Virtual machine allocation

This page requires revision and updating

When a virtual machine is started for the first time, Abiquo selects the most suitable physical machine to hold the new virtual image instance.

General considerations

The allocation operation takes virtual machines one by one.
- When a virtual appliance is started, all its virtual machines are deployed in ascending alphanumerical order.
If the virtual machine is in the power off state, the hypervisor is already assigned, so when you power on again, it will not be allocated again (the same machine is used).
The allocation can fail due to a 3 minute timeout if there are problems checking the target machine state.

Candidate machines

Candidate machines are all the available machines in the infrastructure that are capable of running a certain virtual image. The following rules discard some machines to define the candidate set. The machines considered are:

machines located in the physical datacenter where the virtual datacenter contains the virtual appliance.
machines with the same hypervisor technology as the virtual datacenter containing the virtual appliance.
machines with a datastore enabled with enough free space to hold the virtual image disk file.

Rack selection

Network isolation at rack level

Virtual datacenters use a single rack, so all the virtual machines of all virtual appliances on the same rack are deployed on the same rack.

If there aren't any virtual machines deployed for the current datacenter then the most suitable rack determined by the VLAN Allocation Algorithm is used.

Machine resource limitations

Only machines with enough free resources can be used:

enough free CPU resources.
enough free RAM resources.
enabled datastore with enough free space to hold the virtual image.

CPU ratio

The cpuRatio allows CPU oversubscription of a machine. It affects the real CPU usage.

The cpuRatio is set to 1 by default and we recommend that you use the Workload Level Rule to specify the oversubscription.

When editing a machine you can oversubscribe the resources (discovery manager sets CPU to 1 but you can change it to 2).
See Workload Level Rule for information on how to modify the oversubscription parameters.

Machines in state MANAGED

Machines that are not in the state Managed are not used Compute in Datacenters are not used and the health of a machine is checked before selecting it, only MANAGED machines are considered. The physical machine and Hypervisor are ready to deploy virtual machines.

Enterprise reservation

The cloud administrator can reserve machines for use by a single enterprise. If a machine is reserved it will only be a candidate if the current virtual datacenter belongs to the same enterprise.See the Reservations section for more information about configuring reservations.

Enterprise exclusion rule (affinity)

Enterprises can be explicitly prevented from using physical infrastructure of other enterprises. Defining an enterprise exclusion rule will cause an enterprise to use the same machine for all its virtual machines.
exclusion rule is associative (if enterprise 'a' excludes 'b', 'b' also excludes 'a').
If the current enterprise is covered by an exclusion rule, all the machines with virtual machines owned by the excluded enterprise will not be candidates.
See Restrict Shared Server Rules for more information about configuring these rules.

Workload level rule

See Workload Level Rule for more information about configuring these rules.

Resource allocation limits

Target machine selection

Fit policy rule

See Load balance for more information about configuring these rules.

Progressive

Performance

Check total allocated resources on infrastructure for the Resource Allocation Limits.
Only selects physical machines in the datacenter associated with the virtual datacenter of the virtual appliance.
- These only include physical machines with the appropriate virtualization technology for the given VM template.
Filter the physical machines with enough free resources to fit the virtual machine requirements.
If more than one physical machine can be selected the one with less capacity is used, so as to fully utilize each physical machine.
Finally there is a check process to ensure that the selected physical machine is running, if not, the machine's state is changed and the allocation process starts again.

Common deployment errors

When deploying a virtual appliance or a single virtual machine the user may see the following errors related to the scheduler logic.

Type ____________	Description	Solutions __________________________
Not enough resources	In the virtual appliance's datacenter, there is no physical machine with enough free resources for the virtual machine's requirements.	Add more physical machines. Increase the resources available in a physical machine. Power off some virtual machines.
Resource allocation limits exceeded (Hard limits)	The Resource Allocation Limits configured for the infrastructure do not allow the total resource utilization with the virtual machine requirements.	Power off some virtual machines or change the configured limits (0 for unlimited). Increase the hard limits for the enterprise.
Resource allocation limits exceeded (Soft limits)	The Resource Allocation Limits configured for the infrastructure allow the total resource utilization with the virtual machine requirements but alerts users that their resources are reaching the hard limit	Power off some virtual machines or change the configured limits (0 for unlimited). Increase the soft limits for the enterprise.