NIC Teaming and Uplink Failure

Unanswered Question
Jun 10th, 2010
User Badges:
  • Blue, 1500 points or more

Hi:


The topic is NIC teaming for $500, Alex....


A server is dual-homed to 2 access switches in a NIC team. What happens when the access switch that is part of the active member of the team loses its uplink to the distro layer?  Imagine that is the only uplink. Will the server continue forwarding traffic to that switch and get blackholed? I imagine the answer is no and that the server NIC teaming software has some sort of biuilt in SLA mechanism...


Can anyone confirm and give some detail?


Thanks

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
Hitesh Vinzoda Thu, 06/10/2010 - 07:58
User Badges:
  • Silver, 250 points or more

The Teamed NIC's uses hearbeat to check the status, For example


NIC 1 is primary and NIC 2 is secondary they will exchange heartbeat using the network layer 2, i cant recall the multicast mac address for that, so when the switch connected to NIC 1 is down or cable is disconnected, the Team software detects the link loss and NIC2 is elected as primary


In your case, Where there is only one uplink towards the distribution switch from each switch where the teamed NIC are connected. and uplink between access and dist fails, resulting in loss of hearbeats between NIC1 and NIC2, also note that link loss has not occurred, due to this both the interface will try to get elected, result is failure.


there are few design consideration for Teaming when you want the protection to team in case of upstream link failures, by using the looped topology (square, triangle).



HTH


Hitesh Vinzoda


Pls rate useful posts

Giuseppe Larosa Thu, 06/10/2010 - 09:27
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Victor,

Hitesh is right a sort of heartbeat is used but this can use of a L2 communication or a link local 224.0.0.X multicast  (some SUN teaming does so for example, HP teaming uses L2 address )destination depending on the implementation in this way the server should be able to understand which NIC is still connected to the network and to detect indirect failure


in the case of HP teaming:



>>Network Fault Tolerance (NFT) is the foundation of HP ProLiant Network Adapter Teaming. In NFT mode, two to

eight ports are teamed together to operate as a single virtual network port. However, only one network port, the

Primary port, is used for both transmit and receive communication with the server. The remaining ports are

considered to be in stand-by (or secondary ports) and referred to as Non-Primary ports, and remain idle unless

the Primary port fails. All ports may transmit and receive heartbeats, including Non-Primary ports.


Also the switches CAM tables are updated using these messages



>>>Update Switch MAC table with team MAC address

The Primary Port must ensure that the switch has the team’s MAC address on the correct port, and that

the switch keeps the team’s MAC address in its MAC table [or Content Addressable Memory (CAM)

table]. The Switch MAC table update heartbeat is used with most team types when a failover occurs,

and a Non-Primary port takes over the Primary role.


Destination MAC address 03-00-C7-00-00-EE  this is a multicast MAC address but outside the range used for IPv4 multicast



Hope to help

Giuseppe

lamav Fri, 06/11/2010 - 15:08
User Badges:
  • Blue, 1500 points or more

Thanks guys

Actions

This Discussion