Header Only - DO NOT REMOVE - Extreme Networks

CRCs on ISL port


Hello,

A fabric of six VDXs 6740T-1G running NOS 6.0.2b.

We started observing CRCs and receive errors on one of the ISL links.

Receive Statistics:
60102804042 packets, 51259679161172 bytes
Unicasts: 59225527043, Multicasts: 874747487, Broadcasts: 0
64-byte pkts: 381, Over 64-byte pkts: 18462668286, Over 127-byte pkts: 6404289237
Over 255-byte pkts: 3351627659, Over 511-byte pkts: 2901158210, Over 1023-byte pkts: 1241512515
Over 1518-byte pkts(Jumbo): 27741541072
Runts: 0, Jabbers: 0, CRC: 2373642, Overruns: 0
Errors: 162197, Discards: 21813, TrillportCtrlFrames: 53671211[/code]The fiber carrier (since CRCs are present) normally asked us to replace the adapter and check the cable, but the link was CRCs free for quite some time, so I was wondering is there anything else we can try from the switch side?

The ISL port is connected to the 40G slot (split into 4x10G) via Mellanox adapter. The SFP+ adapter is a Brocade one:
Interface TenGigabitEthernet 79/0/52:1
Identifier 3 SFP
Connector 7 LC
Transceiver 0000000000000010 10_GB/s
Name sw
Encoding 6
Baud Rate 103 (units 100 megabaud)
Length 9u 0 (units km)
Length 9u 0 (units 100 meters)
Length 50u 8 (units 10 meters)
Length 62.5u 3 (units 10 meters)
Length Cu 0 (units 1 meter)
Vendor Name BROCADE
Vendor OUI 00:05:1e
Vendor PN 57-0000075-01
Vendor Rev A
Wavelength 850 (units nm)
Options 001a
BR Max 0
BR Min 0
Serial No XXXXXXXXXXXXX
Date Code 170529
Optical Monitor yes
Temperature 31 Centigrade
Voltage 3277.2 (mVolts)
Current 8.722 (mAmps)
TX Power 595.6 (uWatts)
RX Power 556.3 (uWatts)[/code]Many thanks in advance.

3 replies

Userlevel 6
CRC are always a layer one issue... Optics do go bad some time on the receive side.

Most of the time it is a degradation of the light. You said fiber carrier so are these all in different location and you have a 3rd party providing you fiber for the ISL links?.. The optic you listed seems to be a MM SR which would only be a local link?

From my experience CRC are almost always fiber related. Patches got dirty. Bend or kink in patch. PAtch is not seating in frame or optic correctly.

Out of a several thousands of optic transceivers we have only replaced a hand full where they just go out or start taking CRC's ... We have burnt out a few ER and ZR optics through the years by over saturating the led with an OTDR but that is a human error that can be prevented.

I did find a previous GTAC https://gtacknowledge.extremenetworks.com/articles/How_To/How-to-troubleshoot-CRC-errors-or-general-...
Thank you!
That's correct: there is a 3rd party providing 10G fiber between our data centers with the distance ~5km. From our switches it terminates directly to their OTN boxes (Huawei OptiX OSN 1800), so the requirement from their end was a 10G MM SR.
They also claim that they see CRCs on both ends, but we only see CRCs on one of the VDXs, the other side is CRC free (at least 'sh int' shows 0 CRCs).
Userlevel 6
Have you got optical readings from your links and theirs?? That would be the place to start. I would suggest cleaning/replacing patches first and if that does not clear it up then optics. I would guess patches since you both see CRC's .. I would double check the bend radius along the path and especially at the connectors to make sure that the weight over time has not bent the patch at the connector causing the problem. Is the mm patch ten gig rated comes to mind too...

Reply