DX NetOps

 View Only

Expand all | Collapse all

Alarms in Dx netops for CRC_STOMPED and NON- STOMPED packets on ingress and egress interfaces - FEX connected ports

  • 1.  Alarms in Dx netops for CRC_STOMPED and NON- STOMPED packets on ingress and egress interfaces - FEX connected ports

    Posted Sep 03, 2024 09:59 AM

    Hi,

    Can we report alarms for CRC_STOMPED and NON- STOMPED packets on ingress and egress interfaces - FEX connected ports in DX netops.

    Not only for  Cisco Nexus 7k and 5k or 2k models, for cross platforms like Arista DMF switches, how these errors can be monitored through Dx netops tools 

    Any suggestions and ideas with scenarios , highly appreciated 



  • 2.  RE: Alarms in Dx netops for CRC_STOMPED and NON- STOMPED packets on ingress and egress interfaces - FEX connected ports

    Broadcom Employee
    Posted Sep 04, 2024 08:24 AM

    I think that you need to know the OID corresponding to CRCs so you can do a self-certification to get this information from device and generate the alarm. 



    ------------------------------
    Thank you,
    Carlos A. Radice
    Sr Support Engineer
    Broadcom Software
    ------------------------------



  • 3.  RE: Alarms in Dx netops for CRC_STOMPED and NON- STOMPED packets on ingress and egress interfaces - FEX connected ports

    Broadcom Employee
    Posted Sep 05, 2024 09:24 AM

    As Carlos said, identifying the correct OID values for these metrics is required.  CRC stomped frames are specific to cut through switches to identify frames that it recognizes as corrupt but already began forwarding because they're cut through.  Store and forward switches would have dropped these frames before forwarding.  The Alternate Interface metric family will report on CRC errors with the Ethernet w/DOT3 vendor certification (dot3StatsAlignmentErrors+dot3StatsFCSErrors ) but I don't think that differentiates between stomped and non stomped.  There are also Syslog messages generated for CRC errors : Identify and Trace Nexus 9000 Cloud Scale ASIC CRC

    Cisco remove preview
    Identify and Trace Nexus 9000 Cloud Scale ASIC CRC
    This document describes the steps used to trace the source of CRC errors observed on Cisco Nexus 9000 Cloud Scale ASIC modules.
    View this on Cisco >

    It's not clear to me that there's a distinction between stomped or not.  The Cisco document walks through a 4 step process to go from the syslog message to the conclusion so I suspect not.  

    Looking around at some scripts ( ACI-CRC-FCS-Checker/ACI_CRC_Poller.py at main · CiscoDevNet/ACI-CRC-FCS-Checker )

    GitHub remove preview
    ACI-CRC-FCS-Checker/ACI_CRC_Poller.py at main · CiscoDevNet/ACI-CRC-FCS-Checker
    Contribute to CiscoDevNet/ACI-CRC-FCS-Checker development by creating an account on GitHub.
    View this on GitHub >

    that Cisco has posted to GitHub, it looks like they're polling RMON Dot3Stats FCSErrors and RMON EtherStats CRCAlignErrors then comparing if the values are the same for the time period.  Those values should be available via SNMP as well but I don't see them included in the out of the box certifications.  I would start with seeing if you can use the Alternate Interface metric family on these interfaces if you're not already.  At least you'll have some insight into physical layer problems and can dig in from there.

    -Rob