Atlas disk write-blocking thresholds and recovery
Parent: Atlas storage auto-scaling and disk reclaim via compact · Published reference · snapshot 2026-10-31
↓ Facts as markdownall context files
Source-anchored research for Atlas disk write-blocking thresholds and recovery. 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 also blocks deletes, because deleting documents can temporarily increase used disk space. (low confidence) (verified as of 2026-10-31) [source]
Measurements and reference values
- For the documented 20 GiB–1.25 TiB disk range, Atlas blocks below 4% free and unblocks above 6% free. (low confidence) (verified as of 2026-10-31) [source]
- Atlas Core's documented disk bands are: 8–20 GiB, block below 600 MiB free and unblock above 900 MiB; 20 GiB–1.25 TiB, below 4% and above 6%; at least 1.25 TiB, below 50 GiB and above 75 GiB. These are free-space thresholds. (low confidence) (verified as of 2026-10-31) [source]
Problems, failure modes and limitations
- Auto-Expand Storage can be enabled yet fail to expand before a migration fills the disk, or be unable to expand beyond the cluster's storage maximum; a larger tier is the documented long-term remedy at that maximum. (low confidence) (verified as of 2026-10-31) [source]
- The configurable Disk space % used on Data Partition alert defaults to 90% used capacity; this warning threshold is distinct from the write-blocking policy. (high confidence) (verified as of 2026-10-31) [source]
- Auto-expansion can lose a race with rapid ingestion or hit the cluster's storage limit. Enabling it does not guarantee that write blocking will be avoided. (medium confidence) (verified as of 2026-10-31) [source]
- An independent M0 incident dated 2026-05-23 reports quota error 8000 blocking application status updates while deletion of completed experiments succeeded. This historical Free-tier quota incident does not establish Dedicated-cluster disk-protection thresholds or delete behavior. (low confidence) [source]
Facts and statements
- Atlas Core blocks replica-set writes at critically low primary-node disk space while serving reads; sharded clusters block only writes routed to the affected shard. (high confidence) (verified as of 2026-10-31) [source]
- For the documented 8–20 GiB disk range, Atlas blocks below 600 MiB free and unblocks above 900 MiB free. (low confidence) (verified as of 2026-10-31) [source]
- For disks of at least 1.25 TiB, Atlas blocks below 50 GiB free and unblocks above 75 GiB free. (low confidence) (verified as of 2026-10-31) [source]
- Recovery guidance is to increase storage manually or enable storage auto-scaling; Atlas automatically unblocks writes after the recovery threshold is crossed. (high confidence) (verified as of 2026-10-31) [source]
- Write-blocking protection is documented for M10+ running MongoDB 8.0+, including versions preceding the broader MongoDB 9.0 IWM requirement. (low confidence) (verified as of 2026-10-31) [source]
- Logs distinguish FCV 9.0+ 'Replica set writes blocked, reason: InsufficientDiskSpace' from earlier-FCV 'User writes blocked, reason: DiskUseThresholdExceeded'. (low confidence) (verified as of 2026-10-31) [source]
- The MongoDB server storage README describes a DiskSpaceMonitor that checks database-path space every second and invokes registered threshold actions; its index-build example does not establish Atlas policy thresholds. (low confidence) (verified as of 2026-10-31) [source]
- MongoDB documents manual storage expansion or enabling storage auto-scaling as immediate remedies. If the current tier has reached its storage ceiling, increase the cluster tier. (medium confidence) (verified as of 2026-10-31) [source]
- Atlas automatically resumes writes after the unblocking threshold is satisfied. The Writes have been unblocked on your cluster alert is informational and requires no further recovery action. (medium confidence) (verified as of 2026-10-31) [source]
- eric-sabe's operational notes report a bulk restore aborting with DiskUseThresholdExceeded. This is a firsthand symptom report, not evidence that every auto-scaling operation intrinsically blocks writes. (low confidence) [source]
- The March 2026 Atlas changelog records a change from static to dynamic range-based write-blocking thresholds to reduce unnecessary blocking during disk scaling. (low confidence) [source]
Children
- No children recorded.