Hi Javier<br><br>Thanks for the inputs but I came across another problem while testing:<br><br><span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;" id="internal-source-marker_0.8926840357036702">If opennebula receives multiple vm requests in a short span of time,
the scheduler might take decisions for all these vms considering the
host monitoring information available from the last monitoring cycle.
Ideally, before processing every pending request, fresh host monitoring
information has to be taken into account as the previous set of
requests might have already changed the host’s state. This can result in
over committing when host is being used close to its full capacity. </span><br><br><b>Is there any workaround which helps the scheduler to overcome the above problem ?</b><br><br>steps to reproduce the problem scenario:<br>
<br><span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;" id="internal-source-marker_0.8926840357036702">Host 1 : Total memory = 3GB</span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">Host 2 : Total memory = 2GB</span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">Assume Host1 and Host2 have same number of CPU cores.</span> ( Host1 will have a higher RANK value )<br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;"></span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">VM1: memory = 2GB</span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">VM2: memroy = 2GB</span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;"></span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">Start VM1 and VM2 immediately one after the other. Both VM1 and VM2 will come up on Host1. </span>( Thus over committing )<br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;"></span><br style="color: rgb(0, 0, 0);">
<span style="font-size: 11pt; font-family: Arial; color: rgb(0, 0, 0); background-color: transparent; font-weight: normal; font-style: normal; text-decoration: none; vertical-align: baseline;">Start VM1 and VM2 with an intermediate delay of 60sec. VM1 will come up on Host1 and VM2 will come up on
Host2. This is true because opennebula would have fetched a fresh set of
host monitoring information in that time.</span><br><br><br><div class="gmail_quote">On 4 November 2010 02:04, Javier Fontan <span dir="ltr"><<a href="mailto:jfontan@gmail.com">jfontan@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">Hello,<br>
<br>
It looks fine to me. I think that taking out the memory the hypervisor<br>
may be consuming is key to make it work.<br>
<br>
Bye<br>
<br>
On Wed, Nov 3, 2010 at 8:32 PM, Rangababu Chakravarthula<br>
<div><div></div><div class="h5"><<a href="mailto:rbabu@hexagrid.com">rbabu@hexagrid.com</a>> wrote:<br>
> Javier<br>
><br>
> Yes we are using KVM and OpenNebula 1.4.<br>
><br>
> We have been having this problem since a long time and we were doing all<br>
> kinds of validations ourselves before submitting the request to OpenNebula.<br>
> (there should be enough memory in the cloud that matches the requested<br>
> memory & there should be atleast one host that has memory > requested memory<br>
> ) We had to do those because OpenNebula would schedule to an arbitrary<br>
> host based on the existing logic it had.<br>
> So at last we thought that we need to make OpenNebula aware of memory<br>
> allocated of running VM's on the host and started this discussion.<br>
><br>
> Thanks for taking up this issue as priority. Appreciate it.<br>
><br>
> Shashank came up with this patch to kvm.rb. Please take a look and let us<br>
> know if that will work until we get a permanent solution.<br>
><br>
> ====================================================================================<br>
><br>
> $mem_allocated_for_running_vms=0<br>
> for i in `virsh list|grep running|tr -s ' ' ' '|cut -f2 -d' '` do<br>
> $dominfo=`virsh dominfo #{i}`<br>
> $dominfo.split(/\n/).each{|line|<br>
> if line.match('^Max memory')<br>
> $mem_allocated_for_running_vms += line.split("<br>
> ")[2].strip.to_i<br>
> end<br>
> }<br>
> end<br>
><br>
> $mem_used_by_base_hypervisor = [some xyz kb that we want to set aside for<br>
> hypervisor]<br>
><br>
> $free_memory = $total_memory.to_i - ( $mem_allocated_for_running_vms.to_i +<br>
> $mem_used_by_base_hypervisor.to_i )<br>
><br>
> ======================================================================================<br>
><br>
> Ranga<br>
><br>
> On Wed, Nov 3, 2010 at 2:16 PM, Javier Fontan <<a href="mailto:jfontan@gmail.com">jfontan@gmail.com</a>> wrote:<br>
>><br>
>> Hello,<br>
>><br>
>> Sorry for the delay in the response.<br>
>><br>
>> It looks that the problem is OpenNebula calculating available memory.<br>
>> For xen >= 3.2 there is a reliable way to get available memory that is<br>
>> calling "xm info" and getting "max_free_memory" attribute.<br>
>> Unfortunately for kvm or xen < 3.2 there is not such attribute. I<br>
>> suppose you are using kvm as you tell about "free" command.<br>
>><br>
>> I began analyzing the kvm IM probe that gets memory information and<br>
>> there is a problem on the way to get total memory. Here is how it now<br>
>> gets memory information:<br>
>><br>
>> TOTALMEMORY: runs virsh info that gets the real physical memory<br>
>> installed in the machine<br>
>> FREEMEMORY: runs free command and gets the free column data without<br>
>> buffers and cache<br>
>> USEDMEMORY: runs top command and gets used memory from it (this counts<br>
>> buffers and cache)<br>
>><br>
>> This is a big problem as those values do not match one with another (I<br>
>> don't really know how I failed to see this before). Here is the<br>
>> monitoring data from a host without VMs.<br>
>><br>
>> --8<------<br>
>> TOTALMEMORY=8193988<br>
>> USEDMEMORY=7819952<br>
>> FREEMEMORY=7911924<br>
>> ------>8--<br>
>><br>
>> As you can see it makes no sense at all. Even the TOTALMEMORY that is<br>
>> got from virsh info is very misleading for oned as the host linux<br>
>> instance does not have access to all that memory (some is consumed by<br>
>> the hypervisor itself) as seen calling a free command:<br>
>><br>
>> --8<------<br>
>> total used free shared buffers cached<br>
>> Mem: 8193988 7819192 374796 0 64176 7473992<br>
>> ------>8--<br>
>><br>
>> I am also copying this text as an issue to solve this problem<br>
>> <a href="http://dev.opennebula.org/issues/388" target="_blank">http://dev.opennebula.org/issues/388</a>. It is masked to be solved for<br>
>> 2.0.1 but the change will be compatible with 1.4 as it seems the the<br>
>> only changed needed is the IM problem.<br>
>><br>
>> I can not offer you an immediate solution but we'll try to come up<br>
>> with one as soon as possible.<br>
>><br>
>> Bye<br>
>><br>
>> On Wed, Nov 3, 2010 at 7:08 PM, Rangababu Chakravarthula<br>
>> <<a href="mailto:rbabu@hexagrid.com">rbabu@hexagrid.com</a>> wrote:<br>
>> > Hello Javier<br>
>> > Please let us know if you want us to provide more detailed information<br>
>> > with<br>
>> > examples?<br>
>> ><br>
>> > Ranga<br>
>> ><br>
>> > On Fri, Oct 29, 2010 at 9:46 AM, Rangababu Chakravarthula<br>
>> > <<a href="mailto:rbabu@hexagrid.com">rbabu@hexagrid.com</a>> wrote:<br>
>> >><br>
>> >> Javier<br>
>> >><br>
>> >> We saw that VM's were being deployed to the host where the allocated<br>
>> >> memory of all the VM's was higher than the available memory on the<br>
>> >> host.<br>
>> >><br>
>> >> We think OpenNebula is executing free command on the host to determine<br>
>> >> if<br>
>> >> there is any room and since free would always return the actual memory<br>
>> >> that<br>
>> >> is being consumed and not the allocated, opennebula would push the new<br>
>> >> jobs<br>
>> >> to the host.<br>
>> >><br>
>> >> That's the reason we want OpenNebula to be aware of memory allocated to<br>
>> >> the VM's on the host.<br>
>> >><br>
>> >> Ranga<br>
>> >><br>
>> >> On Thu, Oct 28, 2010 at 2:02 PM, Javier Fontan <<a href="mailto:jfontan@gmail.com">jfontan@gmail.com</a>><br>
>> >> wrote:<br>
>> >>><br>
>> >>> Hello,<br>
>> >>><br>
>> >>> Could you describe the problem you had? By default the scheduler will<br>
>> >>> not overcommit cpu nor memory.<br>
>> >>><br>
>> >>> Bye<br>
>> >>><br>
>> >>> On Thu, Oct 28, 2010 at 4:50 AM, Shashank Rachamalla<br>
>> >>> <<a href="mailto:shashank.rachamalla@hexagrid.com">shashank.rachamalla@hexagrid.com</a>> wrote:<br>
>> >>> > Hi<br>
>> >>> ><br>
>> >>> > We have a requirement where in the scheduler should not allow memory<br>
>> >>> > over<br>
>> >>> > committing while choosing a host for new vm. In order to achieve<br>
>> >>> > this,<br>
>> >>> > we<br>
>> >>> > have changed the way in which FREEMEMORY is being calculated for<br>
>> >>> > each<br>
>> >>> > host:<br>
>> >>> ><br>
>> >>> > FREE MEMORY = TOTAL MEMORY - [ Sum of memory values allocated to<br>
>> >>> > VMs<br>
>> >>> > which<br>
>> >>> > are currently running on the host ]<br>
>> >>> ><br>
>> >>> > Please let us know if the above approach is fine or is there any<br>
>> >>> > better<br>
>> >>> > way<br>
>> >>> > to accomplish the task. We are using opennebula 1.4.<br>
>> >>> ><br>
>> >>> > --<br>
>> >>> > Regards,<br>
>> >>> > Shashank Rachamalla<br>
>> >>> ><br>
>> >>> > _______________________________________________<br>
>> >>> > Users mailing list<br>
>> >>> > <a href="mailto:Users@lists.opennebula.org">Users@lists.opennebula.org</a><br>
>> >>> > <a href="http://lists.opennebula.org/listinfo.cgi/users-opennebula.org" target="_blank">http://lists.opennebula.org/listinfo.cgi/users-opennebula.org</a><br>
>> >>> ><br>
>> >>> ><br>
>> >>><br>
>> >>><br>
>> >>><br>
>> >>> --<br>
>> >>> Javier Fontan, Grid & Virtualization Technology Engineer/Researcher<br>
>> >>> DSA Research Group: <a href="http://dsa-research.org" target="_blank">http://dsa-research.org</a><br>
>> >>> Globus GridWay Metascheduler: <a href="http://www.GridWay.org" target="_blank">http://www.GridWay.org</a><br>
>> >>> OpenNebula Virtual Infrastructure Engine: <a href="http://www.OpenNebula.org" target="_blank">http://www.OpenNebula.org</a><br>
>> >>> _______________________________________________<br>
>> >>> Users mailing list<br>
>> >>> <a href="mailto:Users@lists.opennebula.org">Users@lists.opennebula.org</a><br>
>> >>> <a href="http://lists.opennebula.org/listinfo.cgi/users-opennebula.org" target="_blank">http://lists.opennebula.org/listinfo.cgi/users-opennebula.org</a><br>
>> >><br>
>> ><br>
>> ><br>
>><br>
>><br>
>><br>
>> --<br>
>> Javier Fontan, Grid & Virtualization Technology Engineer/Researcher<br>
>> DSA Research Group: <a href="http://dsa-research.org" target="_blank">http://dsa-research.org</a><br>
>> Globus GridWay Metascheduler: <a href="http://www.GridWay.org" target="_blank">http://www.GridWay.org</a><br>
>> OpenNebula Virtual Infrastructure Engine: <a href="http://www.OpenNebula.org" target="_blank">http://www.OpenNebula.org</a><br>
><br>
><br>
<br>
<br>
<br>
</div></div>--<br>
<div><div></div><div class="h5">Javier Fontan, Grid & Virtualization Technology Engineer/Researcher<br>
DSA Research Group: <a href="http://dsa-research.org" target="_blank">http://dsa-research.org</a><br>
Globus GridWay Metascheduler: <a href="http://www.GridWay.org" target="_blank">http://www.GridWay.org</a><br>
OpenNebula Virtual Infrastructure Engine: <a href="http://www.OpenNebula.org" target="_blank">http://www.OpenNebula.org</a><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br>Regards,<br>Shashank Rachamalla<br>