[RESOLVED] Network config on 0.5 Sardukaar

  • Guys, do we know what Debian file or files control NICs? If we do, can we do a file comparison between 0.4 and 0.5 to see if there are differences. Are the kernal builds the same between 0.4 and 0.5? Also then possibly the files from 0.4 could be copied to a 0.5 system. But who else has this particular NIC down to the same firmware version to test it? Doing this is purely to try and troubleshoot the problem. The reason I say this is it sounds like it may somehow be a Debian issue rather than OMV. Like tekkbebe suggests maybe a backport.


    Could 0.5 be installed onto a Debian squeeze install? Yes I know OMV is built on it but I'm trying to find where the difference lies.


    This I believe is going to take some time and a bit of work to solve.


    About the 3Com 3C2000-T Gigabit Network Interface Card. It seems to be an older card, as the last user review was about seven years ago. An Amazon review is also seven years old and says the card was discontinued.


    Unfortunately the only one I can think of to help us determine where the issue lies is bysard. I can only hope he is willing to help us resolve this.

  • You don't need to focus on 3Com 3C2000-T Gigabit Network Interface Card. I'm experiencing same problem on planet cards and Intel E1000 (single, two-port, 4-port). I will remove old 3com cards anyway and I can understand that support for old hardware will eventually stop.

  • I will do this when I come home from work today. What i've done so far to test exactly what you asked for:


    1. Installed fresh 0.5 OMV
    2. copied interfaces to interfaces.old
    3. tried to create bond which resulted always in network crash
    4. compared the two files - and are exactly the same!


    Will post the file today.


    EDIT:


    Ok, I tested again by the steps described above. Last step is different then before. The files are not the same.


    Old interfaces file before bond creation:
    #########################################
    # The loopback network interface
    auto lo
    iface lo inet loopback
    iface lo inet6 loopback


    # eth0 network interface
    auto eth0
    allow-hotplug eth0
    iface eth0 inet static
    address 10.0.105.10
    gateway 10.0.105.2
    netmask 255.255.255.0
    dns-nameservers 10.0.105.2
    iface eth0 inet6 manual
    pre-down ip -6 addr flush dev eth0
    #########################################



    After bond creation:
    #########################################
    # The loopback network interface
    auto lo
    iface lo inet loopback
    iface lo inet6 loopback


    # bond0 network interface
    auto bond0
    iface bond0 inet static
    address 10.0.105.10
    gateway 10.0.105.2
    netmask 255.255.255.0
    dns-nameservers 10.0.105.2
    bond-slaves eth0 eth1
    bond-primary eth0
    bond-mode 0
    bond-miimon 100
    bond-downdelay 200
    bond-updelay 200
    dns-nameservers 10.0.105.2
    bond-slaves eth0 eth1
    bond-primary eth0
    bond-mode 0
    bond-miimon 100
    bond-downdelay 200
    bond-updelay 200
    iface bond0 inet6 manual
    pre-down ip -6 addr flush dev bond0
    #########################################


    After the page has become inaccessible, I have rebooted OMV from console. Samba and NFS fail to start (because there is no valid ip assigned).
    Also I get this error when booting:


    Reconfiguring network interfaces.../etc/network/interfaces:19: duplicate option
    ifdown: couldn't read interfaces file "/etc/network/interfaces"
    monit: Cannot connect to the monit daemon. Did you start it with http support?
    monit: Cannot connect to the monit daemon. Did you start it with http support?
    monit: Cannot connect to the monit daemon. Did you start it with http support?
    monit: Cannot connect to the monit daemon. Did you start it with http support?
    monit: Cannot connect to the monit daemon. Did you start it with http support?
    monit: Cannot connect to the monit daemon. Did you start it with http support?


    After some 10 to 15 minutes after last monit error I was able to login. I ran ifconfig at it returns nothing, not even lo0. omv-first-aid command not found so I cannot repair it.
    The file /etc/network/interfaces is there and as user root I can access it. I forgot to check for "interfaces" file permissions before and after bond creation as I have already re-installed OMV. :(


    EDIT2:


    Please disregard my stupidity in typing omv-firstaid. Its works, I've tried again, It repairs the network.

  • The following lines are duplicated;

    Code
    dns-nameservers 10.0.105.2
    bond-slaves eth0 eth1
    bond-primary eth0
    bond-mode 0
    bond-miimon 100
    bond-downdelay 200
    bond-updelay 200

    and when this happens it always makes me ask why?


    Did some reading to understand NIC bonding and came up with Bonding and NIC Teaming. I saw the file/program 'ifenslave' is needed. I tried

    Code
    find / -name 'ifenslave'

    on my 0.4.38 install and it was found. Could you also do a find to see if it is in your 0.5 install?


    Also found out the switch needs to support NIC Bonding/Link Aggregation as well. Which I suspect yours does as the 0.4 install worked with it.


    Please be patient with me as I am only at the basic level of understanding NIC bonding.

  • "Also found out the switch needs to support NIC Bonding/Link Aggregation as well. Which I suspect yours does as the 0.4 install worked with it."


    This is true for balance-alb and LACP/802.3ad. For balance-RR as in my case you don't need a switch that supports link aggregation. I don't use switch but a Mikrotik Routerboard RB493G which supports all link agg. protocols.


    ############################
    root@mh0-nas0:~# find / -name 'ifenslave'
    /sbin/ifenslave
    /var/lib/dpkg/alternatives/ifenslave
    /etc/alternatives/ifenslave
    /etc/network/if-up.d/ifenslave
    /etc/network/if-post-down.d/ifenslave
    /etc/network/if-pre-up.d/ifenslave


    EDIT:


    I have deleted the duplicated lines in interfaces file and restarted network. Now I get an error trying to restart. Device "bonding_masters" does not exist.

  • I would suggest revisiting the Bonding & NIC Teaming links in my previous post and see if you can determine how to edit the file to manually add the bonding. Since the error changed it seems as though it is trying to work but it just is not configured properly. I'm sorry I can't be of more help but this is a level of networking I have not dealt with. I've only read about it.


    When I edit files like these I will save the original as you did and then comment out # the lines of code in the copy that I feel are not needed. I also add comments about lines I add or change - to include the date, source and original values, so I can figure out later what it was I was trying to do.


    That Mikrotik Routerboard RB493G is a very interesting piece of hardware. Looking at it and with what you are trying to do it is probably safe to say this setup is not being used in a home.


    Please keep us updated on what is happening so we can improve OMV.


    Good Luck

  • Well,... this is setup is being used at home. I like overkill and also I use ProXmoX cluster. The whole cluster has the main glusterFS storage stationed at OMV so I really need the link aggregation, because a single gigabit card is a bottleneck in my scenario. I can put some more time into this over the weekend. Will keep you posted.

  • Seems like just a bit 'overkill' for the home. Though once it gets all figured out it should work very smoothly. Also just a note to you as I have not seen anyone other posts here in the forum about NIC bonding. You are probably only one of a few doing it with OMV.


    Appreciate you keeping us posted and good luck.

  • I have an average transfer over bonded 4x intel E1000 interfaces on omv 0.4 2gbit/s+, so it's not really an overkill. KVM cluster servers take their tol (zoneminder, zenoss, bacula,...). Back to the bonding subject. I did some trial&eroor research and here are the results.



    My LAB setup is:


    Proxmox KVM server 3.1 as host on 10.4.0.13/24 (single server, no cluster)
    OMV as virtual client on 10.4.0.81/24 over bridged interface on host


    1. I added two more virtual bridges to KVM host and pushed them to OMV client, so I had 3 interfaces in OMV (eth0,eth1,eth2)
    2. Created a simple mode1 (active-backup) bond with eth1 and eth2 --> network crash
    3. edited interfaces file where I deleted double lines and restarted network --> nothing happened, ifconfig shows only lo0 on 127.0.0.1
    4. edited interfaces file again and added lines for eth1 and eth2 (before it was only eth0 setup as manual) with same settings as eth0 (manual)
    5. restarted network --> Device "bonding-masters" does not exist (20 lines of this spam), but the network is accessible again and I can log into OMV over HTTP, active backup setup also works as I was unplugging cables to test this
    6. checked network menu in GUI where I can now see bond0 with 10.4.0.81 IP and also all three eth "physical devices"
    7. changed the IP of bond to 10.4.0.82/24 and applied settings --> network crash
    8. edited interfaces file and saw double lines for bonding again and also no lines defining eth1 and eth2


    This is not yet final conclusion as I'm not done with trial&error tests, but to me it seems as there is some problem with network config in OMV GUI in version 0.5x.

  • It seems as though things are reacting to what you are trying with the different responses you are getting. And my guess is things are fairly close to where they may need to be.


    Going back through this - did you ever get a good working copy of OMV 0.4's interfaces file? I could be wrong but it seems to me the details would be there on how to get the interfaces file set in OMV 0.5.


    Also this may or may not help http://forums.openmediavault.org/viewtopic.php?f=15&t=2363


    I hope some of our other users can help out here. As I'm out of suggestions on how to help. Sorry. Good luck.

  • Hi all. I too have spent a day on trying to get a simple lacp 2 nic 802.3ad bond setup on OMV .5.


    I googled my butt off and there's not much out there. I read all these thread posts and I too have had the bond_masters missing, the double lines in interfaces, etc. i have it up and working but the speed doesn't seem better than 1 gb and I have NO GW access for updates. Only local lan use. GW pings fail.


    I gave this forum one last try and found this thread so I thought I would let the OP, others with this issue and OMV know it's a deal breaker for folks with real and mdadm raid arrays. I need a nas distro that does lacp for throughput for my proxmox vm's.


    I really do enjoy this software. The design is clean and the debian base is easier to work with than freebsd or solaris.


    I hope this gets sorted out soon so I can use it.


    Cheers, Duder

  • Hi.


    My suggestion is to stick with OMV 0.4. It has all you need for Proxmox datastore and the bonding is working without a problem. I tried all bonding types with my HP Procurve 1810G switch. I too couldn't come to the conclusion what's going on with OMV 0.5.


    PhantomSens:


    I tried copying old interface file from 0.4 on the same hardware to 0.5. The bonding works but as soon as I (I said said many times now) only change the IP on bond interface the network setup crashes and you see a new interface file with double lines inserted. Please explain how can this be a Linux distro fault?

  • I can send you files in the evening. Yes, you can test this in any virtual enviroment like I posted some posts above. Vmware, KVM, VirtualBox,... This problem is really easily reproducable.

  • Quote from "bysard"

    I tried copying old interface file from 0.4 on the same hardware to 0.5. The bonding works but as soon as I (I said said many times now) only change the IP on bond interface the network setup crashes and you see a new interface file with double lines inserted. Please explain how can this be a Linux distro fault?


    This may not be a Linux distro fault. And I apologize if some of this is repetitive. Since the file from 0.4 works - until - the IP is changed on the bond interface.


    How is the IP being changed? Is it from the GUI and a selected option, by using OMV-firstaid, automagically from the OS/OMV, or edited by using an editor like nano? Your best thoughts. Mine is by the GUI.


    Also the settings are very likely in the /etc/openmediavault/config.xml file. My single NIC is. I did not mention this file earlier as I did not realize the NIC configs would be there as well. They could be spread through the file though I doubt it. They are probably in the <interfaces> section.


    NOTE: editing the config.xml file could render the OMV system inoperable. Back it up before editing.


    Since the 0.4 file seems to work I would only edit the interfaces & config.xml file with nano and not with the GUI. I would also save a copy of each file. Reason being is when updating OMV, OMV will sometimes overwrite settings we users may make. They are ones we probably shouldn't mess with and they usually are not part of the options in the GUI. By having the copy it is easy to go back to the settings we prefer.


    Please continue to be patient with us, we understand this is easily duplicateable on your system though I personally don't know of any one doing NIC bonding. NIC bonding apparently is done in business and in server farms. Please give tekkbebe some time as he is one of the devs working hard on porting things over to OMV 0.5


    Thanks for understanding.

  • Can I add my confirmation that we have the same issue trying to port bond with mode 802.3ad. I've 2 NICs, 4 ports. Wanted to bond eth2 & eth3 of an add-in card. Using eth0 for local management. If it is done in the web gui and applied, the /etc/networking/interfaces file is reset to having only the loopback adapter config. Used omv-firstaid to restore eth0 (dhcp)
    If I manually add the bond editing /etc/networking/interfaces and bring it up, it appears to come up fine. If I then go into the web interface, The yellow banner warning that I need to apply changes to take affect appears; if I do, it resets the /etc/networking/interfaces again to just a config for the lo interface. There doesn't seem to be a bug report for this, but I can create one if required, with details and configs?
    Apart from that, this has the promise of a great piece of software. Nice that it's on Debian too. Appreciate all the work.
    Patrick

Participate now!

Don’t have an account yet? Register yourself now and be a part of our community!