Last Updated: July 21, 2025 To fine-tune how Job Queue autoscaling works in SaladCloud, you can adjust the following settings. These parameters control how your container groups automatically scale based on the number of jobs in your SaladCloud job queues:

​
Desired Queue Length

  • Description: The target number of jobs per container instance that triggers the Job Queue autoscaler to adjust the number of container instances.
  • Range: Must be between 1 and 100.
  • Example: If you set this to 20, the system will aim to maintain 20 jobs per instance before scaling up or down.

​
Minimum Replicas

  • Description: The minimum number of container instances that the Job Queue autoscaler will scale down to.
  • Range: Must be between 0 and 100.
  • Example: If you set this to 3, the system will never scale below 3 running instances, even during periods with no jobs in the queue. If set to 0, the system will scale down to zero instances when there are no jobs in the queue, which means you do not pay for any resources when the queue is empty. However, this will result in cold starts when new jobs are added to the queue. A cold start happens when the system needs to start new containers, potentially causing delays as the instances are brought online.

​
Maximum Replicas

  • Description: The maximum number of container instances that the Job Queue autoscaler will scale up to.
  • Range: Must be between 1 and 500, or the maximum instance quota assigned to the project, whichever is lower.
  • Note: The maximum allowed value depends on the project instance quota. Ensure that your quota supports the desired number of replicas.
  • Example: If set to 100, the system will never spin up more than 100 instances, even if the job queue continues to grow.

​
Period (Seconds)

  • Description: The time interval in seconds during which the Job Queue autoscaler checks the queue length and applies the scaling formula.
    • Range: Must be between 15 and 1800 seconds.
  • Example: If set to 30, the job queue is checked every 30 seconds, and scaling decisions are made at that frequency.

​
Maximum Upscale Per Minute (Optional)

  • Description: The maximum number of instances that can be added per minute by the Job Queue autoscaler to handle increasing job volume.
  • Range: Must be between 1 and 100.
  • Example: Setting this to 20 means that the autoscaler can spin up a maximum of 20 new instances each minute to accommodate increasing job load.

​
Maximum Downscale Per Minute (Optional)

  • Description: The maximum number of instances that can be removed per minute by the Job Queue autoscaler when job volume decreases.
  • Range: Must be between 1 and 100.
  • Example: If set to 10, no more than 10 instances will be removed per minute during downscaling, ensuring a smooth reduction in resources.
  • Note: When scaling down, the system might remove nodes that are still processing jobs. To prevent job loss, ensure that your application can handle graceful shutdowns and that the Job Queue Worker completes current HTTP requests before termination.
Refer back to the Job Queue Autoscaling Overview or proceed to Set Up Job Queue Autoscaling.