Fresh Installation is loosing network connection

  • Hi,
    first of all, thanks for the great pice of Distri. It was easy to install and setup.


    I have a problem, where i ´am not sure if it comes from using KVM on 12.04.3 with passtroughing network and ibm Serveraid M5015 with 3 WD Red 3 TB harddisk on it.
    Everything seems to works, CIFS, NFS with avahi to VDR. only owncloud and lvm is activated, on the serveraid i defined a lvm with omv.


    Failure:
    - when i copy a large amount of data to the omv attatched lvm it runs for about 1 hour, then the network connection is refused, copyprocess is brocken and stopped
    - it doesn´t matter if copying over cifs or nfs


    What i found in log:

    Code
    Jan 14 09:03:49 omv kernel: [ 2.832747] PM: Starting manual resume from disk
    Jan 14 09:03:49 omv kernel: [ 2.832749] PM: Resume from partition 254:5
    Jan 14 09:03:49 omv kernel: [ 2.832750] PM: Checking hibernation image.
    Jan 14 09:03:49 omv kernel: [ 2.832816] PM: Error -22 checking image file
    Jan 14 09:03:49 omv kernel: [ 2.832818] PM: Resume from disk failed.


    i have disabled pm for hdd, and in powermanagment i also disabled monitoring.




    Why are here white spaces? Where do they come from? These are exactly at the times where i tried to copy my old archiv to the new omv.


    What i also donßt understand in log and doesn´t look normal:


    Some Memory Problems or a resource Limit?, followed of an reboot? But why? is that done by a watchdog?



    Maybe anyone of you have a suggestion or solution?


    Thx

    Images

    Files

    • syslog.zip

      (206.48 kB, downloaded 246 times, last: )

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

    Edited once, last by torsten73 ().

  • These white spaces are "shown" when nothing was logged, e.g. the system wasn't running. Some guys here know a bit more about KVM, let them have a look.


    Greetings
    David

    "Well... lately this forum has become support for everything except omv" [...] "And is like someone is banning Google from their browsers"


    Only two things are infinite, the universe and human stupidity, and I'm not sure about the former.

    Upload Logfile via WebGUI/CLI
    #openmediavault on freenode IRC | German & English | GMT+1
    Absolutely no Support via PM!


    I host parts of the omv-extras.org Repository, the OpenMediaVault Live Demo and the pre-built PXE Images. If you want you can take part and help covering the costs by having a look at my profile page.

  • i should give some more information of kvm konfig:
    - vm with 2 cores on a xeon quad core SB
    - 2 gb ram defined
    - virtio used for system disk
    - vm is on an 250gb ssd


    so i need to find out why the vm resets on high traffic and why the load is to high ?

  • It´s also noticeable, that the wachtdog, needs much CPU consumption(around 20-40%)
    the copyspeed is really slow, much slower than expectet.
    i got aound 130 Mbit/sec starting, but later on ist falls back < 20Mbit.
    the raid xor cpu is vent cooled.
    the overview cpu useage show a load when copying for around 50%, htop also. Normally i would expect values <15% for copying with these Raid and CPU

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

  • a little update:


    i reactivated power managment, also APM 254 and write cache. Then i changed my eth0 to eth1 as VM Bridge with virtio e1000, the pci passtrough drive from my Tyan S5512GM4NR http://tyan.com/product_SKU_sp…=MB&pid=700&SKU=600000226 with Quad Lan Intel 82574L is the same and has no problems with Kernel 3.5 and above.


    Now i had transfered 12h and around 2TB without brake. So the reboot seems to be network driver related ?


    But now i have a poor performance when i copy from my old raid to the new one. My Gigabit connection from SBS 2003 drops from 50%-30% (whats the maximum for windows is) to less than 35%-15%. The System load is reduced, and also the mem page cache looks different.
    But i don´t have an answer what that means. Why i have now such a poor transfer rate?


    I mean in early testings i had some trouble with older kernel and the networkdriver e1000 had to be installed manually.


    But now i cant see network statistics, because eth0 isn´t present. How can i change this?


    Lets have a look on the statistics for comparision:

  • Das Ergibt eine Datenrate bei gerade mal 1,49TB in 11h (alle Videofiles von mind. 1GB) von rund 2,25MB / Minute ?


    Puh there is much to be be optimized. Filesystem ext4 on LVM should be ok. Maybe the controller works in Raid 5 with write cache enabled (only 256 MB onboard) works so poor.

    Images

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

  • Of course. The LAN driver could be the problem. Or was the problem ;-)


    btw. Please update to the latest OMV which was releasd this night ( 0.5.31 ). It optimizes the stripe_cache_size which should give you higher transfer rates as well as less drops.

    OMV 4.x| HP Microserver | 256GB Samsung 830 SSD for system | 4x 2TB in a RAID5
    OMV 4.x| Odroid XU4 | 5TB Data drive | 500GB Backup drive
    OMV 5.x| Raspberry Pi 4 | 6TB Data drive | 500GB SSD drive

  • Hi Torsten
    Even though your stability prblems seem to be gone I think that you should tune your VM instalation to get the most stable installation. I have had my KVM based VMs Debian Wheezy and Ubuntu 12.04.2 running without a hickup for more than 6 months on my small AMD eE350 MB that is also running OMV.


    Primarily I suggest that you use the device model: virtio for networking in stead of e1000


    If you haven't then tune your elevator model for virtualisation in the guest VM by editing /etc/default/grub. In the line GRUB_COMMAND_LINE_DEFAULT insert

    Code
    transparent_hugepage=always elevator=noop


    The do a grub update

    Code
    update-grub


    You can do the same elevator tuning on the KVM host by editing /etc/default/grub. In the line GRUB_COMMAND_LINE_DEFAULT insert

    Code
    transparent_hugepage=always elevator=deadline


    The do a grub update

    Code
    update-grub


    There are more KVM tuning parameters. i have described some of the in my Howto KVM/qemu on OMV http://forums.openmediavault.org/viewtopic.php?f=13&t=2417. Even though you run OMV as a guest vm the same tuning parameters and stability meassures apply.

  • i have done some research last night.


    we are using debian kernel 2.6.32-5 and i found many regressions with using a Intel 82574L. my host is Ubuntu 12.04 since beginnig using this board, i had never any netwok problems. But Ubuntu uses kernel 3.2 from beginning.
    So is there a chance without making other trouble to use e newer kernel, and when yes which one should i use?


    So before i tune virtio i would like better to passtrough one of my 82574L, because i have to much of them :-)

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

  • I would recommend kernel 3.2 or newer before using the system under KVM - both for the guest and for the host. I use kernel 3.2.4 for my host which is very stable.


    Have you upgraded to a newer kernel on your host - like the raring og saucy kernel. This might help you getting the newest drivers for your network/mobo https://wiki.ubuntu.com/Kernel/LTSEnablementStack


    There is a collections of howto and guides regarding many issues in Squeeze/OMV here http://forums.openmediavault.org/viewtopic.php?f=3&t=1289
    The guide for using kernel 3.2.4 on Squeeze/OMV is here http://forums.openmediavault.o…ic.php?f=10&t=2258#p13821
    There are newer kernels for OMV like 3.10.x but i haven't tried them as they are unsupported


    Like you I would start using standard precise 12.04.3 kernel for the host and squeeze/OMV with backports kernel 3.2 on the guest. If you have driver or performance issues i would upograde the host to raring og saucy kernel - but only if there are performance or driver issues.


    Regarding your network speed my best guess is that its the e1000 driver that slow everything down. If you change to virtio you will experience a substantial performance lift. To start you can test by just assigning one host interface (br0) through virtio.


    For your networking in general i would recomment trunking all four interfaces on the hostlevel. Then I would assign the trunked network interface to the OMV guest through virtio. This way you would get the combined network throughput - like 4Gb/s - delivered on one interface to your guest.

  • You can also install the 3.2 backports kernel right from the OMV Gui, if you have installed OMV-Extras.


    Howto for OMV-Extras is here - if you have not already installed it - http://omv-extras.org/simple/i…install-omv-extras-plugin


    Then just click on OMV-Extras.org on the left panel and hit the "Install Backports 3.2 kernel"-button

    OMV 4.x| HP Microserver | 256GB Samsung 830 SSD for system | 4x 2TB in a RAID5
    OMV 4.x| Odroid XU4 | 5TB Data drive | 500GB Backup drive
    OMV 5.x| Raspberry Pi 4 | 6TB Data drive | 500GB SSD drive

  • i have install the debian backport kernel over OMV-Extras. Thats an easy and good way. But it doesn´t solve the problem entirely. It´s better now, cause the system doens´t restart on high network traffic (and therefor on message network unreachable), but the network still stops responding, when i transfer files over it. It´s also on ssh noticeable. And the performance drop is much, altough i now have around 45% with Cifs from SBS2003. Every few seconds it stops transfering data for several seconds.
    I tried to get 3.5 or newer, but this is not so easy to do, the dependencies are to much.


    My ubuntu host uses 3.8 backports and earlier 3.5. Both of them had no problems with my networkcards.


    I have read, that omv 0.6 is comming in the next days, it will use 3.5 and with backports up to 3.12.


    So i think i can stop working on fixing it.

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

  • Quote from "torsten73"

    I have read, that omv 0.6 is comming in the next days, it will use 3.5 and with backports up to 3.12.


    Actually 0.6 is based on Wheezy and uses the same 3.2 kernel that you just installed as the backports kernel. You are right about the 3.12 backports kernel which should help fix a lot of hardware issues :)

    omv 5.3.9 usul | 64 bit | 5.3 proxmox kernel | omvextrasorg 5.2.6
    omv-extras.org plugins source code and issue tracker - github


    Please read this before posting a question.
    Please don't PM for support... Too many PMs!

  • oh sure, your are right. I hope it really will be released, before support end of squeeze, because then it´s latestly beginning febuary.

    Vielen Dank, Regards
    Torsten
    --------------------------------------------------------------------------------------------------------------------------------------
    Tyan Xeon Server S5512, Ubuntu YAVDR 12.04.3 as Host, OMV 1.0 as KVM Guest with Network and Raid Passtrough

  • Luckily, the Squeeze EOL hasn't been announced. I doubt they will EOL it with short notice. So, I think we safe through February and probably March. Not that I want to wait for that long for 0.6 :lol:

    omv 5.3.9 usul | 64 bit | 5.3 proxmox kernel | omvextrasorg 5.2.6
    omv-extras.org plugins source code and issue tracker - github


    Please read this before posting a question.
    Please don't PM for support... Too many PMs!

  • i worked something more on my network setting. First i modified my host and guest with jensk hints:


    Quote

    If you haven't then tune your elevator model for virtualisation in the guest VM by editing /etc/default/grub. In the line GRUB_COMMAND_LINE_DEFAULT insert


    Code: Select all
    transparent_hugepage=always elevator=noop


    then i configured a new br1 on my host similar like before, but on eth3. This one i assigned to omv with virtio again and disabled passtrough. And i corrected the network statistics, so i can make a screenshot.


    the results are much better. Maybe not the level like i would have an optimal working passtrough, but it´s faster like every other setting before.


    First my copy test from a WD Red 3TB single HDD over Host eth0 to Guest br1 with virtio to my IBM Raid 5 with 3 Seagate SV35 ST3000VX000 3 TB:



    Next a copy test from my old sbs2003 raid adaptec 2820SA with 6 * 1 TB WD RE4 (1 Hot Spare) as Raid 5:



    I am not sure if this is a good throughput for a 1Gbit Link between old and new server?

Participate now!

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