Overview
Flavors define the virtual hardware profile for instances — vCPU count, memory allocation, root disk size, and optional hardware extensions. Every instance must be launched with a flavor. Administrators control the flavor catalog available to all tenants, and can create private flavors restricted to specific projects for tailored workload sizing.Prerequisites
- Admin credentials sourced from
admin-openrc.sh openstackCLI installed and configured
Default Flavors
The following flavors are available in a standard Xloud Compute deployment.| Flavor | vCPUs | RAM | Root Disk | Use Case |
|---|---|---|---|---|
m1.tiny | 1 | 512 MB | 1 GB | Lightweight agents, minimal workloads |
m1.small | 1 | 2 GB | 20 GB | Development, testing |
m1.medium | 2 | 4 GB | 40 GB | General-purpose applications |
m1.large | 4 | 8 GB | 80 GB | Web servers, mid-tier services |
m1.xlarge | 8 | 16 GB | 160 GB | Databases, compute-intensive workloads |
Create, Modify, and Delete Flavors
- Dashboard
- CLI
Open flavor management
Log in to the Xloud Dashboard (
https://connect.<your-domain>) and navigate to
Admin → Compute → Flavors.Create a flavor
Click Create Flavor and fill in the required fields:
Click Create Flavor to save.
| Field | Description |
|---|---|
| Name | Unique identifier (e.g., c4.2xlarge) |
| vCPUs | Virtual CPU count |
| RAM (MB) | Memory in megabytes |
| Root Disk (GB) | Size of the instance root disk |
| Ephemeral Disk (GB) | Optional local scratch disk, discarded when instance is deleted |
| Swap (MB) | Optional swap partition size |
| Public | Toggle off to restrict access to specific projects |
The new flavor appears in the flavor list and is immediately available for instance launches.
Modify extra specs
Click the flavor name to open its detail view. Navigate to the Extra Specs
tab to add hardware hints such as CPU pinning policy, huge page requirements,
or NUMA topology constraints.Click Add Extra Spec, enter the key and value, and click Save.
Extra Specs Reference
Extra specs are key-value properties on a flavor that control hypervisor-level behavior. The following table lists commonly used extra specs.| Key | Values | Description |
|---|---|---|
hw:cpu_policy | dedicated, shared | dedicated pins vCPUs to exclusive physical threads |
hw:cpu_thread_policy | prefer, isolate, require | Controls NUMA thread sibling usage |
hw:mem_page_size | small, large, 1GB, 2MB | Requests huge pages for the instance |
hw:numa_nodes | Integer | Number of NUMA nodes to expose inside the instance |
hw:tpm_version | 1.2, 2.0 | Enables virtual TPM with specified version |
hw:tpm_model | tpm-tis, tpm-crb | Virtual TPM hardware model |
os:secure_boot | required, optional, disabled | Controls UEFI Secure Boot policy |
pci_passthrough:alias | <alias>:<count> | Requests PCI device passthrough by alias name |
hw:cpu_max_sockets | Integer | Max vCPU socket ceiling for live scaling |
hw:cpu_max_cores | Integer | Cores per socket for live scaling envelope |
hw:cpu_max_threads | Integer | Threads per core for live scaling envelope |
hw:mem_max_mb | Integer | Max RAM (MB) ceiling for live scaling |
aggregate_instance_extra_specs:<key> | Any string | Match against host aggregate metadata for scheduling |
Extra specs that reference hardware features (CPU pinning, huge pages, PCI passthrough)
require the corresponding scheduler filter to be active. See
Scheduling for the full filter list.
Flavor Metadata (User-Visible Tags)
Flavor (also called flavor tags) are display-layer annotations. Unlike extra specs, they do not influence the scheduler or hypervisor. Use them to communicate workload guidance, compliance designations, or SLA tiers to end users.- Dashboard
- CLI
Open flavor metadata editor
Navigate to Admin → Compute → Flavors. In the flavor row, open Actions → Update Metadata.
Add metadata entries
In the Available Metadata panel, type a custom key (or select from a registered namespace) and enter its value. Click + to add each entry.
| Example Key | Example Value | Purpose |
|---|---|---|
sla_tier | gold | Communicate SLA level to users |
workload | database | Hint recommended use case |
compliance | pci-dss | Tag for regulated workloads |
gpu | v100 | Indicate GPU availability |
cost_category | high-memory | Billing/chargeback classification |
Metadata vs Extra Specs
| Feature | Metadata (Tags) | Extra Specs |
|---|---|---|
| Affects scheduling | No | Yes (via scheduler filters) |
| Affects hypervisor | No | Yes (CPU pinning, huge pages, etc.) |
| Visible to users | Yes | Admin-facing |
| Use for | Display labels, compliance tags, cost categories | Hardware configuration, placement control |
Next Steps
Advanced Features
Use extra specs to enable CPU pinning, huge pages, GPU passthrough, and vTPM.
Quota Management
Set per-project resource ceilings that govern how many flavors a tenant can deploy.
Admin Guide
Return to the Compute Administration Guide index.