NexaSphere DRS: How It Works and How to Configure It

[nexus]

How it works

NexaSphere DRS balances CPU and/or RAM load across the hosts of a cluster by automatically live-migrating VMs.

• Detection: at every scan, DRS checks the current CPU/RAM utilization of each host. If a host exceeds the policy threshold, the cluster is marked Imbalanced.
• Migration decision: to avoid unnecessary migrations caused by temporary spikes, DRS evaluates the host load over a historical window. The value used is the average of the samples at the start and at the end of the window (default: 6 hours).
• A VM is migrated only if the destination host stays below the threshold after the migration.

With the default settings, DRS reacts to sustained load and ignores short peaks. For faster responsiveness, for example during tests or a POC, the window can be shortened.

How to change the historical window

Run on the management node:

zstack-cli LogInByAccount accountName=admin password=<password>

# Check the current value (default: 21600 seconds)
zstack-cli QueryGlobalConfig category=drs name=drs.collectHostMetricDataDuration

# Set the window to 10 minutes
zstack-cli UpdateGlobalConfig category=drs name=drs.collectHostMetricDataDuration value=600

The change is applied immediately, without restart, and applies to all clusters. With a 600-second window, DRS reacts to sustained load within about 10-15 minutes.

To restore the default value:

zstack-cli UpdateGlobalConfig category=drs name=drs.collectHostMetricDataDuration value=21600