cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gaurav Aradhye <gaurav.arad...@clogeny.com>
Subject Re: Scaling up cpu and memory of user vm above host capacity
Date Wed, 16 Oct 2013 11:58:57 GMT
I have logged issue https://issues.apache.org/jira/browse/CLOUDSTACK-4880 for
this. I will check bevavior for changeServiceForVirtualMachine API too and
log issue if confirmed.

Regards,
Gaurav


On Wed, Oct 16, 2013 at 4:46 PM, Harikrishna Patnala <
harikrishna.patnala@citrix.com> wrote:

> Yes Gaurav, please file a bug ticket for this issue. We should also
> consider host cpu cores while scaling up the VM.
> If you want to check for changeServiceForVirtualMachine API, try it on
> stopped vm since the API is meant for only stopped vms.
>
> Thankyou
> Harikrishna
>
>
> On 16-Oct-2013, at 4:16 PM, Gaurav Aradhye <gaurav.aradhye@clogeny.com>
> wrote:
>
> > Hi Nitin,
> >
> > I am able to scale a virtual machine (using scaleVirtualMachine API) to
> use
> > 5 CPU cores where as the host has only 4 physical CPU cores. According to
> > David, this should not be the case. I can also reboot this instance. But
> I
> > can't create a new instance with this scaled up service offering which
> has
> > 5 CPU cores (Which seems to be a valid behavior).
> >
> > Should I file an issue for this?
> >
> > This issue seems to be present only for CPU and not for memory. I can't
> > scale memory above the available memory in host.
> >
> > I will check the behavior again for the old API
> > (changeServiceForVirtualMachine).
> > I think the old API had issue with both CPU and memory.
> >
> > Regards,
> > Gaurav
> >
> >
> > On Fri, Oct 4, 2013 at 10:36 PM, Nitin Mehta <Nitin.Mehta@citrix.com>
> wrote:
> >
> >> changeServiceForVirtualMachine API was the old API to change the service
> >> offering for a stopped vm only.
> >> I think it shouldn't have succeeded for a running vm. Please file a bug
> if
> >> this is the case
> >>
> >> scaleVirtualMachine is the new API introduced in 4.2 for scaling a
> >> running/stopped vm. Do read the link I pointed below when you get a
> chance.
> >>
> >> Thanks,
> >> -Nitin
> >>
> >> On 03/10/13 11:50 PM, "Gaurav Aradhye" <gaurav.aradhye@clogeny.com>
> wrote:
> >>
> >>> Hi Nitin,
> >>>
> >>> I was trying on running vm only, but I was
> >>> using changeServiceForVirtualMachine API instead of scaleVirtualMachine
> >>> API.
> >>> But I wonder why changeServiceForVirtualMachine API succeeded in
> >>> allocating
> >>> more than host capacity.
> >>>
> >>> What is the basic difference between these two operations?
> >>>
> >>> Regards,
> >>> Gaurav
> >>>
> >>>
> >>> On Tue, Oct 1, 2013 at 10:45 PM, Nitin Mehta <Nitin.Mehta@citrix.com>
> >>> wrote:
> >>>
> >>>> Gaurav - Were you trying this on a stopped vm ? If you try and start
> it
> >>>> with an offering
> >>>> above the host capacity (including over provisioning ) then it
> shouldn't
> >>>> start.  Let me know how it goes.
> >>>>
> >>>> More details on scale vm feature @
> >>>>
> >>>>
> >>
> https://cwiki.apache.org/confluence/display/CLOUDSTACK/Dynamic+scaling+of
> >>>> +C
> >>>> PU+and+RAM
> >>>>
> >>>> On 01/10/13 12:02 AM, "Gaurav Aradhye" <gaurav.aradhye@clogeny.com>
> >>>> wrote:
> >>>>
> >>>>> Thanks David. That disabuses my confusion about the CPU
> provisioning. I
> >>>>> was
> >>>>> using the wrong API to scale up the virtual machine, so above
> >>>> observations
> >>>>> stand invalid till I get the same results with the right API.
> >>>>>
> >>>>> About over-provisioning, I have the over provisioning factor set
as 1
> >>>> both
> >>>>> in case of CPU and memory.
> >>>>>
> >>>>> Regards,
> >>>>> Gaurav
> >>>>>
> >>>>>
> >>>>> On Mon, Sep 30, 2013 at 10:55 PM, David Ortiz <dportiz@outlook.com>
> >>>> wrote:
> >>>>>
> >>>>>> A machine won't be able to support more cores on a VM than the
> >>>> physical
> >>>>>> processor.  That should result in problems trying to deploy
it.  I'm
> >>>>>> guessing the service offering is still valid since you could
add a
> >>>> host
> >>>>>> later which has a hex core or two cpus in it.  As far as RAM
goes,
> do
> >>>>>> you
> >>>>>> have overprovisioning enabled?
> >>>>>>
> >>>>>>> From: gaurav.aradhye@clogeny.com
> >>>>>>> Date: Mon, 30 Sep 2013 14:00:04 +0530
> >>>>>>> Subject: Scaling up cpu and memory of user vm above host
capacity
> >>>>>>> To: dev@cloudstack.apache.org
> >>>>>>>
> >>>>>>> Hi,
> >>>>>>>
> >>>>>>> I am trying to automate a scenario here. I have only one
host in
> >>>>>> cluster
> >>>>>>> with 4 CPU cores and 15 GB total memory. When I try to scale
up cpu
> >>>>>> and
> >>>>>> RAM
> >>>>>>> of a running user vm above the host capacity, it doesn't
throw any
> >>>>>> error
> >>>>>>> and I can see the updated values in VM statistics too.
> >>>>>>>
> >>>>>>> For CPU, I am able to change the service offering of user
vm as  5
> >>>>>> cores
> >>>>>> *
> >>>>>>> 100 MHz (even though host has 4 cores). I am not sure how
this
> >>>>>> calculation
> >>>>>>> is done. Definitely many no. of virtual cores can be formed
on host
> >>>>>> (more
> >>>>>>> than 4), but is it possible to allocate 5 cores to single
VM ?
> >>>> When I
> >>>>>> try
> >>>>>>> to deploy new VM with 5 core CPU service offering, then
in this
> >>>> case
> >>>>>> it
> >>>>>>> fails saying not enough server capacity.
> >>>>>>>
> >>>>>>> Also, For memory, I am able to create 17 GB memory service
offering
> >>>>>> and
> >>>>>>> allocate it to any running user vm (although the total memory
on
> >>>> host
> >>>>>> is
> >>>>>> 15
> >>>>>>> GB).
> >>>>>>>
> >>>>>>> Any directions? Is this an issue or am I missing something
here?
> >>>>>>>
> >>>>>>> Regards,
> >>>>>>> Gaurav
> >>>>>>
> >>>>>>
> >>>>
> >>>>
> >>
> >>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message