LACP load balancing

[Wil]

Hi

I have 2 1Gb ethernet ports in LACP/802.3ad to an Ubuntu server. The Ubuntu server has LACP/802.3ad configured and is able to transmit outbound at ~1.8Gb/s successfully, but, after trying many options, it can only receive at 1Gb/s.

After reading through many posts on this forum, I can tell that my server is only receiving data on one of the interfaces (whichever one the 7750 sets as primary).

Does anyone know how you change the hashing mechanism so that the SR will send on both interfaces?

Thanks

[mivens]

It depends the nature of your traffic but if you haven't enabled it, you might like to try

configure system l4-load-balancing

Take a look at the section "LAG and ECMP Hashing" in the SR OS Interface Configuration Guide for how the hashing works for various types of traffic and the various options available.

[thekotaksampah]

Hello guys,

I have different problem. when I use LAG with configuration like this

LACP.JPG

The lag utilization is unbalanced, when the one port is high until 100%, and other port still ~0.00%. I use ostinato to generate traffic and then I inject the traffic using access port (SAP) from VPLS.

I using command lsr-load-balancing lbl-ip but it's not effect with lag.

Code: Select all

#--------------------------------------------------
echo "System Configuration"
#--------------------------------------------------
        name "LAB-A"
        no contact
        no location
        no clli-code
        no coordinates
        chassis-mode c
        mixed-mode
        no config-backup
        dhcp6
            no adv-noaddrs-global
        exit
        dns
            address-pref ipv4-only
        exit
        ethernet
            no new-qinq-untagged-sap
        exit
        no boot-good-exec
        no boot-bad-exec
        no switchover-exec
        power-supply 1 default
        power-supply 2 default
        no l4-load-balancing
        lsr-load-balancing lbl-ip
        no system-ip-load-balancing
        no service-id-lag-hashing
        lacp-system-priority 32768 
        no enable-icmp-vse
        ptp

*A:LAB-A# monitor lag 10 interval 3 rate 

===============================================================================
Monitor statistics for LAG ID 10
===============================================================================
Port-id   Input       Input       Output      Output      Input      Output
          Bytes       Packets     Bytes       Packets     Errors     Errors
-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
At time t = 0 sec (Base Statistics)
-------------------------------------------------------------------------------
4/2/11    96520836    920131      38378355831 37756796    12         0
4/2/13    41032199    386707      75797327056 145551720   2          0
-------------------------------------------------------------------------------
Totals    137553035   1306838     1141756828* 183308516   14         0

-------------------------------------------------------------------------------
At time t = 3 sec (Mode: Rate)
-------------------------------------------------------------------------------
4/2/11    0           0           171         1           0          0
  % Util  0.00        --          ~0.00       --          --         --
4/2/13    202         2           12307653    9618        0          0
  % Util  ~0.00       --          100.00      --          --         --
-------------------------------------------------------------------------------
Totals    202         2           12307824    9619        0          0
  % Util  ~0.00       --          50.00       --          --         --

-------------------------------------------------------------------------------
At time t = 6 sec (Mode: Rate)
-------------------------------------------------------------------------------
4/2/11    0           0           128         1           0          0
  % Util  0.00        --          ~0.00       --          --         --
4/2/13    313         3           12308059    9600        0          0
  % Util  ~0.00       --          100.00      --          --         --
-------------------------------------------------------------------------------
Totals    313         3           12308187    9601        0          0
  % Util  ~0.00       --          50.00       --          --         --

When I Use command, l4-load-balancing, it's also can't give effect to lag

Code: Select all

#--------------------------------------------------
echo "System Configuration"
#--------------------------------------------------
        name "LAB-A"
        no contact
        no location
        no clli-code
        no coordinates
        chassis-mode c
        mixed-mode
        no config-backup
        dhcp6
            no adv-noaddrs-global
        exit
        dns
            address-pref ipv4-only
        exit
        ethernet
            no new-qinq-untagged-sap
        exit
        no boot-good-exec
        no boot-bad-exec
        no switchover-exec
        power-supply 1 default
        power-supply 2 default
        l4-load-balancing
        no lsr-load-balancing
        no system-ip-load-balancing
        no service-id-lag-hashing
        lacp-system-priority 32768 
        no enable-icmp-vse


*A:LAB-A# monitor lag 10 interval 3 rate 

===============================================================================
Monitor statistics for LAG ID 10
===============================================================================
Port-id   Input       Input       Output      Output      Input      Output
          Bytes       Packets     Bytes       Packets     Errors     Errors
-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
At time t = 0 sec (Base Statistics)
-------------------------------------------------------------------------------
4/2/11    96520836    920131      38378368247 37756893    12         0
4/2/13    41065954    387095      76794404182 146323058   2          0
-------------------------------------------------------------------------------
Totals    137586790   1307226     1151727724* 184079951   14         0

-------------------------------------------------------------------------------
At time t = 3 sec (Mode: Rate)
-------------------------------------------------------------------------------
4/2/11    0           0           128         1           0          0
  % Util  0.00        --          ~0.00       --          --         --
4/2/13    583         8           12308646    9569        0          0
  % Util  ~0.00       --          100.00      --          --         --
-------------------------------------------------------------------------------
Totals    583         8           12308774    9570        0          0
  % Util  ~0.00       --          50.00       --          --         --

-------------------------------------------------------------------------------
At time t = 6 sec (Mode: Rate)
-------------------------------------------------------------------------------
4/2/11    0           0           128         1           0          0
  % Util  0.00        --          ~0.00       --          --         --
4/2/13    707         9           12307707    9604        0          0
  % Util  ~0.00       --          99.99       --          --         --
-------------------------------------------------------------------------------
Totals    707         9           12307835    9605        0          0
  % Util  ~0.00       --          49.99       --          --         --

Any advice from this event, Hope you can help me guys.

Thanks

[jeniver]

Hi!

Few questions:
On the drawing I see ports 4/2/11 and 4/2/12, but your LAG is 4/2/11 and 4/2/13. Where is typo?
How many IP flows do you send?

I also see that you send traffic in one direction. Better to send a small stream with reverted mac addresses to be sure that you traffic is treated as unicast, otherwise it will be broadcasted via multipoint queue.

[thekotaksampah]

Hi jeniver,

the port from pict is typo. I'm sorry.

Additional information. I stream 2 IP flow. this the update from my activity. Each stream I using 1000 pps. Each stream I using pair of MAC, so the traffic treated as unicast traffic as you said jen.


This is the result's capture, LAG still unbalanced

Code: Select all

*A:LAB-A# monitor lag 10 interval 3 rate repeat 1 

===============================================================================
Monitor statistics for LAG ID 10
===============================================================================
Port-id   Input       Input       Output      Output      Input      Output
          Bytes       Packets     Bytes       Packets     Errors     Errors
-------------------------------------------------------------------------------

At time t = 3 sec (Mode: Rate)
-------------------------------------------------------------------------------
4/2/11    675         8           236149      2001        0          0
  % Util  0.06        --          22.09       --          --         --
4/2/13    236149      2001        1106        7           0          0
  % Util  22.09       --          0.09        --          --         --
-------------------------------------------------------------------------------
Totals    236824      2009        237255      2008        0          0
  % Util  11.08       --          11.09       --          --         --

===============================================================================

[jeniver]

Hello!

If you send one flow in each direction how do you expect to see a balancing? Internal hashing/balancing works on a flow basis. To make balancing statistically meaningful you have to send 100s of flows in one direction.