Open
Description
- Version: 12.20.1
- Platform: Linux VM-10-6-16-42-17-191 3.10.0-957.12.1.el7.x86_64 deps: update openssl to 1.0.1j #1 SMP Mon Apr 29 14:59:59 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
- Subsystem: docker centos 7
What steps will reproduce the bug?
When there are many new tcp handle send to the master process, it first save it in internal array, and find a free worker process to send to. But there is a low performance problem in the RoundRobinHandle. When the array's length exceeds a certain value, the master process would busy to call shift function of array, its CPU percentage will raise up to 100%, but the children workers have low CPU usages in the other hand.
I have generated a CPU flame graph as follows:
