Change VM resource allocation ratios
search cancel

Change VM resource allocation ratios

book

Article ID: 339886

calendar_today

Updated On:

Products

VMware Integrated OpenStack

Issue/Introduction

  • Edit resource allocation
  • How does VIO determine CPU usage on a node (esx host)
  • Overcommit CPU / RAM  / DISK (Storage) on compute node
  • Disallow/disable resource overcommitment
  • Note: Openstack assumes storage is thick provisioned.  There is no way to change this.  

 

Environment

7.x

Cause

  • Openstack only knows about clusters in vSphere not individual esxi hosts.
  • Openstack keeps track of resource usage in its databases.   
    A query is made to vCenter for cluster level resources (total CPU/RAM/Disk) and then use those numbers and the allocation ratios to determine what resources are available to Openstack.

Resolution

  1. SSH to the primary node

  2. Edit the virtual nova configuration file:

# viocli update nova

  1. Add a new section header called DEFAULT and add the specific allocation ratio under it.

conf:
  nova:
    DEFAULT:
      cpu_allocation_ratio: 10
      ram_allocation_ratio: 1.5
      disk_allocation_ratio: 1.0
    neutron:
      metadata_proxy_shared_secret: .Secret:managedencryptedpasswords:data.metadata_proxy_shared_secret
    vmware:
      passthrough: true
      tenant_vdc: true

Note:  This is in YAML and formatting/indentations are extremely important. 
Note:  Change any of these allocation ratio values to 1.0 to disallow resource overcommitment. 

  1. Save the file to trigger LCM to update the nova services. 

  2. Observe the re-creation of the nova-api-* pods using the following methods:

    1. Review logs:

      # kubectl -n kube-system logs -f -lapp=helm

    2. Observe the nova-api* pods get deleted/created:

      # pods | grep -i nova-api

      openstack            nova-api-metadata-######                                1/1     Running             7          95d

      openstack            nova-api-osapi-#####                                    2/2     Running             6          95d

  3. Once the pods are both running, confirm the Openstack deployment is in a RUNNING status:

    # viocli get deployment

    PUBLIC VIP      PRIVATE VIP       HIGH AVAILABILITY

    ...

    OpenStack Deployment State: RUNNING

  4. Confirm the changes to the allocation ratios:

    1. Validate via the nova-api-osapi* pod:

      # osctl exec -it nova-api-osapi-##### /bin/bash

      # kubectl describe pod/nova-api-osapi-##### -n openstack

    2. Validate the CPU, RAM, Disk allocation ratios in the nova.conf file:

      # grep -i allocation_ratio /etc/nova/nova.conf

      Note:  If the changes aren’t applied, check the status of the osdeployment pod. If the pod stays in a init state for more than 15 seconds, the pod may need to be destroyed manually:

      # osdel po osdeployment-osdeployment1

  5. After the changes has been performed in the nova CR, repeat the same steps in the nova-compute CR:

    # viocli update nova-compute


Note: Changing the CPU allocation ratios will not affect existing instances. They will continue to run as is.

Additional Information

The default allocation ratios are as follows.

cpu_allocation_ratio = 10
ram_allocation_ratio = 1.5
disk_allocation_ratio = 1

To disallow any resource overcommitment:

cpu_allocation_ratio = 1.0
ram_allocation_ratio = 1.0
disk_allocation_ratio = 1.0

 

Test VIO workload placement:
# openstack allocation candidate list --os-placement-api-version 1.10 --resource VCPU=# --resource MEMORY_MB=#### --resource DISK_GB=##
Example:  openstack allocation candidate list --os-placement-api-version 1.10 --resource VCPU=1 --resource MEMORY_MB=2048 --resource DISK_GB=20 

+---+----------------------------------+--------------------------------------+----------------------------------------------+
| # | allocation | resource provider | inventory used/capacity |
+---+----------------------------------+--------------------------------------+----------------------------------------------+
| 1 | VCPU=1,MEMORY_MB=2048,DISK_GB=20 | ########-####-####-####-############ | VCPU=3/40,MEMORY_MB=1536/36862,DISK_GB=3/699 |
+---+----------------------------------+--------------------------------------+----------------------------------------------+

Obtain resource providers:
# openstack resource provider list

+--------------------------------------+-------------------------------------------------+------------+
| uuid | name | generation |
+--------------------------------------+-------------------------------------------------+------------+
| ########-####-####-####-############ | domain-c20.########-####-####-####-############ | 6
+--------------------------------------+-------------------------------------------------+------------+

View resource provider inventory
# openstack resource provider inventory list ########-####-####-####-############

+----------------+------------------+----------+----------+-----------+----------+-------+
| resource_class | allocation_ratio | max_unit | reserved | step_size | min_unit | total |
+----------------+------------------+----------+----------+-----------+----------+-------+
| VCPU           |             10.0 |        8 |        0 |         1 |        1 |    16 |
| MEMORY_MB      |              1.5 |    20479 |        0 |         1 |        1 | 40958 |
| DISK_GB        |              1.0 |      670 |        0 |         1 |        1 |  1199 |
+----------------+------------------+----------+----------+-----------+----------+-------+

Powered by Wolken Software