Atlas Infinite storage model (max storage limit, no auto-scaling)
Parent: Atlas storage auto-scaling and disk reclaim via compact · Published reference · snapshot 2026-10-23 · skill mongodb-cohort-733862797816
↓ Facts as markdownall context files
Source-anchored research for Atlas Infinite storage model (max storage limit, no auto-scaling). 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.
Structure and components
- Atlas Infinite stores data in a managed layer independent of compute. Storage grows with data up to the enforced maximum; users neither choose a per-node disk size nor configure or opt out of storage auto-scaling. The absence of an auto-scaling setting does not mean storage growth is manual. (high confidence) (verified as of 2026-10-23) [source]
How it works
- MongoDB Compass issue COMPASS-10970 records that dbStats/$collStats storageSize and freeStorageSize do not represent meaningful on-disk usage for Infinite. Its fix hides those storage-size displays; use Infinite logical-data monitoring when reasoning about its cap. (medium confidence) (verified as of 2026-08-24) [source]
Measurements and reference values
- During public preview, supported Atlas Infinite tiers share a ceiling of 128 TB of logical data per replica set. A user can set a lower maximum from 1 GB to 128 TB; leaving it unset allows up to 128 TB. Changing the compute tier does not increase this storage ceiling. (high confidence) (verified as of 2026-10-23) [source]
- During public preview, the documented maximum is 128 TB of logical data per cluster. An optional cap accepts 1 GB–128 TB, can be increased or decreased after creation, and defaults to 128 TB when unset. (high confidence) (verified as of 2026-10-23) [source]
Problems, failure modes and limitations
- Atlas blocks writes as logical data approaches the enforced maximum while continuing to serve reads. Its default warning alert fires above 90% of the maximum. Public-preview availability safeguards may throttle writes before the configured limit is reached. (medium confidence) (verified as of 2026-10-23) [source]
- The maximum is an admission limit, not a reserved disk allocation. In the API it is represented by replicationSpecs[].regionConfigs[].autoScaling.storageConfig.shardSizeLimitGB; the documented accepted range is 1 through 131072 GB. The nesting under autoScaling does not establish a configurable storage auto-scaling switch. (medium confidence) (verified as of 2026-10-14) [source]
- Terraform represents the Infinite storage limit as auto_scaling.storage_config.shard_size_limit_gb. The provider example sets database_edition to INFINITE and a 1024 GB limit while compute_enabled is false, illustrating separate configuration of compute auto-scaling and the storage cap. (medium confidence) (verified as of 2026-10-20) [source]
- Inspect effectiveReplicationSpecs[].regionConfigs[].autoScaling.storageConfig.shardSizeLimitGB for the enforced limit. The configured and enforced values can differ due to system-managed changes; a missing configured field does not establish unlimited capacity. (high confidence) (verified as of 2026-10-23) [source]
- Atlas blocks writes at the enforced storage limit while continuing reads. Its default warning alert triggers above 90% of the cap. Public-preview availability protection may throttle writes before the configured limit. (medium confidence) (verified as of 2026-10-23) [source]
- Changing the compute tier does not change the storage capacity ceiling; adding data within the limit does not require a tier upgrade. Throughput still depends on the tier, so independent capacity is not independent storage performance. (high confidence) (verified as of 2026-10-23) [source]
- The API stores the configured cap at replicationSpecs[].regionConfigs[].autoScaling.storageConfig.shardSizeLimitGB. Despite the autoScaling nesting, this is a data-size limit. Terraform exposes auto_scaling.storage_config.shard_size_limit_gb. (high confidence) (verified as of 2026-10-23) [source]
Comparisons and alternatives
- For the storage cap, logical data size means the uncompressed size of collections and indexes, rather than the physical storage Atlas provisions. Comparisons with Atlas Core configured disk capacity must preserve this difference in units. (medium confidence) (verified as of 2026-10-23) [source]
- Preview deployments support replica sets rather than sharded clusters. Consequently the API field named shardSizeLimitGB currently caps the single replica-set cluster; it does not establish preview support for adding shards. (high confidence) (verified as of 2026-10-23) [source]
Facts and statements
- Independent storage capacity does not imply independent storage performance: current product documentation ties throughput and IOPS to cluster tier and offers no provisioned-IOPS or IOPS-auto-scaling setting for Infinite. A larger dataset may therefore still require more compute for acceptable performance. (medium confidence) (verified as of 2026-10-23) [source]
- For storage diagnosis, first establish whether the deployment is Atlas Core, Atlas Infinite, or self-managed and record its version, topology, provider, and relevant capacity metric. This avoids treating Infinite logical-cap pressure as an attached-disk reclamation problem. (medium confidence) (verified as of 2026-10-23) [source]
- Independent launch reporting from TechTarget and CRN identifies Atlas Infinite as a public-preview architecture separating compute and storage. Neither article independently tests the storage cap or its write-blocking behavior. (medium confidence) (verified as of 2026-09-29) [source]
- The Weighted Average distinguishes logical capacity from configured disk and warns that capacity independence does not remove tier-bound throughput. Its storage-ceiling discussion derives from MongoDB documentation, not an independent capacity experiment. (low confidence) (verified as of 2026-09-30) [source]
Children
- No children recorded.