0

I have two linux servers with two NICs each and two unmanaged switches. This is my preferred design:

Server A - eth0 -> Switch 1
Server A - eth1 -> Switch 2

Server B - eth0 -> Switch 1
Server B - eth1 -> Switch 2

Because I want clustering and high availability, I need to understand that if I use active-backup mode of bonding/teaming and Server-A eth0 goes down, will I still be able to communicate with my second server?

I believe that Server-A eth1 (which is on Switch 2) will assume an active role when eth0 (on Switch 1) goes down, but how it will communicate with Server-B which has eth0 in active mode and eth1 in backup?

Do I need any special approach like some kind of link aggregation b/w two switches?

3
  • 1
    Welcome Jaz .. your best bet is to start from what performance you want against a list of risks (broken cable, fuses on switch PSU, NIC crash, etc) and the required recovery times. Then you can figure out how to implement, with a list of tests to perform. There is no substitute for testing. Is there a reason you don't have STP links between the switches, or are they simply omitted from the description?
    – jonathanjo
    Oct 17 '18 at 12:33
  • Those are just two independent unmanaged switches.
    – Jaz
    Oct 17 '18 at 18:09
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can provide and accept your own answer.
    – Ron Maupin
    Dec 25 '18 at 9:43
2

With unmanaged switches you'll want to configure active-passive failover in NIC bonding. An active-active setup with round-robin or the like will mess up the switches' MAC tables.

Note that you've not built a cluster yet. You've got redundant links between the servers but there's no clustered service to anywhere else. If you connect your firewall to both switches then you may have a cluster.

Depending on failover mode on the firewall you can either

  • use L2 failover in a single segment/subnet - this requires an interconnect between the switches and a firewall port group without bridging; cannot (easily) provide host failover, only NICs and switches
  • use L3 failover in two distinct segments/subnets - this requires dynamic port forwarding (destination mapping/DNAT) depending on the host status; can also provide host failover
  • use a mixture of both: virtual IP addresses on the hosts with a IP-to-MAC mapping updated by a monitoring service - same requirements as the L2 setup; can provide host failover
4
  • I will be using pacemaker corosync for high availability. Do i really need firewall? I only want if eth0 goes down, then i should be able to communicate with other server with bonded NIC. I just don't understand how server-A eth1 will communicate with server-B?
    – Jaz
    Oct 17 '18 at 18:07
  • The "firewall" can also be a load balancer, of course. Currently, the servers can communicate across each switch (you're mostly in the "use L3 failover" scenario). In case of a NIC/link failure you'd need some concept still.
    – Zac67
    Oct 17 '18 at 18:13
  • I am still not getting a clear picture. 😦😯 what more logic do i need to put to make it work?
    – Jaz
    Oct 17 '18 at 18:25
  • If the Server1/eth0 link fails, Server2 will still try to talk to it through its eth0 and needs to fall back to eth1 - you're working with two separate segments/subnets as of your diagram.
    – Zac67
    Oct 17 '18 at 19:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.