Oran Agra

what's the point in creating a cluster if it can't do any failovers? how is it useful? you can argue that since we still have a warning, this PR can't...

I think RM_Call is already overloaded with flags, and this (rather low level) feature should be an API of its own. note that the current behavior is not that it'll...

i think that adding a some sort of RM_DrainPropagationQueue or alike is also needed. i.e. in normal cases, a module command which works like a script, is executing several commands...

i think so, and if not we can always let it error. @guybe7 @MeirShpilraien please comment.

ok, that `RM_EndExecutionUnit` API is similar to my `RM_DrainPropagationQueue` (being also combined with `RM_HandleUnblockedClients`). the advantage is that it doesn't deal with internals much and just tells redis to do...

@redis/core-team please approve or comment.

we discussed this in a core-team meeting again, and didn't reach a conclusion. we want to re-consider adding an RM_Call flag to do that. i see that i already spoke...

> Do we still need this change with #12159 ? It seems like you should be able to do everything asked for here. i don't understand how they're related. the...

@madolson you mean that RM_Call with flush the exec-unit after each call, and if anyone wants to avoid it, they'll create a virtual client and wrap it in MULTI-EXEC? i...

> Maybe the better API is to "flush" the context, which closes out all of the pending state associated with it. that's what we're aiming to achieve here. i thought...