[c-nsp] Catalyst 3750-X with SFP-10G-ZR transceiver - troubleshooting‏

[Joe Crap]

On 10/14/2013 02:25 AM, Phil Meyers wrote: 

> Is there any chance you've blown the RX input on one or both xcvr? ZRs 
> typically can't tolerate bright light on RX, and if someone ever 
> connected them without an attenuator... 
> 
> As a general operational thing, I prefer fixed LC simplex attenuators to 
> attenuated patch leads, because you can then leave the attenuator 
> "attached" to the ZR RX input, and avoid any forgetfulness about the 
> nature of what you're doing. Attenuated patch leads OTOH are too easy to 
> forget about. 

No, they haven't been connected without the attenuators in place. We're
using the snap-on inline LC attenuators that you connect to the end of the
cable rather than attenuated leads, which is what I think you are 
recommending here?

Even if the transceivers were ever patched directly without attentuation, 
or lets say one of the attenuators was faulty, it shouldn't actually damage
 the receiver though, right? The Cisco specs
(http://www.cisco.com/en/US/prod/collateral/modules/ps5455/
data_sheet_c78-455693.html) show the damage threshold as +5dBm, which is
 higher than the max output power of the transmitter (4dBm). I don't know 
exact threshold for the 3rd party ones we're using, but I'd guess it 
mirrors the Cisco specs. And if that was to happen (lets say one of the 
attenuators was faulty) and the receiver was in some way damaged it surely
would also break the DOM reporting of the signal strength? Ours is showing
 an Rx signal nicely in the middle of the allowed range.

> The other thing you can try is substituting an LR for ZR at one end; the 
> receivers in the xcvr are normally wideband and in theory a 1310 and 
> 1550 optic can link-up (although the last time I tried this, I think it 
> failed). Obviously you'll need attenuate asymmetrically for this to work 
> - the ZR will likely overdrive the LR RX, and vice-versa, but by 
> different values (~3 and ~7dB, respectively). Again simplex attenuators 
> win here! 

Could try this, though a negative result wouldn't be very helpful here.
What about trying a loopback of a single cable from the Tx to the Rx of
the same transceiver (attenuated obviously) - would that be a valid test
of a single transceiver or is there some reason why this shouldn't work 
due to STP issues, etc.?

> Of course it's also possible that Cisco have messed up in IOS; do you 
> have another device (another vendor?) you can put the xcvr in to test? 

I'm afraid not - we're talking to our reseller to arrange testing inhouse
 with some Cisco-branded transceivers. My worry is that they will work 
fine, and that Cisco have in some way "disabled" 3rd-party interop with 
this new ZR support in the IOS. And I haven't been able to find any 
refurbed Cisco-branded ZRs for sale anywhere :-P

> If it's an xcvr fault, and only one is dead, you might be able to 
> determine which one by disabling autoneg - if one end comes up and the 
> other doesn't, the link-down end has dead RX (or far-end dead TX). 

We've removed all config from the 10GbE interfaces so they are running 
with IOS defaults, including the autoneg. I've tried forcing that to 
10GbE and it doesn't make a difference to the port state. We have MST 
enabled at a switch level, but the ports aren't marked as trunks. 

Thanks for your suggestions so far!

- J