tree 2ebf0e8162241a8014c67d2915932c711e7a4bc4
parent 65482772471ff028b0fbbf2306e01ec2c8b9c728
author wuYin <wuyinpost@gmail.com> 1612838518 +0800
committer GitHub <noreply@github.com> 1612838518 +0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJgIfZ2CRBK7hj4Ov3rIwAAdHIIADIo3Cl+rzSekcimUIe0pJxK
 lAPuMWcp1ufoqBMXXhnXT9K46mxvJDkMX9yMzsKs9B6iuDi9k3IPdtf1HJkIedll
 Bygs7v2HRgoCGjhIhOLVrw3uSB470jv75dMmPciCw3TtBxUx03dpJSQC678OOMZC
 rx3qczh+ynaevbx+Y/yPXmBtOsi42z3gdDz8/zoFE3tzCRzLV8f/S+G5CzUxaDjk
 xlYaU2MA+QvqP2venhQCRrKfOr106ANXJa1NXu2N49Z08T0eYUYejBJ0uZSbxiVR
 R9JQccIOywu95yDzVJ1IlD7xLP7zwO0uWVM9QmD+WOdWlUJ79PhT21ItpvGwfaY=
 =q7Lt
 -----END PGP SIGNATURE-----
 

asynchronized send timeout checking (#460)

Fixes #458

### Motivation

For current producer, send timeout checking triggered by interval batch flush
If connection closed, the producer eventloop will blocked to reconnect to broker, lead to batch flush and send timeout checking take no effective, java-client timer did effective in this situation

### Modifications

Asynchronized send timeout by running in independent goroutine until producer closed, and without a pending queue lock

### Verifying this change

- [x] Make sure that the change passes the CI checks.

### Others

Without pending queue lock, the send timeout checking  gets more complicated, I don't know if it's worth it for performance.