Atlas storage auto-scaling triggers, cooldowns and write blocking
Parent: Oplog sizing under storage auto-scaling · Published reference · snapshot 2026-10-31
↓ Facts as markdownall context files
Source-anchored research for Atlas storage auto-scaling triggers, cooldowns and write blocking. 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
- After disk auto-scaling triggers, subsequent automatic and manual disk scaling may be unavailable for a few hours. This general policy statement does not specify a universal fixed cooldown duration. (low confidence) (verified as of 2026-10-14) [source]
- Atlas may increase the cluster tier to accommodate disk growth even when cluster-tier scaling is disabled; storage growth can therefore alter compute capacity and cost. (low confidence) (verified as of 2026-10-14) [source]
- For Atlas Core, critically low free disk space on the primary blocks replica-set writes while reads continue. Deletes are blocked too because they can temporarily consume disk. Sharded clusters block writes routed to the affected shard. (low confidence) (verified as of 2026-10-14) [source]
- Crossplane provider-mongodbatlas issue 191 reported repeated DISK_SIZE_GB_TOO_SMALL failures when reconciliation reapplied initProvider.diskSizeGb after Atlas had auto-expanded the disk. The author traced it to Upjet TPF initProvider merging; the issue was closed as completed on 2026-10-21 after an Upjet update. (low confidence) (verified as of 2026-10-21) [source]
Measurements and reference values
- Atlas Core enables storage auto-scaling by default; 90% disk usage on any cluster node triggers expansion targeting 70% utilization on AWS, Azure and GCP. Storage does not automatically shrink. (medium confidence) (verified as of 2026-10-14) [source]
- Documented free-space block/unblock thresholds are below 600 MiB/above 900 MiB for 8–20 GiB disks, below 4%/above 6% for 20 GiB–1.25 TiB disks, and below 50 GiB/above 75 GiB for disks at least 1.25 TiB. These are recovery hysteresis thresholds, separate from the 90% auto-scaling trigger. (low confidence) (verified as of 2026-10-14) [source]
Facts and statements
- Auto-expansion can lose the race against rapid writes, including migration or bulk-insert bursts; capacity limits can also prevent expansion. Scale ahead of known bursts and monitor disk headroom. (high confidence) (verified as of 2026-10-14) [source]
- Atlas Core delays auto-scaling until the oplog window covers the estimated scaling time. An enabled auto-scaling switch alone does not guarantee an immediate scale event. (low confidence) (verified as of 2026-10-14) [source]
- Atlas automatically unblocks writes after free space recovers past the applicable recovery threshold. Increasing storage manually or enabling auto-expansion are documented recovery actions. (medium confidence) (verified as of 2026-10-14) [source]
- Stacktape exposes disableDiskScaling separately from disableScaleDown in its MongoDbAutoScaling interface; the latter controls compute-tier downscaling. Treat the two flags as different configuration controls. (low confidence) (verified as of 2021-11-01) [source]
- Storage auto-scaling only increases capacity; reducing storage requires a manual change. (low confidence) (verified as of 2026-10-14) [source]
- Different instance classes for base and analytics nodes disable disk auto-scaling. (low confidence) (verified as of 2026-10-14) [source]
- Write blocking applies to M10+ clusters running MongoDB 8.0 or later, despite the broader IWM overview describing MongoDB 9.0 eligibility. (low confidence) (verified as of 2026-10-14) [source]
- Blocked-write log reasons vary by FCV: 9.0+ uses InsufficientDiskSpace; earlier FCVs use DiskUseThresholdExceeded. (low confidence) (verified as of 2026-10-14) [source]
- Atlas provides default alerts for writes blocked by critically low storage and for writes subsequently unblocked; recovery occurs automatically after the unblocking threshold is crossed. (high confidence) (verified as of 2026-10-14) [source]
Children
- No children recorded.