Scaling and Auto-scaling

MonkOS enables easy manual and automatic scaling of any kit.

To set kit’s scale select it and go to Scale tab in Kit Configuration panel.

You can manually set the number of replicas at any point by changing the number in the Replicas field and clicking Save changes to confirm. Current replica count is indicated on the kit’s top bar.

To set up Auto-scaling, enable it by flipping the switch next to Auto Scale. This reveals more options.

Auto-scaling is always constrained by a range between minimum and maximum replica counts. Set the range using two inputs in the Replicas section - left for minimum replica count, right for maximum replica count.

The Basic Metrics section allows to set resource utilization targets, it is possible to set multiple targets at once, one per metric: CPU, RAM and Disk. When a target is set monk will replicate all containers belonging to the scaled kit dynamically by observing kit’s usage of selected resource across the cluster.

The number of replicas is computed using the following formula per metric:

desiredReplicas = ceil(currentReplicas * (currentMetricValue / desiredMetricValue))

If multiple targets are enabled at once the autoscaler will average out the desiredReplicas calculated for each metric. Number of replicas will not fall below or above limits set in Replicas section.

Replica placement is fully automated and will place replicated containers on separate nodes when possible, this spreads the load evenly.

All network connections to and from replicas are handled transparently spreading the traffic evenly across replicas.

For example: Set max replicas to 3 and flip the switch next to CPU usage, set the slider to 50%. Confirm by clicking Save changes.

MonkOS autoscaler features target-oriented and algorithmic scaling modes, however only the target-oriented method is available in the GUI at this time. For more autoscaling options see the Monkscript documentation.

To disable auto-scaling just flip the switch nest to Auto Scale and confirm by clicking Save changes.