Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


IO Slow with SYS-E32-3
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

IO Slow with SYS-E32-3

jazz1611jazz1611 Member
edited May 2014 in Help

Hi,

I have ordered SYS-E32-3 on us. And i installed Proxmox VE 3.2 on server. Current my server just have 2 VM running. I tested the IO of VM. There is result:

From Node:

root@ns4007xxx:~# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$

16384+0 records in

16384+0 records out

1073741824 bytes (1.1 GB) copied, 10.8188 s, 99.2 MB/s

From VM:

[root@test ~]# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$

16384+0 records in

16384+0 records out

1073741824 bytes (1.1 GB) copied, 16.3143 s, 65.8 MB/s

Server SP-128 of my friend with more 10 VM running, same hard disk model, same Virtio on VM, but i tested the IO on VM (my friend) about ~100MB/s.

root@sv1 [~]# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$

16384+0 records in

16384+0 records out

1073741824 bytes (1.1 GB) copied, 12.4675 s, 86.1 MB/s

root@sv1 [~]# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$

16384+0 records in

16384+0 records out

1073741824 bytes (1.1 GB) copied, 12.3068 s, 87.2 MB/s

I dont know why it very different? Which HardRAID-1 (VM of me) the IO ~65MB/s, SoftRAID-1 (VM of my friend) the IO ~100MB/s

P/s: They are replaced my RAID Card but still ~65MB/s on VM (just 2 VM running). Other VM running same my node not use high IO.

Regards,

«1

Comments

  • 110 MB/s is not too bad

  • @joodle said:
    110 MB/s is not too bad

    110MB/s there IO on VM of my friend. not VM of me. What problem i getting?

    Regards,

  • linuxthefishlinuxthefish Member
    edited May 2014

    Change to ploop file system if you want 'high I/O' on RAID1 7200 rpm drive...

  • WSCallumWSCallum Member
    edited May 2014

    I get 124MB/s on the exact same model, so your results aren't too bad.

    If you've only just had the server delivered then the RAID Building could be a cause of the problem, when I first had mine delivered it took over 24 hours for the RAID to build.

    You can try putting it in rescue mode as well and doing a hardware check.

  • @linuxthefish said:
    Change to ploop file system if you want 'high I/O' on RAID1 7200 rpm drive...

    What is ploop? I created the partition below:

    It will effect the performance?

  • @WSCallum said:
    I get 124MB/s on the exact same model, so your results aren't too bad.

    If you've only just had the server delivered then the RAID Building could be a cause of the problem, when I first had mine delivered it took over 24 hours for the RAID to build.

    You can try putting it in rescue mode as well and doing a hardware check.

    They are replaced the HardRaid Card. But IO still slow on VM, some problem i getting but they and me dont have know. I think that. We are looking find problem.

    Regards,

  • @jazz1611 said:
    Regards,

    When I first ordered mine I put it into rescue mode and did hardware check, RAID Build took over 24 hours so if the cards only just been replaced then RAID could be a cause of slow IO.

  • That layout should be fine, make sure you have enough SWAP to cover any mishaps though. Changing to ploop will just increase the numbers if you like that sort of thing, it's not magic that will speed the disk up though.

    RAM caching could help if you have tons free, although linux does this by default. @mtwiscool knows lots about ram caching!

    http://openvz.org/Ploop

  • jazz1611jazz1611 Member
    edited May 2014

    I use KVM for VPS. Not OpenVZ. With Virtio on KVM, the IO will be better performance but it not. I'm really disappointed. 3 days and they are not resolved anything.


  • later some seconds

    Maybe problem is there. Make low IO performance. :(

  • linuxthefishlinuxthefish Member
    edited May 2014

    @jazz1611 said:
    Maybe problem is there. Make low IO performance. :(

    Wait for RAID to sync. It's basicly a single disk server, anything over 90 MB/s is a bonus!

  • FalzoFalzo Member
    edited May 2014

    THEY are not responsible for what YOU are doing with your UNMANAGED DEDICATED server... ;-)

    I'd say they've been very fair trying to help with changing raid card at least, but guess thats not really an issue here, as of your first post you're getting ~100MB/s on the node...

    not bad looking IMHO and I wonder why just raid should be responsible for lower I/O on your VM (against your node)? seems more a to be a problem with configuration or driver, if at all.

  • Just curious, what do you think the IO speed should be?

  • @Falzo said:

    When i installed Proxmox, i dont config anything on Proxmox, i just create a VM as usual I did.

  • @deptadapt said:
    Just curious, what do you think the IO speed should be?

    almost or same as node, because i installed Proxmox on Server SP-128 (SoftRaid) for my friend then i know IO speed on VM

  • wychwych Member

    You know how many variables are involved with I/O speed right?

    But yes, your array is probably running slow due to the rebuild.

  • If i change back to Boot Hard Disk (not Rescue mode) the rebuild HardRAID will be continue or stopped?

  • wychwych Member

    It should continue once you reboot. However I would let it build then safely switch back to os.

  • but why sync hard raid very slow? i see softraid very fast.

  • WSCallumWSCallum Member
    edited May 2014

    @jazz1611 said:
    but why sync hard raid very slow? i see softraid very fast.

    The RAID Cards on the E3s aren't too great I believe, I think I've seen someone say before that OVH themselves admitted that Soft RAID would perform better than the RAID Cards they use in those machines.

  • @deptadapt said:
    Just curious, what do you think the IO speed should be?

    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 8.92054 s, 120 MB/s

    Cheapest E3 with HW raid, semi busy cPanel server.

  • jazz1611jazz1611 Member
    edited May 2014

    Hi,

    There is result on Rescue Mode
    root@rescue:~# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$ 16384+0 records in 16384+0 records out 1073741824 bytes (1.1 GB) copied, 0.255227 s, 4.2 GB/s

    It is slow or fast? Because the Raid array re-synchronization up to 90%. Please suggest for me.

  • rds100rds100 Member

    It's OK for the ramdisk of the rescue system :)

  • FalzoFalzo Member

    @jazz1611 said:
    It is slow or fast? Because the Raid array re-synchronization up to 90%. Please suggest for me.

    don't do benchmarks if you don't know how to rate them at all.
    such a single number does not say very much about nothing.

    Thanked by 2linuxthefish netomx
  • wychwych Member

    Is your RAID rebuilt...?

    How does it perform in the real world now?

  • @wych said:
    Is your RAID rebuilt...?

    How does it perform in the real world now?

    Current status Hard Raid on Rescue Mode. Im waiting it completed and it high performance IO when it completed. The problem is there.

  • wychwych Member
    edited May 2014

    @jazz1611 said:
    The problem is there.

    I am still not sure you actually have a problem once the RAID is rebuilt; multiple members have said the RAID card's are not top of the line.

  • jazz1611jazz1611 Member
    edited May 2014

    @wych said:

    They are have tell me test hard disk but i dont know how to that?

    The speed of synchronization is unchangeable, it ill depend on how many changes have to be synchronized in between the drives.

    I dont really understand what did you mean with the boot of the VM, the VM layer is beyond of our support. But, in general, yes, the RAID is an independent layer of your infrastructure, you can do whatever you need at the top of the Hypervisor while the RAID syncrhonzation is ongoing.

    >

    The issue might not be with the RAID itself but maybe one of the drive is not giving the right performances. Also, you can run those tests from the rescue mode or directly through ssh/terminal

  • wychwych Member

    @jazz1611 said:

    The issue might not be with the RAID itself but maybe one of the drive is not giving the right performances. Also, you can run those tests from the rescue mode or directly through ssh/terminal

    http://www.ovh.com/us/g920.rescue_mode

    Hardware check interface:

    Memtest: to check the memory (RAM)

    Cpuburn: to check the stability of the CPU

    Fsck: to check the file system

    State: to check the hard drives


    Explorer: to navigate within your folders and files

  • There is result IO on node.

    root@ns4007xxx:~# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$ 16384+0 records in 16384+0 records out 1073741824 bytes (1.1 GB) copied, 7.19945 s, 149 MB/s root@ns4007xxx:/var/lib/vz# dd if=/dev/zero of=test_$$ bs=64k count=16k conv=fdatasync && rm -f test_$$ 16384+0 records in 16384+0 records out 1073741824 bytes (1.1 GB) copied, 7.25761 s, 148 MB/s

    And there is result IO on VM

    Without cache Writeback

    With cache Writeback

    Should enable Cache Writeback on VM? My Server use HardRAID LSI SAS 9211-4i

Sign In or Register to comment.