Howdy, Stranger!

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


possible I/O issues?
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.

possible I/O issues?

so i have a VPS i got recently with a provider.
i keep thinking there is an I/O issue as everytime i try to upload or download something or perform a yum install it somtimes timesout. sometimes it gets very bad.
when i perform the following command:
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

it simply does not finish ( so far ) and the server load goes up and the wa%

here is a screen shot:

now the vps provider keeps saying. everything is fine.

does anyone know of anything else i can do or test to see if it is indeed an issue. Or i should look to scrap the VPS.

«1

Comments

  • How long has the dd command been running? What are the results for ioping?

    Thanked by 2VyprNetworks pablob
  • pablobpablob Member
    edited January 2016

    still going. Even after a fresh reboot it wont finish the command on this VPS.

  • VyprNetworksVyprNetworks Member
    edited January 2016

    Run iostat command or vmstat 1 see what's hitting the disk so hard.

  • i am not to sure what to make of these stats....

  • 50+ wa with no i/o on your VM is pretty bad.

    Thanked by 1VyprNetworks
  • bad yes... but could it be an OS issue do you think? should i scrap it and build it again. Or does it look like it could be something else and i should get the VPS built on different hardware. i am kinda new to trying to troubleshoot VMs
    thanks for your help too :-)

  • @black said:
    50+ wa with no i/o on your VM is pretty bad.

    Not pretty bad but horrible.

    Thanked by 1GM2015
  • Ask to be moved or cancel your server.

    pablob said: bad yes... but could it be an OS issue do you think? should i scrap it and build it again. Or does it look like it could be something else and i should get the VPS built on different hardware. i am kinda new to trying to troubleshoot VMs thanks for your help too :-)

  • blackblack Member
    edited January 2016

    pablob said: bad yes... but could it be an OS issue do you think?

    I doubt it. wa means the time spent waiting for I/O to complete but your container isn't doing any I/O operations which implies that on idle, there's 50% I/O wait. Personally, I do not use / renew services that have more than 5% wa on idle. Is the system responsive? How long does it take to ssh in your server or start a process? How long does it take to ls a directory?

  • jarjar Patron Provider, Top Host, Veteran

    pablob said: but could it be an OS issue do you think?

    Highly unlikely. If you show them that and they blow it off as "there is no problem" I would cancel and never return there again. It COULD be a VM specific issue, but the chances are very low, and any self respecting host would be equally as interested in why your VPS is showing that high iowait if the node is not seeing the same.

    But what we know around here from experience, if you're seeing that kind of iowait and the host says there is no problem, they most likely just don't care.

  • who's your provider.

    pablob said: so i have a VPS i got recently with a provider

    Thanked by 1VyprNetworks
  • hmm, but i have 12 VPSs with these guys and they have been good so far. i will see how it plays out tho and if they don't do good. well i will be moving.

  • pablobpablob Member
    edited January 2016

    also... by just doing that dd command... if i see the wa% hit about 50% (only when i do the test) but it still completes the test in about 6sec at 150MB/s would you say that is worth asking to move to new hardware? as i noticed i am getting that on 2 other VPSs i have with them. but they do seem to be running okay.

    should i name the provider? i don't want to bad mouth them yet i think. see how it plays out first.

  • @pablob said:

    >

    If you do mention them I am embarrassed for them already because that virtual machine is horrid I wonder how many other clients are having that same problem.

  • pablobpablob Member
    edited January 2016

    okay. i am starting to get a bit anoyed now.
    this is what the guy says..

    "Rebooted the server, no uninterruptible task running now. Seems to be fine now. I'll keep on monitoring the server and let you know if there's any issue. "

    even tho i am logged in right now and i can see a server load of 4.00 and wa% between 50 and 100

    i expected better from @dediserve

    Thanked by 1Brad
  • VyprNetworksVyprNetworks Member
    edited January 2016

    @pablob said:

    Load 4.00 and wa% over 5% geez 3wa% is bad don't pass go "just give it up".

  • i trully am hoping they can sort it out. i am asking them to move my VPS to different hardware

    Thanked by 1VyprNetworks
  • @pablob said:
    okay. i am starting to get a bit anoyed now.
    this is what the guy says..

    "Rebooted the server, no uninterruptible task running now. Seems to be fine now. I'll keep on monitoring the server and let you know if there's any issue. "

    even tho i am logged in right now and i can see a server load of 4.00 and wa% between 50 and 100

    i expected better from dediserve

    The support guy, was it Kevin by any chance?

  • i havn't had a chance to experience kevin and his helpfulness yet

  • got the results from the dd test....

    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 1529.17 s, 702 kB/s

  • jarjar Patron Provider, Top Host, Veteran

    @pablob said:
    got the results from the dd test....

    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 1529.17 s, 702 kB/s

    Wow. That's just unusable.

  • blackblack Member
    edited January 2016

    pablob said: 16384+0 records in 16384+0 records out 1073741824 bytes (1.1 GB) copied, 1529.17 s, 702 kB/s

    lol. That's like the speed of a zip drive (if you're old enough to remember those days).

  • still investigating it... i just hope he doesn't come back again saying something like. i rebooted it and it is all good now.
    @black ... if i see the wa% hit about 50% (only when i do the test) but it still completes the test in about 6sec at 150MB/s would you say that is worth asking to move to new hardware?

  • @pablob said:
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 1529.17 s, 702 kB/s

    Thats what I call patience. You actually waited 25 mins for the results...

  • pablobpablob Member
    edited January 2016

    he came back and said "We found some cpu abuse on the hypervisor, we corrected it now."

    so i did another dd test.
    and guess what. still having issues.
    and he still wont reply to me about getting it moved to another node.

    comeon @dediserve , your better then this.

  • "Running a DD is not an accurate way of measuring disk availability"

    they wont accept there is an issue. Can anyone recommend a way for me to know 100% that there is?

  • AmitzAmitz Member
    edited January 2016

    Four letters: M O V E. Really, this is pathetic. It's not really the issue itself, it is about how they handle it.

    Thanked by 2jar inthecloudblog
  • pablob said: "Running a DD is not an accurate way of measuring disk availability"

    While that's generally a true statement, looking at wa & your i/o along with ioping will give you a stronger argument. You should submit ioping and iostat evidence along with your dd result.

    Thanked by 2Amitz inthecloudblog
  • Hi @pablob

    As per your ticket, we can't replicate the wait on another VM on the same HV, so the issue (whetever it is) is localised to your VM - the team are working with you on cloning you on another hypervisor to test, if you need anything pushed or escalated just let me know!

  • vimalwarevimalware Member
    edited January 2016

    @pablob said:

    i expected better from dediserve

    Care to share the location? I've spun up tiny VMs in their Frankfurt, Singapore, Dallas, Amsterdam locations and have always had 6000 iOPS spare in both Xen/KVM.

    A quick fix without support intervention would be to snapshot, convert it to template in the snapshot menu, destroy the problem VM, and re-provision VM from saved template(be sure to ensure template/snapshot shows up as 'created' in the listings.)

    There's no absolute guarantee that the new VM would be spun up on a different HV .

    (Though enforcing some kind of least-utilization provisioning rule, would be the sane thing to do from a provider-perspective.)

    I love providers that let me self-service.

Sign In or Register to comment.