# Flex-to-Dedicated migration triggers (500 ops/sec, 5 GB, feature gates) > Source-anchored research for Flex-to-Dedicated migration triggers (500 ops/sec, 5 GB, feature gates). Run status: COMPLETED. Parent: [Cluster Tier Limits and Feature Gates as Cost Drivers (Connections, IOPS Caps, Minimum Tiers)](https://llms-explorer.com/tree/cluster-tier-limits-and-feature-gates-as-cost-drivers-connections-iops-caps-minimum-tiers/) · 5 facets · 17 facts · page: https://llms-explorer.com/tree/flex-to-dedicated-migration-triggers-500-ops-sec-5-gb-feature-gates/ ## How it works - MongoDB documents downtime when scaling Flex to M10+, instructs operators to halt writes during scaling and restart applications afterward, and disallows native M10+ scale-down back to Flex. Its current instructions also warn that the upgrade permanently deletes existing backup snapshots; preserve required snapshots or export data beforehand. (low confidence) (verified as of 2026-10-30) — [source](https://www.mongodb.com/docs/atlas/scale-cluster.md) ## Measurements and reference values - Private endpoints, network peering, database auditing and customer-managed encryption keys are unavailable on Flex. These requirements can force Dedicated adoption even below 500 ops/sec and 5 GB. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) ## Problems, failure modes and limitations - Flex limits combined reads and writes to 500 operations per second. Exceeding this ceiling triggers per-connection cooldowns and queued operations, potentially waiting more than a second. A higher Flex pricing band cannot remove the hard limit. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - Flex limits storage to 5 GB of uncompressed BSON documents plus associated indexes. Storage cannot be configured or automatically expanded; MongoDB recommends Dedicated when more capacity is required. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - Flex cannot host sharded clusters or dedicated analytics/search nodes, and lacks Performance Advisor. Dedicated workload-isolation nodes require M10+; the documented sharding minimum is M30, so moving to M10 does not unlock every missing feature. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - A user report describes a Flex-to-M10 upgrade failing during migration, then rejecting connections while rolling back. The author later reported creating a new M10 and migrating manually. This is one unverified firsthand report, not evidence of failure frequency or a general root cause. (low confidence) — [source](https://www.reddit.com/r/MongoDB_Official/comments/1wr97d1/atlas_flex_cluster_aws_mumbai_lost_its_primary/) ## Comparisons and alternatives - Flex bills hourly across usage bands, with $8 to $30 for 30 days; its base includes 100 ops/sec, 5 GB and unlimited data transfer. Migration replaces this capped consumption model with Dedicated capacity charges, so evaluate the target configuration rather than assuming a small Flex overage. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/atlas-flex-costs.md) - PointFive recommends comparing peak Opcounters and Connections, rather than averages, with Flex limits before deciding whether a workload fits. This is an assessment method, not a measured universal utilization threshold. (low confidence) (verified as of 2026-09-28) — [source](https://www.pointfive.co/efficiency-hub/inefficiencies/dedicated-mongodb-atlas-cluster-for-workloads-that-fit-a-flex-cluster) - In a serverless-retirement discussion, a user said their dataset exceeded 5 GB and that they were being directed to Dedicated rather than Flex. This is firsthand evidence of a storage-driven tier transition, not independent verification of current pricing or policy. (low confidence) — [source](https://www.reddit.com/r/mongodb/comments/1h01cii) - PointFive recommends checking peak Opcounters and Connections, data plus index size, and required features before selecting a tier. Its sizing advice supports retaining Dedicated where requirements exceed Flex, rather than upgrading solely because a database serves production traffic. (low confidence) (verified as of 2026-09-28) — [source](https://www.pointfive.co/efficiency-hub/inefficiencies/dedicated-mongodb-atlas-cluster-for-workloads-that-fit-a-flex-cluster) ## Facts and statements - Flex takes one daily snapshot and does not support custom backup policies, on-demand snapshots, continuous backup or point-in-time restore. Recovery requirements can therefore force Dedicated adoption before capacity limits. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - MongoDB Search and Vector Search are available on Flex. Requiring search functionality alone does not establish a Dedicated migration requirement; requiring separate Search Nodes does. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/atlas-flex-costs.md) - Private endpoints, network peering, continuous backup and point-in-time restore, database auditing, customer-managed encryption keys, sharding, and dedicated analytics or Search nodes are unavailable on Flex. These requirements can force Dedicated migration below the throughput and storage ceilings; the destination tier must support each required feature. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - Flex takes one daily snapshot and does not support custom backup policies or on-demand snapshots. Recovery objectives needing finer backup control can therefore justify a Dedicated tier before capacity limits are reached. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/reference/flex-limitations.md) - Search and Vector Search support alone do not require leaving Flex. MongoDB positions Flex for development, testing, and low-traffic production; needing consistent dedicated resources or unsupported capabilities is a separate migration rationale. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/atlas-flex-costs.md) - MongoDB documents downtime for a Free/Flex-to-Dedicated upgrade and instructs operators to halt writes during the operation. It does not support scaling that Dedicated cluster directly back to Flex; plan the maintenance window and future reversal separately. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/scale-free-flex-cluster.md) - Moving beyond Flex's $8–$30 monthly range introduces a Dedicated base charge and possible additional storage, backup, data-transfer, and feature charges. AceMQ's cost analysis recommends comparing full operating costs and fixing sizing or indexing problems before buying more hardware. (medium confidence) (verified as of 2026-10-11) — [source](https://www.mongodb.com/docs/atlas/billing/atlas-flex-costs.md)