Understanding RedisURI vs TimeoutOptions in Lettuce

[wmxl]

I'm working with Lettuce as my Redis client and have some questions regarding timeout settings and their observed behavior:

RedisURI Timeout vs TimeoutOptions: Could someone clarify the nuance between setting a timeout on RedisURI and using TimeoutOptions, particularly when I don't specify a particular command type? I understand RedisURI provides a general timeout setting, and TimeoutOptions offers more detailed control, but I'm not clear on how they interact or override each other.

Actual Command Execution Time Exceeding Set Timeout: Despite setting timeouts both on RedisURI and via TimeoutOptions, I've noticed that the actual execution time of some commands exceeds the set timeout value. What could be the reasons for this discrepancy? Is it related to factors like network latency, server load, client-side processing, or potential misconfigurations in my setup?

(In my tests using Lettuce as a Redis client, I've observed differing behaviors with regard to timeouts for GET and MGET commands. It appears that the GET command consistently adheres to the set timeout values, but MGET does not. Does this suggest that TimeoutOptions in Lettuce are applicable on a per-command basis, specifically affecting single commands differently than multi-key commands like MGET?)

Any insights or experiences with similar issues would be greatly appreciated!

[mp911de]

There's a bit of history involved. Initially, we had only RedisURI.timeout that was applied in some places (connect timeout, command timeout through the synchronous API but not the async API). TimeoutOptions defines a full set of how to determine a timeout. TimeoutOptions can either apply the command timeout that is set on RedisURI or you can specify a custom TimeoutSource to control timeouts for individual commands.

Timeout starts as soon as the command is dispatched into a connection. That means that a command isn't written to Redis yet when the timeout starts.

On Redis Cluster, a command (such as MGET) is partitioned across hash-slots and Lettuce sends multiple MGET commands with slices of your keys that share the same hash-slot. You might observe there a different timeout behavior than in the Redis Standalone variant.

If you have further questions, ideally provide a scenario that you want to understand better and please provide some code/some arrangement so that we talk the same things.