I have been bit by flow control more times than I care to remember. See
if you have any Tx or Rx pause frame counters on either or both ends. If
not that then I would suggest a hardware issue of some kind, maybe a
failing SFP module.
On 9/26/2014 10:25 AM, Adam Moffett via Af wrote:
The customer is having real issues, so I don't think it's a matter of
how pings are handled compared to other traffic.� There's also
pretty much zero load on any of this stuff.
You see this with Ciscos all the time, because pings are handled at
the process level rather than the interrupt level.� I would suspect
the Mikrotik of something similar, but since MT is linux-based, AFAIK
all pings are handled in the kernel at interrupt level.� But that's
just a guess, so perhaps MT is handling the pings at the process
level for some reason.
On 09/25/2014 03:25 PM, Sterling Jacobson via Af wrote:
I�ve seen that before, but not with fiber anywhere.
�
My current deployments with RB2011 don�t show this and it�s
similar to your setup.
�
�
*From:*Af [mailto:[email protected]] *On
Behalf Of *Adam Moffett via Af
*Sent:* Thursday, September 25, 2014 2:41 PM
*To:* Animal Farm
*Subject:* [AFMUG] Mikrotik-tik-tik-tik
�
So I've got several setups like this: CCR <-> SFP <-> Fiber <-> SFP
<-> RB2011
sometimes pinging the RB2011 I can see this once per second
delay.� Those pings are at an interval of .2 seconds (ping -i .02)
so you can see the delay on every 5th packet corresponds to a once
per second "tick" of some sort.� If I vary the interval, the tick
still occurs every one second.� I have multiple installations that
do this, and multiples that don't....and I cannot find any rhyme or
reason to it.� Connected to one CCR on SFP2 I have an RB2011 that
has the symptom, and then I made a virtually identical installation
on SFP3 that doesn't do it.� The only thing different is the IP
addresses and the length of the fiber (3 feet on the good one, a
couple thousand feet on the bad one).�
The delay varies anywhere from a few ms to upwards of a hundred ms,
and when it's high it affects VoIP so it is a real issue.� I have
a few more combinations of things to test, but I wonder if somebody
has seen this already who can save me a ton of time.� Anybody?
P.S.:� I emailed [email protected]
<mailto:[email protected]> yesterday.� Do they eventually
respond or is it a blackhole?
!DSPAM:2,542496be61881139814307!
