Autoscaling application servers in the public cloud is a popular design choice to provide elastic, on-demand scale for the applications while keeping resource costs in check.
Avi Load Balancer provides autoscaling capability based on metrics such as application load, connections, and so on. The scheduled scaling feature complements the autoscaling capabilities to provide a deterministic capacity during a specified period.
Use Case
Scheduled autoscaling allows you to create a policy for a duration, for example, two months, during which your application capacity can be maintained at a specified state, overriding the metrics-based autoscaling.
The feature provides deterministic capacity at these periods so that the application servers are already provisioned and ready for traffic, rather than waiting for metrics-based autoscaling to kick in. This reduces the lead time for application capacity to be available.
Example
An e-commerce provider uses autoscaling policies already to scale up/ down application servers based on incoming traffic (connections per second).
During the holiday shopping season between November through January, the company sees high capacity from afternoon to midnight.
By creating a scheduled autoscaling policy, the company can keep the application capacity at a high value, thus ensuring their application response times are optimum and there is no delay in servicing the added load.
Configuration
Scheduled autoscaling is configured within the existing Server AutoScale Policy, and attached to pool.
The scheduled_scalings configuration has the following sub-parameters:
Parameter |
Values |
Notes |
|---|---|---|
enable |
True / False |
Configure this to True to enable scheduled scaling. |
desired_capacity |
Value up to 1000 |
Desired number of servers during the scheduled time interval.This could result in scale in or scale out depending on the value entered. |
cron_expression |
Cron-like expression must be within the start_date and end_date. Supports a fixed integer for minutes and hours. |
A cron-like expression at which time the policy is activated for autoscaling_duration hours. |
start_date |
Date and Time in ISO8601 format. |
Date and Time in ISO8601 format. |
end_date |
Date and Time in ISO8601 format. |
End date and time for scheduled autoscaling. |
autoscaling_duration |
Integer, hours |
The duration in hours after cron criteria is met, for which the desired capacity will be maintained. |
schedule_max_step |
Integer |
The number of servers to simultaneously scale. Helps in faster provisioning. The autoscale policy’s scaleout_cooldown parameter is used as cooldown duration between successive scale-out events. |
Example Configuration
In the below configuration:
The policy is active between 1st October 2021 and 1st February 2022 (based on start_date and end_date).
The scheduled scaling must happen every day at 9:10 AM (based on cron_expression).
The capacity must be maintained at 50 servers (desired_capacity).
The ramp-up must happen in steps of 5 (schedule_max_step).
The ramp-up must be maintained for 2 hours (autoscaling_duration).
configure serverautoscalepolicy asp-example scheduled_scalings autoscaling_duration 2 index 1 #index assignment should happen after at least on other field enable #no enable if we want to disable desired_capacity 50 schedule_max_step 5 start_date 20211001T000000Z #timezone should be UTC end_date 20220201T000000Z cron_expression '10 9 * * *' save (save scheduled_scalings) save (save auto scale policy)
+-------------------------------------+------------------------------------------------+ | Field | Value | +-------------------------------------+-----------------------------------------------+ | uuid | serverautoscalepolicy-69ce3c2d-b401-4b9a-a834-08fbec8921d9 | | name | asp-example | | intelligent_autoscale | False | | intelligent_scaleout_margin | 20 | | intelligent_scalein_margin | 40 | | max_scaleout_adjustment_step | 1 | | max_scalein_adjustment_step | 1 | | scaleout_cooldown | 300 sec | | scalein_cooldown | 300 sec | | use_predicted_load | False | | delay_for_server_garbage_collection | 0 | | scheduled_scalings[1] | | | enable | True | | desired_capacity | 50 | | cron_expression | 10 9 * * * | | start_date | 20211001T000000Z | | end_date | 20220201T000000Z | | autoscaling_duration | 2 hours | | schedule_max_step | 5 | | tenant_ref | admin | +-------------------------------------+----------------------------------------------+
Considerations
Only one
scheduled_scalingconfiguration is supported for each autoscale policy.If
scheduled_scalingis disabled, the other fields such asmax_size,min_size,max_scaleout_adjustement_step,max_scalein_adjustment_step,desired_capacity, andschedule_max_stepwill start to control the pool server size.
