Based on the possible obvious peaks and troughs of your business within a certain period, select Repeat or Execute once as the execution frequency, and configure scale-out and scale-in rules. For repeated execution, you can set the expiration time of rule activation by configuring the rule validity period. After the validity period, the scaling rule is no longer triggered.
Example:
If your business load starts to increase at 22:00 and decrease at 06:00 every day and this is expected to last for one month, you can configure the By Time policy type and set two scaling rules (one for scale-out and one for scale-in) or one scale-out rule (scheduled termination).
Scale-out rule: It is executed daily. Configure the scale-out rule at 22:00 every day, lasting for one month.
Scale-in rule: It is executed daily. Configure the scale-in rule at 06:00 every day, lasting for one month.
Scale-out rule+scheduled termination: It is executed daily. Configure the scale-out rule at 22:00 every day. The batch of resources is used for 8 hours (which corresponds to 06:00 the next day), repeated daily for one month. Repeated execution supports Daily, Weekly, and Monthly. Set the frequency according to your actual needs. For other rule configuration items and usage instructions, see Setting Time-based Scaling. Attention:
1. The above time points for supplementing resources to the cluster are ideal. The actual time consumed for scaling out resources depends on the quantity per scale-out. It is recommended to set the time rule at least 5 minutes in advance based on your needs.
2. During peak order placement, scale-out may result in the actual number of scale-out machines failing to reach the elastic target quantity due to resource contention. It is recommended to enable the resource replenishment retry policy for scale-out rules.
3. Nodes may be running tasks when the scale-in action is triggered. To avoid immediate node release, it is recommended to enable graceful scale-in. For details, see Graceful Scale-In.