[j-nsp] 1000Base-LX link problem

[Enrico Rete]

I have a strange problem on a Single Mode Fiber link (1000Base-LX) between
my J4350 and service provider using a Cisco 6509. Most of the time the link
is good and data passes no problem, but my provider started calling me
stating that they saw the link flap for a second at such and such a time.
This is a big deal as the link is used to pass critical traffic. When I
looked in my logs I saw no errors (no link up down) and no errors show on my
interface nor theirs. Since this was first reported this has started
happening almost every day, sometimes twice, at random times of day. There
is a secondary set of routers (another J4350 of mine connecting to a
redundant Cisco 6509) and it has the same problem, again at random about
once a day.

Here is what I have determined:
Both my provider and I tested our fiber with loopback testing and it appears
clean. The provider apparently is using a Cisco GBIC which does
1000Base-LX/LH- theoretically it should be compatible with my card. There is
about 500 meters of fiber run between their cabinet and mine (maybe less).
Juniper helped me look at the the SFP and I found lots of TX errors in the
SFP (vty fwdd; show sfp 1 alarms) and voltage high low errors in the SFP.
Juniper sent me an unsupported 1000Base-LH SFP which I installed in one of
the two J4350s- no SFP errors in the fwdd, but the link problem stays the
same (flaps about once a day).

I tried to have my tech use a light meter to measure the power levels to see
if I am oversaturating my end or their end, or if I am receiving or sending
too little light, but as far as I can figure (I am not very good at
measuring light levels) everything is within the acceptable parameters of
the specs. The problem is when measuring dBm the meter I am using has to be
set to a wavelength, and given the mix of PICs I am not sure if the level
should be 1310 or 1550 nanometers. Regardless the levels seem right.

Question:
1) Is there some sort of fundamental incompatability between Cisco and
Juniper cards here that could cause this problem (sounds stupid- but I am at
my wits end here).
2) Should I be using an attenuator at this short distance on either my RX or
TX end?
3) Has anyone else come across a problem like this and been able to resolve
this?