Ciena 3930: must a 1GE SFP be Ciena-coded, or is matching the 1000BASE-LX spec enough
We pick up a 1GE carrier handoff on a Ciena 3930 and I am putting together a spares kit for it. The module quoted for that circuit is XCVR-A10Y31, a 1000BASE-LX/LC part: 1310 nm, 10 km single mode. Nothing exotic about it, and I have a drawer of generic optics with exactly that specification sitting three metres from the rack.
- Ciena 3930, 1GE fibre port facing the carrier
- in the port now: XCVR-A10Y31, reported as 1000BASE-LX/LC
- what I would rather stock: generic 1310 nm 10 km LX SFPs, same optical spec, no Ciena coding
What I looked at on the box:
port xcvr show
port xcvr show supportedThe supported output prints the transceiver models the running release knows about, and every entry there is a Ciena part number. The one in the port matches, fine. What that output does not tell me is whether the list is advisory or whether the platform does something to a module that is missing from it.
What I have already done:
- compared the specs line by line - wavelength, reach, connector and rate all match on the generic parts
- weent looking for one compatibility list covering the platforms we run, and found only per-platform, per-release articles
So which is it: does the module have to be one sold by Ciena, or is any SFP of the same optical specification fine as far as the box is concerned?
Comments 3
It is the coding, not the specification. Ciena certifies transceiivers by what the module reports about itself, so an optically identical LX part that is not coded for the platform is simply not in the certified list of the release you are running. Matching 1310 nm and 10 km buys you nothing at that layer.
The good news is that SAOS does not behave like IOS here. No err-disable, no refusal to bring the port up. A module outside the list comes up in the operational state UCTF - uncertified transceiver - and the port still links and passes traffic in that state. Three commands cover what you need:
The first one just confirms the port is really up. The second reads the module back at you - who built it, which part it claims to be, and the medium/connector line, which on a decent spare prints 1000BASE-LX/LC exactly as the Ciena part does. The third decides your question: it lists what the loaded release certifies, and anything outside it wears the UCTF marking.
Two caveats before you build the whole spares kit on this. The certified list is tied to the software release, so an upgrade can move it under you, and behaviour is not identical across the family, so verify on the actual model rather than on a sibling. And the obvious one: on a circuit the carrier hands you, an uncertified module is the first thing anybody points at when a ticket is opened, whatever the counters say. I would keep XCVR-A10Y31 on the handoff port and put the generic parts on internal links.
Tried it on a lab port before going anywhere near the handoff. Generic 1310 nm LX in, link up, and port xcvr show reads the third-party vendor string with 1000BASE-LX/LC as the medium, exactly as described. State shows UCTF, the port keeps forwarding, and nothing in the log looks like a complaint.
So it works, and I am still taking the compromise rather than the win: XCVR-A10Y31 stays on the carrier-facing port because that circuit is under contract, and the generic parts go into the internal links where an uncertified marking costs me nothing. Not the answer I was hoping for, but at least I now know what the box actually does.
Same picture on the older Carrier Ethernet boxes. A pile of decommissioned kit - an LE-311v and a 3911 - swallowed a WWP-branded 1000BASE-LX and a copper 1000BASE-T SFP that identified itself as a Cisco GLC-T, and neither box refused either module. Both came up uncertified and carried traffic. One difference worth knowing: the 3911 reported a distinct transceiver type per port, while the LE-311v showed no difference at all between the modules tried, so do not read too much into what a single platform prints.
On the spares-kit question, there is no single cross-platform matrix and you will lose an afternoon looking for one. Every list I have found covers one platform on one software train, so two boxes in the same rack can disagree because their releases differ. From the lists we keep for our own kit, XCVR-A10Y31 passes on the 39xx and 51xx boxes - 3926 and 3928 on one side, 5142, 5160, 5170 and 5171 on the other - and no further. The 6500 and the 8700 are packet-optical, their lists are their own, and nothing carries over from the switches. port xcvr show supported on the box in front of you, re-run after every upgrade, is the only answer that stays true.