Please see my comments below. Thanks.
-- Evgeniya On Thu, Sep 18, 2014 at 9:43 AM, Andre Moruga <[email protected]> wrote: > Looks good to me, thanks Dmitry > > > > > > *From:* Dmitry Savenkov [mailto:[email protected]] > *Sent:* Thursday, September 18, 2014 9:42 AM > *To:* Andre Moruga > *Cc:* [email protected]; Juan Victor > Izquierdo; Evgeniya Shumakher > *Subject:* OpenStack APS - Excessive Resource Usage + Resellers & > Supertenants > > > > Andrey: > > > > Following our recent meeting, I've put together the statement of the > problem, which can be broken down into the following two parts: > > > > *1) OpenStackDeployment + Lack Of Resources:* > > > > *Initial Data:* > > > > 1.1) We have an OpenStack deployment (cluster) that was created by means > of deploying a number of nodes (it's comprised of a set of control, > compute, and storage nodes). > > 1.2) All the storage nodes have some RAM, CPU (cores), and Disk capacity. > This physical capacity will be used by VMs once they are created/launched. > > 1.3) One physical CPU core could map to a number of virtual ones (say 1 > physical core would enable the creation of 8 virtual ones) > > > > *Problem:* > > > > 1) Figure out if it's possible to query the current resource utilization > from an OpenStack cluser (all compute nodes) (*Environment *is the right > OpenStack term) so that we can display it on the DataCenter statistics tab. > > 2) Investigate the possibility of catching/handling events having to do > with the lack of resources (Once we have run out of CPU, RAM, or Disk > Space, we need to take respective actions/alert the user). > Ceilometer does it for virtual resources. It's deployed and configured for KVM. It has Alerts which work like a events you mentioned. Zabbix can be installed to collect information about physical resources. It needs to be installed and properly configured. > > > *Constraints:* > > > > Since KVM/libvirt may allow the user to create more virtual machines than > the cluster can handle (in terms of memory/CPU/Disk), we may have to come > up with some empirical formulae on average resource thresholds (CPU, RAM, > Disk) > There are some rules. We apply them during a deployment planning. Please check: https://www.mirantis.com/blog/openstack-hardware-bom-calculator/ https://www.mirantis.com/openstack-services/bom-calculator/ > > > *2) OpenStack APS + Resellers:* > > > > *Statement of the Problem:* > > > > Find out if OpenStack enables the creation of a user that could be called > 'SuperTenant'. In other words, that user would have privileges sufficient > for him to connect to the main provider (as a reseller) and create > subtenants that would belong to him, where some limited supertenant quota > applies (all the subtenants cannot consume more resource than allocated). > > > > That supertenant should have credentials enabling him to connect to a > DataCenter and create those subtenants. > Looks like in OpenStack terms it's an admin, here you can find details: http://docs.openstack.org/openstack-ops/content/projects_users.html > > > Andrey, please feel free to comment on whatever you want, correct > something, add what is being missed. > > > > This will be tackled tomorrow. The first point is being already discussed > with some OpenStack specialists. > > > > Thank you! > > > > Kind Regards, > > Dmitry. > > > > > > > -- -- Regards, Evgeniya Mirantis, Inc Mob.phone: +7 (968) 760-98-42 Email: [email protected] Skype: eshumakher [image: Register today: OpenStack Silicon Valley | 16 September 2014] <http://www.openstacksv.com>
-- Mailing list: https://launchpad.net/~openstack-aps-package-dev Post to : [email protected] Unsubscribe : https://launchpad.net/~openstack-aps-package-dev More help : https://help.launchpad.net/ListHelp
