Autoscaling automatically adjusts an Atlas cluster tier, storage capacity, or both in response to use. Scaling down reduces provisioned capacity to help control costs while preserving managed operation. section 10.0 section 10.2.3
Configure autoscaling around these operational boundaries:
- Set minimum and maximum cluster sizes to constrain automatic tier changes. section 10.2.3
- Allow scaling down when lower capacity is appropriate; Atlas performs adjustments on a rolling basis without downtime. section 10.2.3 section 10.7
- Expect Atlas to evaluate CPU and memory use, while remaining within the configured range and cluster class. section 10.2.3
- Check configuration defaults: UI-created clusters enable cluster-tier autoscaling by default, whereas API- or CLI-created clusters disable it by default. section 10.2.3
Bounded tier scaling
A cluster currently running at M50 can use a minimum of M30 and a configured maximum to limit automatic tier changes; enabling scale-down permits Atlas to reduce capacity when use falls. section 10.2.3
(memoryTotal - (memoryFree + memoryBuffers +
➥memoryCached))/(memoryTotal) * 100
Review the memory-use calculation Atlas uses when evaluating cluster-tier scaling decisions. section 10.2.3
