Pod autoscaling

Autoscaling by Horizontal Pod Autoscaling (HPA)

Note: "metrics-server" add-on must be installed to use pod-autoscaling Ressource request must be configure on the desired ressource (deployment, replica set,...)

Policy Name	Name of the policy to be created. Set this parameter as required.	Default values
Associated Workload	Workload with which the HPA policy is associated.
Cluster Name	Cluster to which the workload belongs.
Cooldown Period	Set a cooldown period, within which scaling will not be triggered again when a policy is triggered.	For scale-down = 5m For scale-up = 3m
Namespace	Namespace to which the workload belongs.
Pod Range	Enter the minimum and maximum numbers of pods. When a policy is triggered, the workload pods are scaled within this range.
Rules	Metric: You can select CPU usage or Memory usage. Usage = CPUs or memory used by pods/Requested CPUs or memory. Expected Value: Enter an expected value. The expected value of the selected metric works with the threshold. The actual number of pods to be scaled is calculated using the following formula: Expected Value/Threshold x Number of current pods Threshold: When the metric value is less than the scale-in threshold, scale-in is triggered. When the metric value is greater than the scale-out threshold, scale-out is triggered. You can click  Add Rule again to add more scaling policies.

Basic template:

Policy Name: mem-cpu

Cluster Name: < cluster-name >

Namespace: < namespace >

Associated Workload: < deployment, replica set,... >

Pod Range: 1 to 10 (must be adapted to the needs)

Cooldown Period: scale-down = 5m, scale-up = 5m

Rules:

Metric	Expected Value	Threshold
CPU usage	50%	Scale down 33% Scale up 80%
Memory usage	50%	Scale down 33% Scale up 80%

Autoscaling by AOM (Not tested for the moment)

Metric-based policy

Parameter	Description
Policy Name	Enter the name of the scaling policy.
Policy Type	Set this parameter to Metric-based policy.
Metric	Set the metrics that describe the resource performance data or status. CPU Usage: indicates the CPU usage of the measured object, that is, the percentage of the CPU cores actually used by the measured object to the total CPU cores that the measured object has requested.  Physical Memory Usage: indicates the percentage of the physical memory size used by the measured object to the physical memory size that the measured object has applied for.
Trigger Condition	You can set this parameter based on CPU Usage or Physical Memory Usage.
If you set this parameter to the average physical memory usage > 70%, the scaling policy will be triggered when the average memory usage exceeds 70%.
Monitoring window	Metric statistics period. Select a value from the drop-down list box.
If the parameter is set to 20s, metric statistics is collected every 20 seconds.
Threshold Crossings	If the parameter is set to 3, the action is triggered if threshold is reached for three consecutive measurement periods.
Action	Action executed after a policy is triggered. Two actions are available: add or reduce pods.

Scheduled policy

Parameter	Description
Policy Name	Enter the name of the scaling policy.
Policy Type	Set this parameter to Scheduled policy.
Trigger Time	Time at which the policy is enforced.
Action	Action executed after a policy is triggered. Three actions are available: add pods, reduce pods, and set the number of pods.

Periodic policy

Parameter	Description
Policy Name	Enter the name of the scaling policy.
Policy Type	Set this parameter to Periodic policy.
Time Range	Specify the time for triggering the policy.
Action	Action executed after a policy is triggered. Three actions are available: add pods, reduce pods, and set the number of pods.