Auto-scaling tier ranges
Parent: Atlas Cluster Topology and Tier Economics · Published reference · snapshot 2026-10-23 · skill mongodb-cohort-88cc8939dcad
↓ Facts as markdownall context files
Source-anchored research for Auto-scaling tier ranges. Run status: COMPLETED.
These notes link each claim to its source. Some citations identify archived research reports that are no longer hosted on this site. A published reference means the content is available; it does not certify independent review or accuracy.Read the editorial policy and follow the sources before relying on a claim.
How it works
- Atlas Core predictive scaling respects minimum and maximum instance sizes, scales upward by at most two tiers per event, and uses reactive scaling for downscaling. Eligibility requires M30 or higher and, when downscaling is enabled, a minimum bound of at least M30. (low confidence) (verified as of 2026-10-14) [source]
- Cloud Cost Room recommends setting bounds from actual demand and reviewing scaling events because higher tiers can persist after spikes. MongoDB likewise presents bounds as capacity and cost controls. This is operational guidance, not proof of a hard billing cap or guaranteed savings. (medium confidence) [source]
Problems, failure modes and limitations
- On Atlas Core, storage growth can raise the minimum auto-scaling tier when the configured minimum cannot accommodate the new disk configuration. (medium confidence) (verified as of 2026-10-14) [source]
- On Atlas Core, storage auto-scaling can increase the compute tier even when Cluster Tier Scaling is disabled. If the configured maximum cannot hold the new storage, Atlas raises that maximum to the smallest sufficient tier, scales to it, and disables automatic downscaling until manually re-enabled. (low confidence) (verified as of 2026-10-18) [source]
- If a smaller tier cannot support existing disk capacity or provisioned IOPS, Atlas Core refuses the downgrade. At the configured maximum it disables automatic downscaling; otherwise it raises the minimum to the current tier without disabling downward scaling. Storage growth can also make the former minimum invalid. (high confidence) (verified as of 2026-10-14) [source]
- A Crossplane provider report describes storage auto-scaling beyond initProvider.diskSizeGb causing repeated DISK_SIZE_GB_TOO_SMALL failures when reconciliation resubmits the old size. An October 15, 2026 comment says an Upjet update should resolve it; this is a reported integration failure, not proof that all current versions remain affected. (low confidence) (verified as of 2026-10-15) [source]
Facts and statements
- Atlas lets operators configure minimum and maximum cluster sizes. Compute auto-scaling ordinarily stays within that range; the minimum supports baseline capacity and the maximum constrains compute growth, subject to storage-driven exceptions. (high confidence) (verified as of 2026-10-14) [source]
- Atlas Core compute auto-scaling moves between tiers in the same class. It supports General and Low-CPU classes; it does not switch automatically between those classes or support the Local NVMe SSD class. (low confidence) (verified as of 2026-10-14) [source]
- For Atlas Core Gen2, the current tier and auto-scaling bounds must share a generation. M10 and M20 are generation agnostic: the maximum bound determines whether they scale into Gen1 or Gen2. Incompatible generation changes produce INVALID_ATTRIBUTE. (low confidence) (verified as of 2026-10-14) [source]
- Stacktape documents autoScaling.minClusterTier and maxClusterTier, plus disableDiskScaling and disableScaleDown. Its example uses an M10–M30 range with both disabling flags true; configuring a range does not itself ensure downward scaling is enabled. (low confidence) (verified as of 2026-10-23) [source]
- Infrastructure reconciliation can undo Atlas scaling: a Kubernetes Operator v1.2.0 issue describes M30 scaling to M40 and then reverting to M30 during an unrelated settings update; a Terraform provider issue independently documents reapplying a configured instance size after auto-scaling. (medium confidence) (verified as of 2024-12-27) [source]
- Atlas exposes compute enabled, scaleDownEnabled, minInstanceSize and maxInstanceSize controls in its API configuration. Terraform exposes corresponding compute_enabled, compute_scale_down_enabled, compute_min_instance_size and compute_max_instance_size settings. (high confidence) (verified as of 2026-10-14) [source]
- With Terraform use_effective_fields, a new range that excludes the current effective instance size fails Atlas validation, even if instance_size changes in the same apply. The provider documents disabling auto-scaling and applying a size inside the intended range, then re-enabling auto-scaling with the new bounds. (low confidence) (verified as of 2026-10-20) [source]
- MongoDB recommends pre-scaling programmatically for unexpected one-time spikes: reactive scaling has a delay and predictive scaling targets recurring patterns. An allowed higher maximum alone does not guarantee capacity will arrive before a sudden index-build workload. The index-build application is an inference from that guidance. (low confidence) (verified as of 2026-10-14) [source]
Children
- No children recorded.