Skip to main content

Autoscaling storage

Introduced in v.1.50.

CCX supports autoscaling of storage for datastores. Autoscaling is a feature that automatically adjusts the storage capacity of a datastore based on the current disk usage.

End-users have to activate this feature on a per datastore basis by setting autoscale storage to true in the UI. It is disabled by default.

If enabled for a datastore, and the datastore has any node that exceeds a defined threshold, the cluster will automatically scale up by a defined percentage.

This disk increase may be repeated multiple times if the disk usage continues to exceed the threshold after each increase.

If the end-user has enabled notifications for a datastore then the designated email addresses will receive a notification when the datastore storage is scaled up. Email notifications is configured by the end-user under the Datastore Settings.

Configuration

The threshold and percentage values are defined in helm values as autoscaling.storage.threshold and autoscaling.storage.percentage.

The default autoscaling.storage.threshold is 70 and autoscaling.storage.percentage is 20.

To disable storage autoscaling globally (users won't be able to enable it for their datastores) set autoscaling.storage.enabled: false

End-user enablement

The end user must enable the autoscale storage feature per datastore, see Datastore Settings for more information.

Alarm notifications

When the autoscaling.storage.threshold has been reached for a storage volume in a datastore, then AlertManager/VM_alert will raise an alert called HostAutoScaleDiskSpaceReached. It may be look like this:

#76390: [Prometheus]: [FIRING:1] HostAutoScaleDiskSpaceReached /dev/mapper/VG_data-data 1.2.3.4:9100 warning (<datastore uuid>>
Disk usage on instance 1.2.3.4:9100 is above the autoscale threshold. Current usage: 70.01082600415612% Disk autoscale threshold has been reached on instance 1.2.3.4:9100 70.01082600415612%
Alerts Firing:
Labels:
- alertname = HostAutoScaleDiskSpaceReached
- ClusterID = NNN

If the end-user has not activated the autoscaling, then the alarm will be triggered repeatedly.