Bonded network problem - needs single connection to work?

1. offizieller Beitrag

    I have a twin intel nic card in my server as well as the nic on the mobo. Up till now the mobo nic has been used.


    I tried to set up bonding on the other two today. Seems simple enough I set it too 802.3 on the switch I aggregated two ports and hooked it all up, lights are flashing green seems to be working.


    I want all traffic to go through the bond so removed the single ethernet. Now it does not work, still flashing green but cannot get to the GUI or samba etc. When I plug the single nic back in it all works again leading me to believe the bonds is passing through the single NIC.


    What have I done wrong?

    HI,


    The single nic has ip ending 15


    The twin nic I have set to bond has ip ending 16


    If 15 is not plugged in, 16 does not work. This has me much confuse!

    Hi.


    I just went into network and created new>bond


    Here are the outputs


    I am not sure I understand?


    The server gets hit pretty hard for much of the day I was aiming to double the link speed. Whilst I appriciate this does not double things it seems to help for aggregation on a pair of switches I have where by with one link the throughput seems to go up and down and with bonded that throughput flatlines at peek rate with no peeks and troughs


    Was hoping for the same with OMV.

    Zitat von garyi

    I was aiming to double the link speed


    That's not how bonding/LACP works. Link speed remains the same and only if you've a bunch of clients and the appropriate algos chosen on switch and the Linux host many clients accessing concurrently won't saturate the same link but end up on different links (link speed always being the same). I've seen even large companies where exactly this did not work since the Cisco certified admins did not chose the correct algos and all clients behind the EtcherChannels (cisco speak) ended up on the same link (check xmit_hash_policy option on Linux)


    Most probably it's just a waste of resources and energy. But back to your problem: When you let a few NICs accessing the same IP subnet with addresses within the same range (same netmask) what you saw is quite common. Once the primary interface is down nothing seems to work any more (due to ARP).


    Most probably it will work out of the box with eth0 disconnected and Linux host and switch restarted at the same time. But as already said: you can not double the link speed with LACP, that's what more intelligent and modern approaches do on different layers (e.g. SMB Multichannel)

    • Offizieller Beitrag

    I tried adding a pair of nics in a vm and making a bond. When hitting apply I get errors. Did you get errors then? Even with the error I can ping and get to the webui. Can you ping the ip of the bond from the omv? Or another machine?


    Will look at the error later.
    root@OMV4:~# systemctl status networking.service● networking.service - Raise network interfaces Loaded: loaded (/lib/systemd/system/networking.service; enabled; vendor preset: enabled) Active: failed (Result: exit-code) since Sun 2018-08-26 08:03:15 MDT; 2min 43s ago Docs: man:interfaces(5) Process: 533 ExecStart=/sbin/ifup -a --read-environment (code=exited, status=1/FAILURE) Process: 512 ExecStartPre=/bin/sh -c [ "$CONFIGURE_INTERFACES" != "no" ] && [ -n "$(ifquery --read-environment --list --exclude=lo)" ] && udevadm settle (code=exited, status=0/SUCCESS) Main PID: 533 (code=exited, status=1/FAILURE)

    Aug 26 08:03:12 OMV4 systemd[1]: Starting Raise network interfaces...Aug 26 08:03:13 OMV4 ifup[533]: ifup: waiting for lock on /run/network/ifstate.ens18Aug 26 08:03:15 OMV4 ifup[533]: RTNETLINK answers: File existsAug 26 08:03:15 OMV4 ifup[533]: ifup: failed to bring up bond0Aug 26 08:03:15 OMV4 systemd[1]: networking.service: Main process exited, code=exited, status=1/FAILUREAug 26 08:03:15 OMV4 systemd[1]: Failed to start Raise network interfaces.Aug 26 08:03:15 OMV4 systemd[1]: networking.service: Unit entered failed state.Aug 26 08:03:15 OMV4 systemd[1]: networking.service: Failed with result 'exit-code'.


    What is your output of systemctl status networking.service


    What bond type did you choose? 802.3ad?

    Yes I went with 802


    I have unplugged the bonding cables for now as the server is not using them with the single cable on the other nic so not sure if the output says anything



    tkasier are you saying restart the server? To be honest I have not tried that lol.

    • Offizieller Beitrag
    Zitat von garyi

    Device "bond0" does not exist

    What does cat /etc/network/interfaces look like.


    # Include additional interface stanzas.
    source-directory interfaces.d



    # The loopback network interface
    auto lo
    iface lo inet loopback



    # ens18 network interface
    auto ens18
    allow-hotplug ens18
    iface ens18 inet dhcp
    dns-search example.com
    iface ens18 inet6 manual
    pre-down ip -6 addr flush dev $IFACE



    # bond0 network interface
    auto bond0
    iface bond0 inet static
    address 10.0.83.220
    gateway 10.0.83.1
    netmask 255.255.255.0
    mtu 1500
    dns-nameservers 10.0.83.215
    dns-search example.com
    bond-slaves ens19 ens20
    bond-mode 4
    bond-miimon 100
    bond-downdelay 200
    bond-updelay 200
    iface bond0 inet6 manual
    pre-down ip -6 addr flush dev $IFACE

    Zitat von tkaiser

    I tried to explain the problem and why LACP won't work when it's about to 'double the link speed'. To no avail obviously.

    Obviously. I am not entirely sure why you are being shirty with me, if it helps this is all just a learning experience for me, I am not an expert or even a capable novice, sorry if I am not up to your standard. It can come over as aloof, you know like linux people are constantly criticised for.


    If none of this make any difference whats the point of having it in OMV?

    Zitat von garyi

    If none of this make any difference whats the point of having it in OMV?


    Again: 'if you've a bunch of clients and the appropriate algos chosen on switch and the Linux host many clients accessing concurrently won't saturate the same link but end up on different links (link speed always being the same)'


    Such installations (many clients) exist and some of them use OMV. Also LACP provides redundancy.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!

Benutzerkonto erstellen Anmelden