Network setting and behavior for teamd (load-balance and round robin mode)

Latest response

I found good explanation about runner setting on documentation but I would like to learn,

  1. how should you connect LAN cables to multiple switches
  2. how does a teamd daemon tell the switch which port to forward the packet to

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Networking_Guide/sec-Understanding_the_Network_Teaming_Daemon_and_the_Runners.html

Is there a area in documentation where it covers the idea of network topology for teamd setting?

Responses

I assume you've found that you need to configure an EtherChannel or Port-Channel or similar sort of interface on the switch for these modes.

how should you connect LAN cables to multiple switches?

Your switches will need to support Multi-Chassis Link Aggregation (MLAG). Cisco's version of this is called Virtual Port Channel (VPC) because it makes one Port-Channel interface out of multiple.

That's all done inside the switch and Teaming won't care, the team will just transmit out each slaves as per the runner's balancing algorithm.

how does a teamd daemon tell the switch which port to forward the packet to?

The switch controls its own transmit, neither Teaming or Bonding can influence the incoming load balancing algorithm.

We intentionally made the decision not to document specific switch configuration in the RHEL product documentation, such content risks being incorrect or outdated as we are not a network hardware vendor, nor do we wish to maintain an ever-growing collection of "recipes" for brands and models which we have no experience with.

One more point, you probably don't want to use the round-robin runner. See: What is the best bonding mode for TCP traffic such as NFS, ISCSI, CIFS, etc?

Hello Jamie.

Thanks for the detailed explanation. Very clear to understand.

May I ask for similar description with active-backup mode?

Active-backup doesn't require any switch config, everything is done on the Teaming or Bonding system. Configuring an EtherChannel with active-backup mode is incorrect.

The failover with active-backup is handled by Gratuitous ARP on failover updating the switch's MAC table, and by that GARP updating remote hosts ARP tables if the MAC address does not change (eg: bonding's fail_over_mac=active).

Thank you .

Close

Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.