So I was recently gifted some Mellanox 40gig network cards that I installed in my NAS and my desktop and connected with AOC fiber. I gave them both static IP addresses on their own dedicated subnet that’s not used anywhere else in my network. I was able to run iperf3 between both computers, and that worked exactly as expected.

At that point, I edited /etc/fstab to update the IP addresses for my mounted network shares. I ran # mount -a successfully and thought all was well.

The problem is, my computer defaults to my one gig lan connection for some reason, despite the entries in fstab using a completely different subnet.

The only way I’ve found to force it to work properly is to disable my LAN connection, then remount the network shares, then reenable the LAN port.

On one occasion I noticed that a file I was duplicating on my NAS was being downloaded via my LAN to my computer to duplicate, then being uploaded back to the NAS via the fiber connection.

Does anyone have any clue why this may be happening or how to fix it more permanently?

The NAS is Debian, my desktop is Manjaro.

  • @[email protected]
    link
    fedilink
    14
    edit-2
    1 year ago

    I was recently gifted some Mellanox 40gig network cards

    Where can I find such friends/relatives?

    • Transient PunkOP
      link
      fedilink
      English
      8
      edit-2
      1 year ago

      Marry way out of your league 😉

      That’s what I did at least

      • metaStatic
        link
        fedilink
        11 year ago

        I’m already at the lowest league unless you count people who marry their animu body pillow.

  • 30021190
    link
    fedilink
    91 year ago

    Are you sure you’re not using local hostnames / DNS resolution?

    You may also need to updates your nfs exports file for the new subnet. And also update systems on the fstab changes (daemon-reload).

    • Transient PunkOP
      link
      fedilink
      English
      11 year ago

      I did run a daemon-reload. I’m directly referencing the IP address in my fstab. On my desktop I have DNS turned off for the Mellanox interface.

      I’ll have to look into the nfs exports file. I’m using OMV, and I just have my exports wide open. But, I guess I could solve this by limiting the connection to the mounts by client IP.

      • 30021190
        link
        fedilink
        11 year ago

        There might also be some magic/weirdness with IP routing in the kernel. Have a look at net.ipv4.ip_forward system variable.

  • Gray
    link
    English
    61 year ago

    Sounds like route tables are finding a priority match on the 1gb interface. Are you 100% sure the NAS connection is truly not an overlapping subnet with the 1gb nic?

    • Transient PunkOP
      link
      fedilink
      English
      3
      edit-2
      1 year ago

      100% sure.

      LAN: 192.168.69.0/24

      Fiber: 10.42.69.0/24

      Also: yeah, yeah. I know…

      • @[email protected]
        link
        fedilink
        English
        101 year ago

        you need to look at the routing tables on your computer. these tables store the prioritized rules for how packets leave your host machine.

        it might be that something is adding rules, or, there is some overly broad rule taking priority (like a rule that says all 10.0.0.0/8 traffic go to your home router over 192.168.69.0/24, etc)

        it’s also suspect that you can reach the NAS over the 1gb card. That to me means one of two things:

        1. something is not actually using the IP you’ve configured in your fstab and is using some IP that is on the 1gb interface
        2. you have some weird network routes configured that is leading to this issue. if 10.42.69.0/24 is accessible over the 192.168.69.0/24 network, then you might need to create a static route explicitly telling your OS to send packets out the 40gb card

        ultimately, i suggest you run something like tcpdump or wireshark on your computer (ideally on the NAS too) so you can start to visualize how the packets are being addressed and transferred over your networks.

        sincerely, a fellow 10.0.69.0/24 enjoyer

        • Natanael
          link
          fedilink
          21 year ago

          It definitely sounds like he has redundant routes (this literally shouldn’t be possible otherwise) so yeah, he needs to fix priority

        • @[email protected]
          link
          fedilink
          21 year ago

          A bad route would be the first thing I’d check, too.

          It sounds like u/[email protected] is already pretty familiar with the Linux command line, but just in case, you can check the routing table (on both the NAS and the client machine) with ip route (that should show the whole routing table) and get the specific route to the remote device with ip route get 10.42.69.XXX (change the XXX to make it the address of the remote system). If either side shows that the route is going over your default gateway, that’s your problem.