Description

Traffic imbalance could occur on the aggregated interfaces, and whole PFE traffic could be affected, if the LACP interface experience aggressive link flaps with 100ms interval. 

Symptoms


Solution

When link flaps repeatedly within a very short time frame  on 100G interface on MPC10E line card supported platforms(MX240,MX480,MX960,MX2010,MX2020), traffic stops egressing the affected interface and report syslog messages during link down event. When there are continous  flaps and if those flaps are very fast under 1 second and continuous then this issue will be seen.

The workaround is to reboot the affected FPC.

 

This issue is fixed via below PR:

https://prsearch.juniper.net/problemreport/PR1719682

Affected Platforms: all-evo:all-junos:mx/x

Root-Cause: 1. Issue halt/disengage bits need not be enabled/disabled during link flaps as experiments point      to the Tx credits loss because of that. 2. Performing Tx flush enable instantly on link down event no matter link up processing is finished or not.

Modification History

2023-10-31 : Article Created

Related Information

This issue is tracked on the internal PR1719682: https://gnats.juniper.net/web/default/1719682#fixinfo_tab