Juniper EX4200 shows wrong RX diagnostic values when in virtual chassis setup

A small hint for your installation with Juniper gear. We faced a very interessting situation with two EX4200 in virtual chassis setup. The following behaviour can drive you crazy:analyzing the optical parameters of a SFP+ straight on the Interface is fine. In this case here we had two SFP+ connected back-to-back with 2m Singlemode fiber:
#show interfaces diagnostics optics xe-1/1/2
Physical interface: xe-1/1/2
  Laser bias current                    : 33.174 mA
  Laser output power                    : 0.8950 mW / -0.48 dBm
  Module temperature                    : 45 degrees C / 113 degrees F
  Module voltage                        : 3.2000 V
  Receiver signal average optical power : 0.8341 mW / -0.79 dBm
after activation of Virtual Chassis and keeping the wiring as before you can't access the transceiver parameter no longer from the interface directly (since JunOS 12.3 R3 possible again). To gather the diagnostic values you have to use the Virtual Chassis port. More interessting is the fact that somehow "gremlins" added an imaginary attenuator of 20dB into the link although the phyiscal setup wasn't touched!
# show virtual-chassis vc-port diagnostics optics vcp-255/1/2
Virtual chassis port: vcp-255/1/2
  Laser bias current                    : 27.606 mA
  Laser output power                    : 0.8730 mW / -0.59 dBm
  Module temperature                    : 44 degrees C / 112 degrees F
  Module voltage                        : 3.1730 V
  Receiver signal average optical power : 0.0089 mW / -20.52 dBm
To find a proof for the 20dB delta we added an attenuator of 10dB to our back-to-back link. With this attenuator we were able to reproduce this pattern (our tests were done with 5dB and 15db as well). Looking as a software developer the decimal mWatt value of "Receiver signal average optical power" is shifted by two digits. Actually this is the problem. In the field people use Dezibel rather than mWatt. Just keep a 20dB Offset in mind.
without Virtual Chassis - with 10dB:
Physical interface: xe-1/1/2
  Laser bias current                    : 32.690 mA
  Laser output power                    : 0.8900 mW / -0.51 dBm
  Module temperature                    : 44 degrees C / 112 degrees F
  Module voltage                        : 3.2010 V
  Receiver signal average optical power : 0.0576 mW / -12.40 dBm
with Virtual Chassis - with 10dB:
Virtual chassis port: vcp-255/1/2
  Laser bias current                    : 27.364 mA
  Laser output power                    : 0.8690 mW / -0.61 dBm
  Module temperature                    : 44 degrees C / 111 degrees F
  Module voltage                        : 3.1740 V
  Receiver signal average optical power : 0.0006 mW / -31.90 dBm
This issue is known for JunOS 12.2R4 and 12.R5. The diagnostic values can only be gathered via the VCP
Update: with JunOS 12.3 R3 you can get the diagnostic values directly from the inteface again, even when the switch is in virutal chassis config. And hurray, the values are correct.

1 comments

Peter O'Kane Peter O'Kane November 4, 2016 at 14:56

This is still happening with JunOS 12.3R3.4 and 10G LRM optics P.1396.2M-JUN02. All receiver related number both measurements and thresholds moved by 20dB. On 12.3R6.6 I am seeing received power scaled the other way, received power 1.0767 mW / 0.32 dBm where the transmitted power at the other end is 0.6450 mW / -1.90 dBm. No way 80 meters of OM2 has 2.22dB gain.

Reply

Leave a Reply

Your email address will not be published.
* Required field
Your data will be gathered and processed according to our privacy policy.