Practical Tutorial on Setting Time-based Scaling Rules

Recent Pages

Practical Tutorial on Setting Time-based Scaling Rules

Last updated: 2024-10-30 12:32:24

Based on the clear peaks and valleys in business activity over a certain period, you can choose between setting the execution frequency to Repeat or Execute only once. Configure scale-out rules and scale-in rules accordingly. When choosing Repeat, you can set the rule’s effective deadline by configuring the rule’s validity period, after which the scaling rules will no longer be triggered.
Example:
Your business activity starts increasing at 10 PM and begins to decrease at 6 AM daily, and this pattern is expected to last for one month. You can configure a time-based policy by setting up two scaling rules (one for scale-out and one for scale-in) or a single scale-out rule with scheduled termination.
Scaling Rule: Set to repeat daily. Configure the scale-out rule to be triggered at 10 PM each day for one month.
Scaling-down rule: Set to repeat daily. Configure the scale-in rule to be triggered at 6 AM each day for one month.
Scaling Rule + Scheduled Termination:scheduled termination: Set to repeat daily. Configure the scale-out rule to be triggered at 10 PM each day, with the allocated resources scheduled for 8 hours of use (equivalent to terminating at 6 AM the next day). This configuration will continue for one month. Support for Daily, Weekly, or Monthly repetition is available, so adjust based on your actual requirements. For more details on other rule configuration items and usage, see Setting Time-Based Scaling.
Note:
1. The timing for adding resources to the cluster above represents an ideal scene. In practice, the actual scale-out time depends on the number of resources requested. It is recommended to set the time rules at least 5 minutes earlier based on your needs.
2. During peak periods, resource contention may prevent the actual scale-out number from reaching the elastic target number of machines. It is recommended to enable the Resource Replenishment Retry Policy for your scale-out rule.
3. When the scale-in action is triggered, nodes may still be executing tasks. To avoid immediate release of the nodes, it is recommended that you enable graceful scale-in. For more details, see Graceful Scale-In.