The project is an application that implements the ability to dynamically change the number of service instances under high load. The application receives all services that have the swarm.autoscale label enabled, calculates the average value of the CPU utilization and, based on this, either increases the number of instances or decreases it.
Currently, only the CPU is used for autoscaling in the project. By default, if the CPU load reaches 85%, the service will scale, if it reaches 25%, it will be scaled down. But the minimum and maximum values of CPU utilization can be changed through environment variables.
Also, for each service, you can set the maximum and minimum number of replicas to prevent a situation with an uncontrolled increase in the number of replicas (or too much decrease)
- Deploy Swarm Autoscaler using
swarm-deploy.ymlfrom this repository - Add label
swarm.autoscale=truefor services you want to autoscale.
deploy:
labels:
- "swarm.autoscale=true"For better resource management, it is recommended to add resource constraints to your service. Then it will definitely not eat more resources than necessary, and auto-scaling will work much better and will save resources in idle time.
deploy:
resources:
limits:
cpus: '0.50'The application is configured through environment variables
| Setting | Default Value | Description |
|---|---|---|
AUTOSCALER_MIN_PERCENTAGE |
25 | minimum service cpu utilization value in percent (0-100) for decrease replicas |
AUTOSCALER_MAX_PERCENTAGE |
85 | maximum service cpu utilization value in percent (0-100) for increase replicas |
AUTOSCALER_DNSNAME |
tasks.autoscaler |
swarm service name for in stack communication |
AUTOSCALER_INTERVAL |
300 | interval between checks in seconds |
AUTOSCALER_DRYRUN |
false | noop mode for check service functional without enable inc or dec service replicas count |
Services in docker swarm are configured via labels
| Setting | Value | Default | Description |
|---|---|---|---|
swarm.autoscale |
Boolean | false |
Required. This enables autoscaling for a service. Anything other than true will not enable it |
swarm.autoscale.min |
Integer | 2 |
Optional. This is the minimum number of replicas wanted for a service. The autoscaler will not downscale below this number |
swarm.autoscale.max |
Integer | 15 |
Optional. This is the maximum number of replicas wanted for a service. The autoscaler will not scale up past this number |
swarm.autoscale.disable-manual-replicas |
Boolean | false |
Optional. Disable manual control of replicas. It will no longer be possible to manually set the number of replicas more or less than the limit. Anything other than true will not enable |
swarm.autoscale.percentage-max |
Integer | AUTOSCALER_MAX_PERCENTAGE |
Optional. Custom maximum service cpu utilization for increase replicas |
swarm.autoscale.percentage-min |
Integer | AUTOSCALER_MIN_PERCENTAGE |
Optional. Custom minimum service cpu utilization for decrease replicas |
swarm.autoscale.decrease-mode |
String | MEDIAN |
Optional. Service utilization calculation mode to decrease replicas. Modes: MEDIAN, MAX |