Switching

Just trying to see if anyone else has experienced this before, but I am executing some tests in our lab which consists of a pair of EX-8208s in a Virtual Chassis to a pair of Cisco 7609s.  I am running ae interfaces to each 7609 across different 8208 member switches and everything is working very well, however in some cases I am noticing that LDP/BGP drops on the Cisco shortly after one of the two ae member links comes back online from a controlled line card restart as a result of an NSSU

 

If I manually fail over the XREs NSR/NSB works perfectly every time and manually restarting a line card that is member of ae doesn't seem to impact it.  The pattern that I see emerging though is it is always within 10-30 seconds after the the member link comes back online. 

 

I get the feeling there is a certain set of conditions that occur during the link coming up that is impeding the BGP/LDP control plane packets from reaching the RE which results in them dropping.  They automatically come back within about ~30 seconds, which makes me believe there is something to do with the period slow lacp pdus playing a role, whereby the link isn't fully in the ae until the first lacp pdu arrives and anything sent on the link is ignored until then.

 

Has anyone experienced similar issues with protocols dropping after an LACP member port comes back online?

 

Jerrad

Hi,

 

Have you enable the traceoptions for your switches?

May I see the records please.

Hi Jerrad_D. Have you check this forum? (it isn't same architecture, but it can help you). https://supportforums.cisco.com/thread/2111664 I think similar, it can be a LACP issue, being one of the side en slow mode or a STP problem on Cisco side (etherchannel guard). Can you check it? What kind of STP mode is ussing? Bye.

I identified the issue as a lack of "vrf-table-label" in my vrf's, when I originally read about this command it seemed to indicate it was required for devices without tunnel pics, as sort of software hack, however upon further investigation it appears that broadcast ethernet based MPLS should have this on all of the time as it assists with the MAC learning process.  Once I enabled it I never experienced the issue again, the unfortunate thing is without it present in the config you can still achieve a working LDP signaled MPLS L3 VPN, however things start to behave erratically when protocols restart etc...

 

Thanks for the feedback though