Turning up a new network segment on a newly installed/certified SMF plant, I had the option to either aggregate a couple switches at an intermediate switch, or home-run all of them directly to the site's core router. I opted for the latter.
One of the three switches home-runs directly to the core on a 24-strand SMF. No problems.
The other two, however, follow cabling paths that jump one or two intermediate fiber runs, before joining that same 24-strand back to the core.
Link lights came right up. CDP/LLDP shows neighbors. CAM tables populated. Looks normal so far.
Then ping gateway from switch: first 5 succeed, then intermittent failures, like this: !.!..! or !.... and finally ..... And, not just ping: snmp, ssh, etc. all fail as well.
ARP/CAM looks good on both ends. Links are stable--not flapping. No incrementing interface errors on either end. No errors in logs. sh int transceiver detail shows tx/rx power within spec, on both ends.
Finally I decide to aggregate the two problematic switches through the one that worked. This eliminated one fiber jump for both problematic switches. They came up without issue. Ping/ssh/snmp all succeed, everything's normal.
I'm calling the cable installer to re-certify the pairs in question. But I'm wondering if I missed something.
I know about insertion loss, and that for every fiber connection I lose some dB of signal. But I would have expected a problem like that to show up as a flapping interface, or low tx/rx power, or incrementing interface error counters, or something. I saw none of the usual suspects for a L1 problem. Is there a limit to the number of passive jumpers between active nodes?
This is on Cisco 3850/3650 kit with Cisco SMF-LR optics on both ends.
No comments:
Post a Comment