Compute Service Advance Configuration

You may occasionally need to adjust the default Compute Service configuration to enable or disable specific settings. This document describes the steps to do that.

What is nova_override.conf

nova_override.conf is the configuration file used by Private Cloud Director compute service to store all the default configuration values.

You can use this override file to:

• Adjust logging levels or enabling more verbose logging by enabling debug mode. (Mostly commonly used case).

• Customize compute driver options.

• Modify quota defaults for specific tenants.

The nova_override.conf file is located in the following directory.

Bash

/opt/pf9/etc/nova/conf.d/nova_override.conf

What to know before

• Updating nova_override.conf file is an advance operation and should only be performed by Administrators with complete understanding of the property being edited and the possible impact. We highly recommend that you do this only under guidance from the Platform9 support or solution architect teams.

• Any changes made to the config file persist across upgrades to Private Cloud Director, ensuring that your customized settings remain intact across upgrades.

• Any change to the config file requires service restart for the changes to take effect.

• The nova_override file must be edited on each hypervisor host to ensure that the changes to apply consistently across all hosts.

Important

Updating nova_override.conf file is an advance operation and must only be performed by Administrators. We highly recommend doing this only under Platform9 support guidance.

Restart Service

After editing the nova_override.conf file, you must restart the compute service for the changes to take effect. Run the following command on the host:

Bash

systemctl restart pf9-ostackhost

Common Settings

For a complete list of Compute Service configuration options, refer to nova.conf.

Below we have some common scenarios and configurations options you may wish to edit.

Enable Debug Logging

This is the most commonly used option. You can turn this on when troubleshooting issues with VM creation or VM volume attachment. Note that enabling this will generate highly verbose logs and may exhaust disk space on the host. We recommend disabling it in production after troubleshooting.

Bash

[DEFAULT]
debug = true  # Enable detailed logging (useful for troubleshooting)
log_dir = /var/log/nova  # Log file storage location

Scheduler Configuration

This parameter determines the number of times the virtual machine scheduler will try to provision a virtual machine instance before failing. Only adjust this under specific guidance from Platform9 support team.

Bash

[scheduler]
max_attempts = 5  # Number of retries for scheduling instances

Resource Over-Provisioning

cpu_allocation_ratio and ram_allocation_ratio define the CPU and memory over commitment ratios at the hypervisor host level. The defaults are 1:5 for RAM and 1:16 for CPU over commitment.

Bash

[DEFAULT]
cpu_allocation_ratio = 16.0  # Overcommit CPU resources (Default: 16)
ram_allocation_ratio = 1.5  # Overcommit RAM resources (Default: 1.5)

Reserving Resources for System Overhead

You can reserve CPU and Memory resources at each hypervisor level that will not be allocated to virtual machines that get provisioned on that hypervisor. This is crucial for production environments to ensure that the hypervisor remains functional even when the virtual machines provisioned on it may be under heavy load. __reserved_host_memory_mb represents memory in MB to be reserved for system overhead and reserved_host_memory_mb represents host cpus to be reserved. The default values for these are 512 for memory and 0 for CPU.

Bash

reserved_host_memory_mb = 512  # host memory to reserve for system overhead 
reserved_host_cpus = 0 # host cpus to reserve for system overhead

Info

Kernel-level overcommit checks during VM creation are disabled by setting /proc/sys/vm/overcommit_memory to 1. This enables dynamic VM creation and hotplug operations. After host upgrades, the file is automatically reset to 1 even if it was manually modified.

Advance Settings

Warning

These settings impact VM operations and performance. Modify only if necessary and test changes in a non-production environment.

Compute Driver Configuration

Bash

[DEFAULT]
compute_driver = nova.virt.libvirt.LibvirtDriver

Do not change unless you are configuring a different virtualization technology (e.g., VMware, Xen, KVM).

Networking Configuration

Bash

[neutron]
service_plugins = router,metering

Changing networking parameters without proper testing may break VM connectivity.

Disk & Storage Settings

Bash

[libvirt]
images_type = qcow2  # Default disk image format
live_migration_flag = VIR_MIGRATE_LIVE

Modify only if you are implementing specific storage optimizations.

CPU Mode and Model Configuration

Bash

[libvirt]
# Specifies the CPU mode for virtual machine instances.
cpu_mode = custom
# Defines the custom CPU model(s) to be used with the hypervisor.
cpu_models = Broadwell-noTSX-IBRS

This configuration is typically used when you need consistent CPU features across different hypervisors.

Enable Storage Multipath

1. Make sure that service multipathd is running on all hosts.

2. Use the following config option to enable storage multipath support.

Bash

volume_use_multipath = True