After 1 month and 2 weeks I found a solution :)! The main idea is to redirect outgoing traffic to ifb device from every queue of the real eth interface.
Example: tc qdisc add dev eth0 root handle 1: mq tc qdisc add dev eth0 parent 1:1 handle 8001: htb tc filter add dev eth0 parent 8001: u32 ...... action mirred egress redirect dev ifb0 ... tc qdisc add dev eth0 parent 1:4 handle 8004: htb tc filter add dev eth0 parent 8004: u32 ...... action mirred egress redirect dev ifb0 2017-10-10 18:00 GMT+06:00 Sergey K. <simker...@gmail.com>: > I'm using Debian 9(stretch edition) kernel 4.9., hp dl385 g7 server > with 32 cpu cores. NIC queues are tied to processor cores. Server is > shaping traffic (iproute2 and htb discipline + skbinfo + ipset + ifb) > and filtering some rules by iptables. > > At that moment, when traffic goes up about 1gbit/s cpu is very high > loaded. Perf tool tells me that kernel module > native_queued_spin_lock_slowpath loading cpu about 40%. > > After several hours of searching, I found that if I remove the htb > discipline from ifb0, the high load goes down. > Well, I think that problem with classify and shaping by htb. > > Who knows how to solve?
