improve latency when a client is unblocked by module timer

[tzongw]

Describe the bug

1. client block on command XREAD BLOCK 0 STREAMS mystream $
2. in a module, calling XADD mystream * field value via lua from a timer callback
3. client will receive response after some latency up to 100ms

Reason
When XADD signal the key mystream as ready, beforeSleep in next eventloop will call handleClientsBlockedOnKeys to unblock the client and add pending data to write but not actually install a write handler, so next redis will block in aeApiPoll up to 100ms given hz config as default 10, pending data will be sent in another next eventloop by handleClientsWithPendingWritesUsingThreads.

Calling handleClientsBlockedOnKeys before handleClientsWithPendingWritesUsingThreads in beforeSleep solves the problem.

Related issues #8125 #7880 #7903

[oranagra]

@tzongw thank you.
i see you mentioned the XADD was from a module via a timer event, but as far as i can tell the exact same problem will happen for a normal client sending XADD.
the 100ms delay will happen when redis is not experiencing any traffic (i.e. if someone sends a high rate of traffic or event PINGs) it'll dramatically shorten the latency.
anything i'm missing?

[tzongw]

@oranagra I didn't see the same problem happen for a normal client sending XADD. I notice that handleClientsBlockedOnKeys is called in processCommand, so handleClientsWithPendingWritesUsingThreads in beforeSleep in next eventloop will send the pending data, no aeApiPoll call between them.

[oranagra]

I didn't see the same problem happen for a normal client sending XADD

@tzongw do you mean that the problem doesn't exist in that case? or just that that's not the case in which you found it?
i wanna simplify the description at the top (for the sake of release notes), and specify that it affects any unblocked client.

[tzongw]

do you mean that the problem doesn't exist in that case? or just that that's not the case in which you found it?

@oranagra The problem doesn't exist for a normal client.

i wanna simplify the description at the top (for the sake of release notes), and specify that it affects any unblocked client.

Sure, fell free to modify the description.

[oranagra]

ohh, now i see what you wrote about handleClientsBlockedOnKeys being called from processCommand (and the code comment which i didn't bother to read).
thanks.