterOptions.ReadOnly and/or ClusterOptions.RouteByLatency.

[valentyn-hrynevich]

Okay, but I don't want to configure cluster stack. May be will be good to implement parameter for get random slave instead of getting master permanently.
What do you think about it?
E.x https://github.com/luin/ioredis has parameter role.

[totoyotonet]

I do not know how to do it

[vmihailenco]

Closing since I don't understand what you are trying to achieve and there is no good place for method like "slave_for".

[agparadiso]

Im trying to do the same it was asked here, get a sentinel slave in order to split reads/writes. haven't found a way of doing it with this library. btw the closing reason doesn't seems to be appropriated.

[vmihailenco]

#997 looks promising, but it requires more work.

[eafzali]

@vmihailenco I'm happy to work on that if you give it a review and could help me understand your build pipeline

[yushizhao]

I have the same need, namely, route all read commands to slaves.
From what I read here, I think @vmihailenco have suggested to create a go-redis ClusterClient with ClusterOptions.ReadOnly. Were you suggesting then all the read commands passed by ClusterClient::Do will route to a slave while all the write commands go to master? I see there is a work around in clusterState. However, I want to confirm when Sentinel failover a master, how would clusterState get updated? Thank you.

[landrade]

Any workaround to resolve it? Write on master and read from slaves is a common approach for applications that don't need sharding data but have a high throughput. It will be good if we have a alternative to do it.