# MongoDB Atlas Multi-Cloud > - Designing Atlas PrivateLink on Azure — steps, CLI, Terraform, AKO Parent: [MongoDB Atlas](https://llms-explorer.com/tree/mongodb-atlas/) · 19 facets · 65 facts · page: https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/ ## When to Use This Skill - Designing Atlas PrivateLink on Azure - steps, CLI, Terraform, AKO — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Setting up GCP Private Service Connect for Atlas — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Planning multi-cloud replica set topology (write-concern latency tradeoffs) — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Evaluating Atlas marketplace billing and MACC/CUD applicability — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Architecting cross-cloud DR with Atlas — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Auditing egress cost for multi-cloud clusters — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) - Skip for: Deep Azure-only content (DNS zones, NSG, Entra ID OIDC, AKS) → use mongodb-atlas-azure; Deep GCP-only content (Shared VPC, Workload Identity, GKE, Vertex AI) → use mongodb-atlas-gcp; AWS-only networking → use mongodb-aws-networking. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#when-to-use-this-skill) ## Setup Steps - Request Private Endpoint in Atlas UI/API: POST /api/atlas/v2/groups/{groupId}/privateEndpoint/azure/endpointService — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#setup-steps) - Atlas creates Azure Private Link Service backed by Standard Load Balancer — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#setup-steps) - Azure creates NIC in your subnet with private IP — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#setup-steps) - Status sequence: Creating → Available — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#setup-steps) ## DNS - Private DNS Zone .mongodb.net with A records pointing to private endpoint NIC IP. Zone must be linked to every VNet needing resolution. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#dns) - Connection string: mongodb+srv://cluster0-pl-0..mongodb.net — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#dns) ## Legacy vs Port-Mapped - Legacy (deprecated Apr 30 2027): _pl- prefix; 50 IP addresses per region; 50 forwarding rules — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#legacy-vs-port-mapped) - Port-Mapped (current): _psc- prefix; 1 IP address; 1 forwarding rule; requires portMappingEnabled: true — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#legacy-vs-port-mapped) ## DNS for PSC - Cloud DNS private zone. Allow TCP 27017 + high ports (1024-65535) through VPC firewall rules for SRV connection strings. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#dns-for-psc) ## Topology Options - 3-node across 2 clouds (HA across clouds): — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) - Primary: AWS US_EAST_1 — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) - Secondary: GCP EASTERN_US — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) - Secondary: Azure EASTUS — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) - Write-concern latency: W:majority in multi-cloud = at least one write acknowledged from another cloud. Cross-cloud RTT typically 10-50ms. Use wtimeoutMS to prevent write stalls. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) - Read preference: Route reads to local secondary with readPreference: nearest + maxStalenessSeconds. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#topology-options) ## Cross-Cloud DR Pattern - Primary cloud (AWS): 3 electable nodes — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#cross-cloud-dr-pattern) - DR cloud (Azure/GCP): 1-2 non-electable nodes (priority=0) or hidden secondaries — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#cross-cloud-dr-pattern) - Hidden secondary in DR cloud: receives replication without being elected — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#cross-cloud-dr-pattern) - Failover: promote DR nodes by changing priority — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#cross-cloud-dr-pattern) - RPO: Near-zero (replication lag typically <1 second for healthy cluster) RTO: 10-30 seconds for automatic election after primary failure — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#cross-cloud-dr-pattern) ## Egress Cost Analysis - Cross-cloud data transfer incurs egress charges: — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) - AWS→GCP: ~$0.08/GB (US East) — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) - AWS→Azure: ~$0.08/GB (US East) — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) - GCP→AWS: ~$0.08/GB (US Central) — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) - For a 3-replica-set with 10 GB/day replication across clouds: ~$0.80/day in egress costs. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) - Mitigation: Keep primary and majority of nodes in lowest-cost cloud. Use hidden secondaries in DR cloud to limit replication traffic. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#egress-cost-analysis) ## Azure MACC - Atlas purchased via Azure Marketplace counts toward Microsoft Azure Consumption Commitment (MACC). Direct MongoDB invoices are NOT MACC-eligible. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#azure-macc) - MACC-eligible purchases: Atlas cluster compute/storage, Advanced Security, Dedicated Search Nodes, Stream Processing. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#azure-macc) - Azure Native MongoDB (ANM): First-party Azure resource in Azure Portal. Billing on Azure invoice (MACC-eligible). Generally at feature parity with standard Atlas. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#azure-macc) ## GCP Committed Use Discounts (CUD) - GCP Marketplace Atlas purchases are EDP (Estimated Discount Program) eligible. GCP startup credits can be stacked. Committed use via GCP CUD contracts. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#gcp-committed-use-discounts-cud) ## AWS ISV Accelerate / ISV Workload Migration Program - MongoDB is an AWS ISV Accelerate partner. Atlas clusters on AWS can qualify for AWS partner funding for customer migrations. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#aws-isv-accelerate-isv-workload-migration-program) ## Azure Preferred Solutions / MACC Impact Program - MongoDB Atlas is an Azure Preferred Solution. Purchases via Marketplace impact partner competency and incentive credits. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#azure-preferred-solutions-macc-impact-program) ## GCP Startup Ecosystem - MongoDB participates in Google for Startups Cloud Program. Atlas credits available for eligible startups. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#gcp-startup-ecosystem) ## Anti-Patterns - Multi-cloud without read preference tuning: Default primary read preference sends all reads cross-cloud → high latency. Set readPreference: nearest. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#anti-patterns) - Majority writes in high-latency multi-cloud setups without wtimeoutMS: Writes can block indefinitely if cross-cloud replication stalls. Always set wtimeoutMS. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#anti-patterns) - Using VNet peering instead of Private Link for new Azure deployments: Private Link is recommended and simpler (no IP overlap constraints). — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#anti-patterns) - Legacy GCP PSC endpoints without migration plan: Must migrate to port-mapped before April 30, 2027. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#anti-patterns) - Forgetting cross-cloud egress costs in TCO analysis: Multi-cloud replication incurs ~$0.08/GB egress. — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#anti-patterns) ## References - MongoDB Atlas Private Endpoint Management — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#references) - Multi-Cloud Cluster Documentation — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#references) - GCP PSC Port-Mapped Architecture — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#references) - mongodbatlas Terraform provider — [source](https://llms-explorer.com/sources/mdb-context-hub/mongodb-atlas-multicloud/#references) ## Where this helps - Designing a replica set that spans AWS, GCP, and Azure for cross-cloud high availability or disaster recovery. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Setting up private connectivity, such as Azure Private Link or GCP PSC, for a multi-cloud Atlas topology. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Evaluating whether multi-cloud egress costs and cross-cloud write latency are actually worth the availability benefit for a given workload. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Applying marketplace billing programs, such as Azure MACC, GCP CUD, or AWS ISV Accelerate, across a multi-cloud Atlas footprint. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Project ideas - Build a 3-node, 2-cloud replica set, such as AWS primary-eligible nodes plus a GCP secondary, and measure real cross-cloud write-concern latency with W:majority before setting wtimeoutMS in production. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Design a cross-cloud DR pattern using a hidden, non-electable secondary (priority=0) in a second cloud, and test promoting it during a simulated primary-cloud failure. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Migrate a legacy GCP PSC endpoint (the _pl- prefix) to the current port-mapped (_psc-) architecture ahead of the April 30, 2027 deprecation deadline. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Build an egress-cost model comparing a single-cloud cluster against a 2-cloud or 3-cloud topology, using the roughly $0.08/GB cross-cloud transfer rate as the baseline. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Common mistakes - Leaving read preference at its primary default in a multi-cloud cluster, which sends every read cross-cloud instead of routing to a local secondary with readPreference: nearest. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Using majority writes in a high-latency multi-cloud setup without setting wtimeoutMS, which can let a write block indefinitely if cross-cloud replication stalls. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Choosing VNet peering over Private Link for a new Azure deployment, when Private Link is the recommended, simpler option with no IP-overlap constraints. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Ignoring cross-cloud egress costs, around $0.08/GB, in a multi-cloud TCO analysis, which can be a meaningful recurring cost for high-replication-volume clusters. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Known issues - Legacy GCP PSC endpoints, the _pl- prefix architecture, are deprecated and must be migrated to the port-mapped architecture before April 30, 2027. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Cross-cloud replication RTT is typically 10–50ms, which sets a hard floor on write-concern latency for any topology that requires acknowledgment from another cloud. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - A hidden secondary in a DR cloud still incurs full replication traffic and its associated egress cost, even though it never serves reads or gets elected. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Marketplace billing eligibility, such as Azure MACC, GCP CUD, or AWS ISV programs, depends on purchase channel — direct MongoDB invoices are generally not eligible for these programs, only Marketplace purchases are. — [source](https://llms-explorer.com/tree/mongodb-atlas-multi-cloud/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Context files - [MongoDB Atlas Multi-Cloud](https://llms-explorer.com/downloads/sources/mdb-context-hub/mongodb-atlas-multicloud.md)