Hi Daniel,

Resource calculation will not be greatly affected by 4Mhz – you should be safe 
here. Keep in mind you are running with something like a 75% alert and 85% 
disable threshold on your clusters (see cluster settings / global settings) – 
and unless you increase the disable threshold to 100% you will never completely 
max out to the extent 4MHz will make a difference.

CPU overprovisioning is all about the overall amount of CPU resources available 
– not what speed each VM will receive.

Regards,
Dag Sonstebo
Cloud Architect
ShapeBlue

On 21/07/2017, 08:57, "daniel.herrm...@zv.fraunhofer.de" 
<daniel.herrm...@zv.fraunhofer.de> wrote:

    Hi Dag, Hi Ivan,
    
    thanks to both of you for your reply. I would like to build on the question 
of my colleague Christian.
    
    Can you elaborate what would happen if we’d just change the service 
offering in the database? From my understanding the main problem is, that 
during VM creation the old and slightly higher value of 1999MHz has been used 
for resource calculation, while only 1995MHz will be freed again when one of 
the existing VMs are destroyed. Is this correct, are there any other potential 
implications?
    
    If this is the only implication, this could be “corrected” with a slightly 
increased overprovisioning factor of the CPU resource. This might not be a very 
elegant solution, but changing the service offering is a hard problem as well, 
because we cannot easily reboot all VMs currently hosted by CS.
    
    Thanks and regards
    Daniel
    
    Am 21.07.17, 09:51 schrieb "Dag Sonstebo" <dag.sonst...@shapeblue.com>:
    
        Hi Christian – my twopence worth – for the sake of 4Mhz the DB change 
should be OK, it seems a bit overkill to create and replace all your service 
offerings just to accommodate your new 1995MHz hardware.
        
        Regards,
        Dag Sonstebo
        Cloud Architect
        ShapeBlue
        
        On 21/07/2017, 08:07, "Ivan Kudryavtsev" <kudryavtsev...@bw-sw.com> 
wrote:
        
            Hi. You just have to restart all affected VMs after offering change,
            because running VMs only get new resources after restart.
            
            It might be better to configure CPU overprovisioning in case you 
met system
            limits.
            
            21 июл. 2017 г. 13:59 пользователь 
<christian.nieph...@zv.fraunhofer.de>
            написал:
            
            > Hi Ivan,
            >
            > thanks für die quick reply.
            >
            > Would you mind elaborating somewhat further on the potential 
implications.
            > Can I avoid unfaire resource provisioning by modifiying all 
existing
            > service offerings equally, e.g. changing the CPU Speed of all 
offering from
            > 1999 to 1995 MHz?
            >
            > Cheers,
            > Christian
            >
            > > On 21. Jul 2017, at 03:37, Ivan Kudryavtsev 
<kudryavtsev...@bw-sw.com>
            > wrote:
            > >
            > > Hi, you can actually do it thru DB, but it can lead to several
            > > implications, like unfare resource provisioning. The better way 
is just
            > > delete the offering, create the new with the same name and 
switch all VMs
            > > either automatically or asking users.
            > >
            > > Have a good day.
            > >
            > > 21 июл. 2017 г. 2:55 ДП пользователь <christian.niephaus@zv.
            > fraunhofer.de>
            > > написал:
            > >
            > > Dear all,
            > >
            > > as there is no means to modify an existing Cloudstack Service 
Offering
            > > neither  via Cloudstack API nor with the GUI, I’m wondering 
what would
            > > happen if the CPU speed of the service offerings is changed 
directly in
            > the
            > > cloud DB (table service_offering). Does this have any impact on 
existing
            > > VMs? Would this be a valid way to modify an existing Service 
Offering?
            > >
            > > We did some very brief test and it seem to work fine, but 
before doing
            > the
            > > change in our production environment I’d like to know if anyone 
else has
            > > done something similar?
            > >
            > > The reason why I’m trying to do this is as follows:
            > > In all our Service Offerings for user VMs we have set the CPU 
Speed to
            > 1999
            > > Mhz. Unfortunately, the CPUs of our most recent hosts only 
provide 1995
            > > MHz, leading to the situation that no VM is deployed on these 
servers as
            > > the hosts do not have the proper cpu capability (speed 1995 is 
provided
            > but
            > > 1999 is required).
            > >
            > > Cheers, Christian
            > >
            > > PS: We’re still on Cloudstack 4.5.1
            >
            >
            
        
        
        dag.sonst...@shapeblue.com 
        www.shapeblue.com
        53 Chandos Place, Covent Garden, London  WC2N 4HSUK
        @shapeblue
          
         
        
        
    
    


dag.sonst...@shapeblue.com 
www.shapeblue.com
53 Chandos Place, Covent Garden, London  WC2N 4HSUK
@shapeblue
  
 

Reply via email to