Oleksandr Samoylyk wrote:
Jack Vogel wrote:
Oh, I just had a thought, increase the RX processing limit,
that only allows you to process 100 packets in one pass.

First change it to 250 and see what it does, you might
also set it to -1 which will allow you to process til you
drain the ring, the risk is that you cause other problems
by doing that, but heck at this point anything is worth
trying, right?


Nothing has helped. :(

I need to unplug and plug in again patch cords each time when my CPUs with emX go 100% in order to keep my server alive with a descent pings.

I mentioned that "100%: emX taskq" occurs only on that interfaces where GRE packets are being processed.

External interface to Internet feels great. Pings are <0ms and load is
9.57%  with 14kpps (input/output).

Maybe interesting:
According to kgmon:
  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  name
 39.9      93.10    93.10  1643247     0.06     0.07  rip_input [10]

Is it em related or mpd related or something else?

Back to releng_6? Not sure though. :(


Now I'm on 6.3-STABLE:

last pid: 31566; load averages: 7.61, 7.25, 7.07 up 0+01:34:00 22:30:05
82 processes:  10 running, 58 sleeping, 14 waiting
CPU states: 0.2% user, 0.0% nice, 52.1% system, 47.8% interrupt, 0.0% idle
Mem: 63M Active, 10M Inact, 152M Wired, 8K Cache, 35M Buf, 1752M Free
Swap: 4011M Total, 4011M Free

  PID USERNAME  THR PRI NICE   SIZE    RES STATE  C   TIME   WCPU COMMAND
   12 root        1 -44 -163     0K     8K WAIT   0  42:43 93.60% swi1: net
24 root 1 43 0 0K 8K CPU0 0 29:50 48.34% em0_rx_kthread_1 23 root 1 43 0 0K 8K RUN 0 29:50 46.78% em0_rx_kthread_0 28 root 1 43 0 0K 8K RORDER 1 3:30 2.25% em1_rx_kthread_0 29 root 1 43 0 0K 8K RUN 0 3:30 2.15% em1_rx_kthread_1 10 root 1 171 52 0K 8K RUN 1 11:22 0.00% idle: cpu1 11 root 1 171 52 0K 8K RUN 0 9:01 0.00% idle: cpu0

The results aren't good as well.

--
 Oleksandr Samoylyk
 OVS-RIPE
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to