MongoDB to Azure Cosmos DB Migration Center¶

The definitive resource for migrating from MongoDB (Atlas, Community, Enterprise) to Azure Cosmos DB for MongoDB -- wire-protocol compatible, fully managed, globally distributed.

Who this is for¶

This migration center serves CTOs, CDOs, platform architects, data engineers, application developers, and DevOps engineers who are evaluating or executing a migration from MongoDB to Azure Cosmos DB. Whether you are running MongoDB Atlas (M10 through M700), self-hosted Community or Enterprise editions on VMs or Kubernetes, or a hybrid deployment -- these resources provide the evidence, patterns, and step-by-step guidance to execute confidently.

MongoDB powers over 46,000 customers globally. The migration to Cosmos DB is driven by one or more forcing functions: an Azure-first mandate, the need for turnkey global distribution with financially backed SLAs, tighter integration with the Azure analytical platform (Fabric, Purview, Power BI), HTAP via analytical store, vector search for AI/RAG, or federal compliance requirements that are simpler to satisfy on a fully managed Azure-native service.

Quick-start decision matrix¶

The Cosmos DB for MongoDB decision: vCore vs RU-based¶

Before diving into migration guides, choose the right Cosmos DB deployment model. This is the single most important architectural decision in the migration.

Cosmos DB for MongoDB vCore¶

• Architecture: Cluster-based, similar to Atlas or self-hosted MongoDB. Dedicated compute nodes with local SSD storage.
• Compatibility: Full MongoDB wire protocol (5.0+), including $lookup, $graphLookup, $merge, $out, native aggregation pipeline.
• Best for: Lift-and-shift from Atlas or self-hosted; workloads needing complex aggregation; applications where connection-string swap is the goal.
• Scaling: Vertical (up to 128 vCores, 2 TiB RAM per node) + horizontal (sharding). Burstable tier for dev/test.
• Vector search: Native integrated vector search for AI/RAG scenarios.
• Pricing: Per-node, per-hour. Predictable costs similar to Atlas cluster pricing.

Cosmos DB for MongoDB (RU-based)¶

• Architecture: Request-unit throughput model. Globally distributed, multi-region writes, automatic partitioning.
• Compatibility: MongoDB wire protocol (3.6, 4.0, 4.2, 5.0, 6.0, 7.0 selectable). Most aggregation stages supported; some limitations on $merge to different databases and certain $lookup patterns.
• Best for: Globally distributed applications; event-driven architectures needing change feed; HTAP via analytical store; serverless or unpredictable workloads.
• Scaling: Horizontal (unlimited partitions, autoscale 10x range). Manual or autoscale throughput provisioning. Serverless tier for dev/test.
• Analytical store: Column-oriented, fully isolated HTAP layer queryable from Fabric Spark or Synapse.
• Pricing: Per-RU/s provisioned (or per-operation for serverless). Globally distributed storage billed per GB per region.

Decision flowchart¶

flowchart TD
    Start["MongoDB Migration"] --> Q1{"Need global distribution<br/>with multi-region writes?"}
    Q1 -->|Yes| RU["Cosmos DB for MongoDB<br/>(RU-based)"]
    Q1 -->|No| Q2{"Need analytical store<br/>(HTAP) or change feed<br/>to Event Hubs?"}
    Q2 -->|Yes| RU
    Q2 -->|No| Q3{"Complex aggregation?<br/>$lookup, $graphLookup,<br/>$merge to other DBs?"}
    Q3 -->|Yes| vCore["Cosmos DB for MongoDB<br/>vCore"]
    Q3 -->|No| Q4{"Workload pattern?"}
    Q4 -->|"Predictable, steady"| vCore
    Q4 -->|"Bursty, unpredictable"| RU
    Q4 -->|"Dev/test"| Q5{"Budget priority?"}
    Q5 -->|"Lowest cost"| RUServerless["RU-based Serverless"]
    Q5 -->|"Simplest migration"| vCoreBurst["vCore Burstable"]

How CSA-in-a-Box fits¶

Cosmos DB does not exist in isolation within csa-inabox. It integrates with the broader data platform through three primary pathways:

Change feed to Fabric (real-time streaming)¶

Cosmos DB change feed captures every insert, update, and delete. An Azure Functions processor or a Kafka-protocol consumer pushes events to Event Hubs, which feeds Fabric Real-Time Intelligence (RTI). Data lands as Delta tables in the Fabric lakehouse within seconds, available for Power BI Direct Lake dashboards, dbt transformations, or downstream AI workloads.

CSA-in-a-Box reference: examples/iot-streaming/, csa_platform/data_activator/, ADR-0005 docs/adr/0005-event-hubs-over-kafka.md

Analytical store (zero-ETL HTAP)¶

For RU-based deployments, analytical store provides a column-oriented mirror of operational data. Fabric Spark notebooks or Synapse Link query this store directly -- no ETL pipeline, no RU consumption on the operational side. This is the fastest path to analytics over Cosmos DB data.

CSA-in-a-Box reference: docs/patterns/cosmos-db-patterns.md, docs/guides/cosmos-db.md

Purview governance and lineage¶

Purview scans Cosmos DB accounts, discovering collections, inferring schemas, applying data classifications (PII, PHI, financial data), and building lineage graphs from operational source through Fabric lakehouse to Power BI report. Every Cosmos DB collection becomes a governed data product in the csa-inabox data marketplace.

CSA-in-a-Box reference: csa_platform/csa_platform/governance/purview/purview_automation.py, ADR-0006 docs/adr/0006-purview-over-atlas.md

Integration architecture¶

flowchart LR
    subgraph Operational["Operational Tier"]
        Mongo["MongoDB<br/>(source)"]
        Cosmos["Cosmos DB<br/>for MongoDB"]
    end

    subgraph Integration["Integration"]
        DMS["Azure DMS<br/>(migration)"]
        CF["Change Feed"]
        AS["Analytical Store"]
        EH["Event Hubs"]
    end

    subgraph Platform["CSA-in-a-Box Platform"]
        Fabric["Fabric Lakehouse<br/>(Delta tables)"]
        Purview["Microsoft Purview<br/>(governance)"]
        PBI["Power BI<br/>(Direct Lake)"]
        AI["Azure AI<br/>(vector search)"]
        Marketplace["Data Marketplace"]
    end

    Mongo -->|Online CDC| DMS
    DMS -->|Migrate| Cosmos
    Cosmos --> CF
    Cosmos --> AS
    CF --> EH
    EH --> Fabric
    AS -->|Synapse Link| Fabric
    Cosmos --> Purview
    Fabric --> PBI
    Fabric --> AI
    Purview --> Marketplace

Strategic resources¶

These documents provide the business case, cost analysis, and strategic framing for decision-makers.

Migration guides¶

Domain-specific deep dives covering every aspect of a MongoDB-to-Cosmos DB migration.

Tutorials¶

Step-by-step walkthroughs with screenshots and validation checkpoints.

Technical references¶

Migration timeline expectations¶

Audience guide¶

Related resources¶

• Migration playbook (concise): mongodb-to-cosmosdb.md
• Cosmos DB patterns: docs/patterns/cosmos-db-patterns.md
• Cosmos DB guide: docs/guides/cosmos-db.md
• Decision trees:
  • docs/decisions/fabric-vs-databricks-vs-synapse.md
  • docs/decisions/batch-vs-streaming.md
• Companion migration playbooks: sql-server-to-azure.md, oracle-to-azure.md
• Compliance matrices:
  • docs/compliance/nist-800-53-rev5.md
  • docs/compliance/fedramp-moderate.md
  • docs/compliance/cmmc-2.0-l2.md

Maintainers: csa-inabox core team Last updated: 2026-04-30