27 km span on an MES3324F: the same modules link up in an MXA-64 but Eltex takes the port to PORT_SUSPENDED
A link between two of our sites: about 27 km of single-mode fiber, MES3324F on both ends. Installed a pair of 1550 nm FANG HANG SFPs sold as 40-km modules. The link comes up, lives for a few dozen seconds, and drops, over and over, until the port gets blocked.
- 2 x Eltex MES3324F, firmware 4.0.7
- SFP FANG HANG 1550 nm, part number S1253271221932, rated 40 km
- ~27 km single-mode span, intermediate patch panels at both sites
- patch cords from the same batch as at neighboring links
console# show fiber-ports optical-transceiver
Port ... Output Power Input Power LOS
gi1/0/23 -6.11 dBm -22.00 dBm No
%LINK-W-PORT_SUSPENDEDWhat we've already checked:
- measured the run, no issues with the cable
- moved this same pair of modules with the same cords into an MXA-64 on the same fiber - the link comes up immediately and stays up with no flapping
- swapped the modules and moved them to neighboring optical ports, same behavior
I don't like the -22 dBm on the receive side myself, but then it's unclear why everything holds on the MXA-64 over this same fiber. Is the MES3324F just this intolerant of working near the sensitivity limit, or is it the modules after all and I need to find a proper 40 km pair? I'd rather not go get new modules only to run into the exact same thing.
Show original (Русский)
Also available in: Русский
العربية
Deutsch
English
Español
Français
हिन्दी
Bahasa Indonesia
Italiano
日本語
한국어
Nederlands
Polski
Português
Türkçe
Tiếng Việt
简体中文
Comments 4
-22 dBm on the receive side for gigabit at 1550 nm is already right at the edge, where any extra connector makes the difference. A few questions so we're not guessing:
show fiber-ports optical-transceivershow on the far end? Interested in the Input Power there specifically, not just on your side.In my experience, "40 km" in a listing's description and the module's actual reach are two different things, and at 27 km that's the first thing to surface.
The picture is mirrored on the far end, receive is just as low, the difference is within margin of error. About the modules: the seller eventually admitted these are 20-km units, not 40, so there was no budget margin for 27 km to begin with.
In parallel we wrote to Eltex support. Short answer: check that the transceivers actually comply with DDM, update the firmware, and, since the modules are gigabit, add on both ends
Did all three - the port still reaches
%LINK-W-PORT_SUSPENDED, just a bit later. Didn't check levels on the MXA-64, I admit, only looked at whether the link was up.Now waiting for a proper 1550 nm pair with budget margin, until then I consider the question open.
Ran into something similar on Cisco, only there it showed up not as a port block but as garbage in the counters: one of two optical uplinks racked up 46 thousand input errors and 42 thousand CRC, and the log showed an alarm Rx power low, -20.2 dBm against a -18.4 dBm threshold.
It wasn't fixed on the switch: cleaned both ends, measured TX and RX with a meter, checked the run length, splices, and patch quality. Some errors went away, but the levels never fully returned to normal, and the module ended up being replaced too. The point is that at the edge of the budget hardware behaves differently: one racks up CRC and pretends to work, another honestly shuts the port down. So your -22 isn't "barely working", it's already the cause.
About measuring with a meter, I'll add: don't take DDM numbers from cheap modules at face value. Even vendors that build their own optical diagnostics allow for a tolerance of around 3 dB on Tx and Rx and about 3 degrees on temperature - and that's for their own modules. On a third-party one you can easily get a couple of decibels of discrepancy from reality, and sometimes the module doesn't report anything meaningful at all, so there's no early power alarm in principle. The thresholds in DDM also come from the module itself (SFF-8472), the switch doesn't compute them.
And a work order that saves time: first compare Rx against the module's operating range, not against the fact that "the link is up", then replace the patch cord, then check the intermediate patch panels - has anyone been in there recently - and only after that suspect the far-end module. I once had a case where receive dropped to -25.69 dBm, and the culprit turned out to be the run, not the module.