Mikael Abrahamsson <[email protected]> writes:

> I don't know how the Linux AQM system works, but exposing some of
> these counters through the proc subsystem (or something else), might
> work, and then a userspace daemon could poll this data as often as it
> wanted, then you don't even need the kernel to do averaging and
> keeping the data.

It's exposed via the Netlink interface, which is what `tc -s` shows:

qdisc fq_codel 110: dev enp2s0 parent 1:10 limit 1000p flows 1014 quantum 500 
target 5.0ms interval 100.0ms ecn 
 Sent 16981968691 bytes 17816348 pkt (dropped 8, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0 
  maxpacket 13216 drop_overlimit 0 new_flow_count 6502039 ecn_mark 0
  new_flows_len 1 old_flows_len 5

The 'backlog 0b 0p' shows that the queue is currently empty, while the
lines below show qdisc-specific stats. But as Dave said, getting the
statistics has some locking issues that may (or will, probably) cause
trouble if trying to poll it too frequently...

-Toke

_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm

Reply via email to