1

I am currently trying to implement redundancy whereby hosts have 2 eth ports and both ports connect to different switches. If either switch fails (or if any port fails) there will still be a way to reach the other host. The diagram below is adapted from bonding.txt section 11.2 in the linux drivers documentation

                |                                     |
                |port3                           port3|
          +-----+----+                          +-----+----+
          |          |port2       ISL      port2|          |
          | switch A +--------------------------+ switch B |
          |          |                          |          |
          +-----+----+                          +-----++---+
                |port1                           port1|
                |             +-------+               |
                +--------eth0-+ host1 +-eth1----------+
                |             +-------+               |
                |                                     |
                |             +-------+               |
                +--------eth0-+ host2 +-eth1----------+
                |             +-------+               |

Generalizable Questions

  • Does eth0 and eth1 of host1 have the same IP & MAC address?
  • Are packets from host1 to host2 duplicated?
    • If yes how would host2 know to ignore the duplicate packet?
  • If link availability checks are done instead of duplicate packets how would host1 know to stop using eth0 and start using eth1?
  • What type of switch configuration would be needed?
6
  • Unfortunately, questions about host/server configurations are off-topic here. You could try to ask this question on Server Fault for a business network.
    – Ron Maupin
    Apr 20, 2020 at 17:18
  • 1
    This is definitely more of a network engineering question than a server fault question because it's about how the network (i.e. switches and devices) transmit and handle the data. Redundancy IS a component of networks. Please reopen.
    – philn
    Apr 20, 2020 at 19:44
  • There are a lot of options. It does depend somewhat on your switches, too. If you have switches which support a stacking feature with cross-stack LACP, that will be the easiest way to configure the switches & servers. In the absence of stacking (or similar MLAG features) things get more complex and are more likely to require a routing protocol or application-awareness. Apr 20, 2020 at 20:12
  • If you really feel that this is something for NE, then you need to provide some information about the network devices, such as the models and configurations. The only thing you referred to was the documentation on the Linux drivers, and that is off-topic here. How you configure the hosts is off-topic here. We may be able to help with the configuration of the network devices, but we cannot simply guess what models and configurations. You can refer to the Network Engineering Question Checklist for guidance, then edit your question.
    – Ron Maupin
    Apr 20, 2020 at 22:35
  • Also, your questions about the hosts: "Does eth0 and eth1 of host1 have the same IP & MAC address?" and "If yes how would host2 know to ignore the duplicate packet?" and "If link availability checks are done instead of duplicate packets how would host1 know to stop using eth0 and start using eth1?" are off-topic here, so you should remove those when you edit the question.
    – Ron Maupin
    Apr 20, 2020 at 22:37

0

Browse other questions tagged or ask your own question.