etherchannel loadbalance algo not supported ?

[tommyboay]

Hi all,

In one of our datacenters, I have a server-farm distribution switch for HP blades : WS-CBS3120G-S.

It is connected by LACP etherchannel using 7 ports to my core switch.

Unfortunately, I noticed the load balancing algorithm to be pretty unefficient for a server-side portchannel :

agfr01esxswt1a#show etherchannel load-balance
EtherChannel Load-Balancing Configuration:
        src-mac

EtherChannel Load-Balancing Addresses Used Per-Protocol:
Non-IP: Source MAC address
  IPv4: Source MAC address
  IPv6: Source MAC address

I wanted to change this to src XOR dst mac addresses or at least dest mac addresses. The options seemed available :

agfr01esxswt1a(config)#port-channel load-balance ?
  dst-ip       Dst IP Addr
  dst-mac      Dst Mac Addr
  src-dst-ip   Src XOR Dst IP Addr
  src-dst-mac  Src XOR Dst Mac Addr
  src-ip       Src IP Addr
  src-mac      Src Mac Addr

But once set to any alternative option, no data passes through the port channel any longer. It is still up though.

Is there any obvious reason for that ? Current IOS version is cbs31x0-universal-mz.122-40.EX3.bin

Thanks.

Tom

[Satya Narra]

When you change the load-balancing algorithm, try clearing the mac table on that vlan and see if it works.

[tommyboay]

Thanks for the advice. I'll try this next week and mark your answer if it fixed the issue.

[tommyboay]

After quite some time spent on getting a maintenance window, I finally got the switches upgraded and LB algorithm modified. I took time to test the mac table clearing but it was a no go : I finally rebooted the whole stack to get it to work.

Problem : I now experience LOTS of overruns on the Po interfaces. Would this be due to the ASICS getting swamped by the extra calculation when moving from src-mac to src xor dst mac ?

[smogra]

If the capacity of the interface is exceeded, the frame that is  currently being received is dropped and the overrun counter is incremented. In the majority of cases, it indicates that the receiving  capability of the interface was exceeded. If possible, the rate that frames are coming should be controlled at the remote end of the connection.

Sweta

Please rate useful posts.

[tommyboay]

Hi,

Unfortunately, this is not possible : These are blade switches WS-CBS3120G-S offering connectivity to a server farm. My uplinks are getting flooded due to what I believe to be traffic burst from the servers to clients or backup servers. I'm already at the max bundle with 8 active interfaces. See here :

Port-channel1 is up, line protocol is up (connected)

....

  Full-duplex, 1000Mb/s, link type is auto, media type is unknown

  input flow-control is off, output flow-control is unsupported

  Members in this channel: Gi1/0/23 Gi1/0/24 Gi2/0/23 Gi2/0/24 Gi3/0/23 Gi3/0/24 Gi4/0/23 Gi4/0/24

  ARP type: ARPA, ARP Timeout 04:00:00

  Last input 00:00:01, output 00:00:02, output hang never

  Last clearing of "show interface" counters never

  Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 4164677

  Queueing strategy: fifo

  Output queue: 0/40 (size/max)

  30 second input rate 8167000 bits/sec, 3449 packets/sec

  30 second output rate 21989000 bits/sec, 3462 packets/sec

.... (All error / drop counters for input and output are null)

I have no QoS policy configured nor active on these devices but receiving over 100 000 000 transmit discards from SNMP polling. Strange thing is that I never had such values on IOS 10.2-40 EX3. I'm now on 12.2-58 SE1.

[Somasundaram Jayaraman]

Hi,

Can you check whether are we hitting the oversubscription issue?

I can see there are two ports in a single module (Members in this channel: Gi1/0/23 Gi1/0/24 Gi2/0/23 Gi2/0/24 Gi3/0/23 Gi3/0/24 Gi4/0/23 Gi4/0/24) a part of port-channel.

Could you try moving the interfaces to a different Asic group?

For example if the interfaces 1/0/23 & 1/0/24 belongs to same asic group, then the Asic will get over utilized

and then there is more probability of packets getting dropped.

So if you could move the ports say 1/0/1 & 1/0/23 it may be of different port-asic groups and see it helps.

I am not sure of the interface-port asic maps.. you may need to check the datasheet of the product to check the mappings.

Hope this helps

Cheers

Somu

Rate helpful posts.

[tommyboay]

Hi Somu,

Thanks for the hint. I wouldn't have thought about this at all. But I'm not sure myself how to identify existing ASIC groups on this model. I drilled through the technical bulletin of that model without finding the info :

http://www.cisco.com/en/US/prod/collateral/switches/ps6746/ps8742/ps8749/data_sheet_c78-439133.html

I have to add that I'm pretty limited on the interface choice : I have 4 SFP ports (x/0/19 - 22) and 4 copper interfaces (x/0/23 - 26). All other interfaces are provided inside the blade chassis. But I'm ready to try the SFP ports for a test.

Again, I'd like to thank all the people that took time into this thread.

Tom