Howdy, Stranger!

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


Get Load Of Host Node
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.

Get Load Of Host Node

Is there a way in an openvz or KVM environment that you can get the host nodes i/o load?

Mun

Comments

  • blergh_blergh_ Member
    edited January 2014

    From the container or from hostnode? htop? top? iotop? iostat?

  • MunMun Member

    The Hostnode. Usually htop, top, and iotop report your containers load.

  • @Mun said:
    Is there a way in an openvz or KVM environment that you can get the host nodes i/o load?

    Mun

    What's IOWAIT like in the container/VPS?

  • MunMun Member

    @Virtovo said:
    What's IOWAIT like in the container/VPS?

    Couldn't tell you as I have no server I am actively testing, just wanted a way to see if I could get the load of a host server.

  • @Mun said:
    Couldn't tell you as I have no server I am actively testing, just wanted a way to see if I could get the load of a host server.

    On HW virtualization like KVM, definitely can't. But I'm somehow not sure about OpenVZ!

  • MunMun Member

    Yeah I was thinking that, I was just curious if it was possible because you could possibly monitor full deployments if you spaced things out right.

  • @Mun said:
    The Hostnode. Usually htop, top, and iotop report your containers load.

    Not sure I've understood your initial question right, but if you ran any of those on the hostnode you'd get.. reports of load(s) on your hostnode.

  • Your vm is effectively sandboxed. without something on the host specifically giving you the info, you wont be able to get it.

  • VDS6VDS6 Member
    edited January 2014

    In kvm iotop will list qemu processes and their read/write activity. The one (qemu process), at the very top would be using most of io resources.

    In ovz there are multiple ways to find abusive vz containers.

    Find 10 most active in write activity containers

    for i in vzlist -1; do echo $i cat /proc/bc/$i/ioacct |grep write;done|sort -rn -k3|head -n10

    Find 10 most active in read activity containers

    for i in vzlist -1; do echo $i cat /proc/bc/$i/ioacct |grep read;done|sort -rn -k3|head -n15

    So if you are looking for some abusers on your ovz node, you can find them with these simple commands mentioned above.

    Hope it helps in any way.

  • jarjar Patron Provider, Top Host, Veteran

    I believe the correct answer to this question is no :)

    But you can guess that when running "top" takes 100% CPU and iowait is high while you're performing no notable reads/writes that the load is likely very high on the node.

    Thanked by 1AnthonySmith
  • MunMun Member

    Alright :(

  • drserverdrserver Member, Host Rep

    Degradation of service level without obvious reason is 1st sign of high HW node load

  • @Zen said:
    which you could build into it I guess.

    s/you/the provider/

  • perennateperennate Member, Host Rep

    Zen said: This topic isn't necessarily provider specific.

    Well I think it is, if you own the host node then you could just run htop. The point of the topic was to get load of host node through from inside a container without requiring modifications of the virtualization technology; which generally isn't possible unless you find a bug in the virtualization technology that allows you to escalate your privileges :)

    Thanked by 1Zen
  • This method is by no means perfect by any amount. But it requires using the "st%" in top. This stands for steal and from what i remeber shows the amount of time that that ur vm requested cpu access but was given to another vm (not sure if it includes host also). Because there are many ways cpu can be used its hard to know if its cpu or io wait that is casuing steal. But for example if you have a high io wait and a low system or user cpu usage and then there is cpu steal that would most likly be the time other vms were getting the IO you wanted.

Sign In or Register to comment.