Live vCPU & RAM Scaling

Overview

Live vCPU and RAM scaling lets you adjust the compute resources of a running instance without rebooting. Changes take effect immediately while the instance continues serving traffic. The Dashboard provides a slider-based dialog for intuitive scaling within the bounds configured on the instance’s flavor.

Xloud-Developed — Bidirectional CPU hotplug (add AND remove vCPUs) and combined memory scaling (balloon + DIMM + virtio-mem) are developed by Xloud and ship with XAVS / XPCI.

Prerequisites

Instance must be in Active status
The instance must not be locked
The flavor must have hot-add enabled (hw:cpu_min or hw:mem_min extra specs) — if you do not see Adjust Resources in the actions menu, the flavor does not support live scaling

Scale vCPU or RAM

Dashboard
CLI

Open the Resource Adjustment dialog

Navigate to Compute > Instances. Click the More dropdown on the instance row, then select Adjust Resources under the Configuration Update group.

This action only appears for instances in Active status whose flavor has hot-add enabled. If you do not see it, contact your administrator to configure a hot-add-enabled flavor.

Adjust resources with sliders

The dialog shows the current state and adjustment controls:

Field	Description
Instance	Instance name (read-only)
Current Info	Current vCPU count and memory allocation (read-only)
vCPUs	Slider input — drag or type to set between minimum and maximum vCPUs
Memory (GiB)	Slider input — adjust between minimum and maximum memory in 0.25 GiB steps
Make permanent	Checkbox — persist the configuration after a soft reboot (default: checked)

The dialog fetches the live instance status to determine the current and allowed resource ranges. Color-coded hints indicate the scaling method:

Color	Meaning
Green	virtio-mem — fully adjustable in both directions
Yellow	No balloon driver detected — limited memory scaling
Blue	DIMM hotplug information

The vCPU slider step aligns to the CPU thread count configured in the flavor. For example, with 2 threads per core, vCPUs adjust in steps of 2.

Apply the change

Click Confirm. The change takes effect immediately — no reboot required.

Scaling down RAM while applications are using it may cause out-of-memory events inside the guest. Check application memory usage before reducing RAM.

Verify inside the instance

SSH into the instance and confirm the new resources:

Check vCPU count

lscpu | grep "^CPU(s):"

Check available RAM

free -h

The guest OS shows the updated vCPU count and memory — live scaling applied successfully with zero downtime.

Source credentials

source openrc.sh

Live resource adjustment uses the Xloud-developed Nova API extension. The openstack CLI does not have a native command for this operation.

curl -s -H "X-Auth-Token: $TOKEN" \
  "$NOVA_ENDPOINT/v2.1/servers/<INSTANCE_ID>/xloud-status" | python3 -m json.tool

curl -X POST -H "X-Auth-Token: $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"current_vcpus": 4}' \
  "$NOVA_ENDPOINT/v2.1/os-xloud-adjust/<INSTANCE_ID>"

curl -X POST -H "X-Auth-Token: $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"current_memory_mb": 8192}' \
  "$NOVA_ENDPOINT/v2.1/os-xloud-adjust/<INSTANCE_ID>"

curl -X POST -H "X-Auth-Token: $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"current_vcpus": 4, "current_memory_mb": 8192}' \
  "$NOVA_ENDPOINT/v2.1/os-xloud-adjust/<INSTANCE_ID>"

Comparison: Live Scaling vs. Flavor Resize

	Live vCPU/RAM Scaling	Flavor Resize
Downtime	None — instance stays running	Reboot required
Scope	vCPU and RAM only	vCPU, RAM, and disk
Direction	Up or down within flavor envelope	Up or down (disk: up only)
Confirmation step	No — instant	Yes — must confirm or revert
Dashboard action	More > Configuration Update > Adjust Resources	More > Configuration Update > Resize
Use case	Responding to live load changes	Permanent tier change

Troubleshooting

Adjust Resources not visible in the menu

Scaling request rejected

The requested value is outside the allowed range. The minimum and maximum are determined by the flavor’s hw:cpu_min / hw:mem_min (minimum) and the flavor’s vCPUs / RAM (maximum). Try a value within the slider range.

Guest OS does not show new vCPUs

On older Linux kernels (before 4.15), new CPUs may need to be brought online manually:

Bring hotplugged CPUs online

for cpu in /sys/devices/system/cpu/cpu*/online; do echo 1 > $cpu; done

Modern kernels (4.15+) bring CPUs online automatically.

Memory change has no effect

The guest OS memory balloon driver may not be loaded. Check inside the guest:

Check balloon driver

lsmod | grep virtio_balloon

If not loaded: modprobe virtio_balloon and add virtio_balloon to /etc/modules for persistence.

Next Steps

Admin: Enable Live Scaling

Configure flavors with hot-add extra specs and resource bounds

Resize Instance

Standard flavor resize — change vCPU, RAM, and disk with a reboot

Flavors

View available flavors and their hot-add capabilities

Launch an Instance

Create an instance with a hot-add-enabled flavor

Core Services

Other Services

Live vCPU & RAM Scaling

Overview

Scale vCPU or RAM

Open the Resource Adjustment dialog

Adjust resources with sliders

Apply the change

Verify inside the instance

Comparison: Live Scaling vs. Flavor Resize

Troubleshooting

Next Steps

Admin: Enable Live Scaling

Resize Instance

Flavors

Launch an Instance

​Overview

​Scale vCPU or RAM

Open the Resource Adjustment dialog

Adjust resources with sliders

Apply the change

Verify inside the instance

​Comparison: Live Scaling vs. Flavor Resize

​Troubleshooting

​Next Steps

Admin: Enable Live Scaling

Resize Instance

Flavors

Launch an Instance

Overview

Scale vCPU or RAM

Comparison: Live Scaling vs. Flavor Resize

Troubleshooting

Next Steps