Troubleshoot LAG/MLAG LACP-PDU Rate Issues

Version 4

    This post discusses LAG/MLAG configured with LACP LLDP-PDU rate different than the server configuration (Linux and Windows OS).

     

    References

     

    Overview

    LACP has two speed operations:

    • Fast mode - every 1 second :LACP-PDU is sent
    • Slow mode - every 30 seconds LACP-PDU is sent

     

    By default, Mellanox switches transmit LACPDU every 30 seconds (slow mode). However, many server vendors will require LACP-PDU every second the fast mode.

    This would lead to a mismatch in the expectations when the server and switch configured with LAG or MLAG.

     

    How to Tell if There is LACP Rate Mismatch

    From the switch side, you can run the command "show lacp counters". From the output, you can see if the switch transmit LACPDU packet number is lower than that which it receives.

    If so, that means that the server side LACP is running in fast mode and the switch itself is running in slow mode.

    For example:

    switch (config) # show lacp counters

     

    LACPDUs     Marker     Marker Response        LACPDUs

    Port      Sent Recv      Sent Recv           Sent Recv        Illegal    Unknown

    -----------------------------------------------------------------------------

    Mlag-port-channel: 16

    ------------------

    1/16       0   0           0   0             3646 128            0       0 

    Mlag-port-channel: 17

    ------------------

    1/17       0   0           0   0             4172 144            0       0 

     

    It is recommended that both ends have the same LACP-PDU speed configuration. So you can either change the switch side to fast mode or the server side to slow mode.

     

    Solution: Changing the LACP-PDUs Rate

    On the switch side, to change the LACP-PCU rate to fast mode, run:

    switch (config) # interface ethernet 1/1 lacp rate fast

    This will change the interface's LACP mode to fast.

     

    If the server's OS is Microsoft Windows, refer to microsoft KB to modify the rate to slow mode https://support.microsoft.com/en-us/kb/3109099

     

    If server OS is Linux, you need to change the bond configuration file. Change bond-lacp-rate 1 (fast rate) to bond-lacp-rate 0 (slow rate).

    Here are some examples: