Ran Shidlansik
Ran Shidlansik
@liyaoxuan - would like to ask the status of this change? I believe we all left it aside for some time.
@liyaoxuan thank you for this effort!! I just think we can try and encapsulate the readHandler qps logic and move it to a dedicated function in order to keep better...
> @ranshid I've added some new commits to fix bugs and hope you can review them, so I click the re-request review button. I don't know if it is a...
> @ranshid @yoav-steinberg expecting approval or questions or suggestions :) LGTM
@shawn-higgins1 great to see this work is being done to promote the cluster mode benchmark! I placed some comments which I think might help to simplify things. Thank you!
@shawn-higgins1 LGTM Thank you! I still feel the name of the parameter 'readonly' - does not directly reflect the outcome of activating this feature (allow read from replicas) but I...
> maybe we can start by creating an alternative, and using it in the few places that have performance concern and the few with complicated error handling (note that there...
> How about remaining the api but substituting the function implementation, thus the caller code wouldn't be broken. @uvletter the problem with this is that the current API return a...
> @ranshid i usually try to avoid refactoring, specifically if not strictly necessarily. for one, it causes a mess in the blame log, making it harder to find out the...
@oranagra - we have made some internal benchmark tests after apply the change over ALL use cases of dictGetIterator. as expected we were unable to identify a noticeable performance improvement,...