How does PfR calculate byte loss and packet loss

[taburley]

Hi there…I’m looking for some information about some timers and calculations of PfR =

Topology is a 4351 Master Controller, the Borders are 2921 running 3.16.1a

Development lab is running smoothly with no errors or issues.  All parameters are in compliance.

Production is reporting byte loss at the MC.  An iPERF test was performed to validate the  byte loss, and iPERF results did NOT coincide with the byte loss reported by PfR.  PfR reports loss, iPERF does not.

Here are a few questions we have regarding the situation:


1.  If the reachablity timer has been configured for 3600, is that the aloted time the routers have to establish reachablitiy or is that the time that the channel is unreachable before it is reported?

2.  Does out of order packets have an impact on byte loss calculations with 3.16.1a?

3.  When does the byte loss calculation get reset?  Is it timer based or real time?

4.  If a path takes several minutes to establish reachability how can the issue of the delay of establishing the reachablity be trouble shot?

5.  The CU has a phase iii path that doesn’t establish reachablitiy until both spokes have been reset.  The channel tries to establish, but cannot until both spokes in the phase iii tunnel are reset.  What is the best way to troubleshoot the PfR issues related to the channel establishing?

many thanks!

Tammy

[dlongpre]

Hi Tammy,

I know your post is 3 months old, but did you get answer to your questions ?

Regards,

Dominic

[kmckaypason]

I too am interested in your findings on byte loss. I've seen the same problems on a few trains of code.. 3.13.3-5 and 3.16.1.aS. I have requested the details of the Byte loss calculation from TAC but never got my answer unfortunately. My findings via iPerf are about the same as yours.. generating out of order packets definitely triggers byte loss but this was supposed to be resolved in later releases of the 3.13 train according to the release notes.

Looking forward to hearing from you.

Kyle

[damienpicard]

Same issue here on a recent IWAN 2.1 deployment. Please let us know if you managed to get any information on this issue.

[Vasilii Mikhailovskii]

Hello.

As the time goes by questions may change. What questions do you have today?

PS: if you have any live issue - please share details and logs from MC.

[damienpicard]

Hello Vasilii,

We have a TAC case open for this issue (SR 682418695), I noticed that you were in cc of this case, you can review all the logs from MC and other devices in this case. Probably better to work on this case and then post the solution in this thread if we find one.

I was looking on the community forum for additionnal help but the only other post I found was this one and it is bad news as we do not want to turn off PfR... https://supportforums.cisco.com/discussion/13242066/cisco-pfr-loss-rate-bytes-vs-loss-rate-packets

All in one, we keep getting TCA due to "bytes losses" but it seems to be false positive (and we do not see any "packet losses") so I was wondering how this bytes loss was calculated.

[Vasilii Mikhailovskii]

Hello.

Thanks for pointing to another topic - I've replied there; if you have any other questions regarding loss calculation - let's keep it there.

I've taken over the ticket your mentioned and reviewing the case notes.