MongoDB Atlas

MongoDB Atlas is the fully managed cloud database service — provisioning, patching, backups, monitoring, and scaling handled by MongoDB Inc. Most production deployments start here rather than self-hosting.

Why Atlas? link

Atlas runs on AWS, GCP, and Azure in 100+ regions worldwide.

Getting Started link

1. Create account at cloud.mongodb.com
2. Create a project (organizational unit for clusters, users, alerts)
3. Build a cluster — start with M10+ for production (M0/M2/M5 are shared/dev tiers)

  # Install Atlas CLI
brew install mongodb-atlas-cli

# Login and create cluster
atlas auth login
atlas clusters create myCluster --provider AWS --region US_EAST_1 --tier M10
  

Cluster Tiers link

Production minimum: M10 (dedicated resources, backup, SLA).

Network Access link

IP Access List link

Restrict connections to known IPs:

  atlas accessLists create --ip 203.0.113.50 --comment "Office"
atlas accessLists create --ip 10.0.0.0/8 --comment "VPC"
  

Atlas blocks all other IPs by default.

VPC Peering and PrivateLink link

For production — no public internet traffic:

  Application (VPC) → VPC Peering / PrivateLink → Atlas Cluster
  

  # AWS VPC peering
atlas networks peering create aws --providerAccountId 123456789 \
  --vpcId vpc-abc123 --routeTableCidrBlock 10.0.0.0/16 \
  --region US_EAST_1
  

PrivateLink (AWS) and Private Service Connect (GCP) provide private connectivity without peering route table management.

Database Users link

  atlas dbusers create admin --role atlasAdmin
atlas dbusers create appUser --role readWriteAnyDatabase
  

Use least-privilege roles — never atlasAdmin for applications:

Connection string format:

  mongodb+srv://appUser:[email protected]/mydb?retryWrites=true&w=majority
  

Backups and Recovery link

Atlas continuous cloud backup (M10+):

• Continuous backup — oplog-based, point-in-time recovery
• Snapshot schedule — daily, weekly, monthly retention
• Restore — to same cluster, new cluster, or download

  # List snapshots
atlas backups snapshots list --clusterName myCluster

# Restore to new cluster
atlas backups restores create --clusterName myCluster \
  --snapshotId 67890 --deliveryType automated
  

Test restores quarterly — an untested backup is not a backup.

Monitoring and Alerts link

Atlas provides built-in dashboards:

• Operations — opcounters, connections, network I/O
• Performance — slow queries, index suggestions (Performance Advisor)
• Replication — lag, oplog window
• Hardware — CPU, memory, disk I/O

Configure alerts:

  atlas alerts settings create --event REPLICA_SET_MIGRATED --enabled \
  --notificationType EMAIL --intervalMin 60
  

Key alerts to configure:

• Replication lag > 60 seconds
• Connections > 80% of limit
• Disk usage > 80%
• Query targeting scan-to-return ratio > 1000

Performance Advisor link

Atlas analyzes slow queries and suggests indexes:

• Review suggested indexes before creating — test impact on writes
• Dismiss suggestions for one-off admin queries
• Auto-create indexes with approval workflow (Enterprise)

Scaling Operations link

Vertical Scaling link

  atlas clusters update myCluster --tier M20
# Rolling upgrade — brief failover during election
  

Horizontal Scaling (Sharding) link

Enable sharding on M30+ clusters:

  atlas clusters upgrade myCluster --enableSharding
  

Choose shard key before loading data — Atlas supports resharding but it is disruptive.

Read Scaling link

Add analytics nodes (Dedicated) or enable Read Preference to secondaries:

  mongodb+srv://.../?readPreference=secondaryPreferred
  

Atlas Search link

Full-text search without Elasticsearch:

  // Create search index (Atlas UI or API)
{
  "mappings": {
    "dynamic": false,
    "fields": {
      "title": { "type": "string" },
      "body": { "type": "string" },
      "category": { "type": "stringFacet" }
    }
  }
}

// Query
db.articles.aggregate([
  {
    $search: {
      index: "articles_search",
      text: { query: "mongodb atlas", path: ["title", "body"] }
    }
  },
  { $limit: 20 }
])
  

Faceted search, autocomplete, and relevance scoring built in.

Atlas Triggers and Functions link

Serverless event processing:

• Database Triggers — react to insert/update/delete (like change streams, managed)
• Scheduled Triggers — cron jobs
• Functions — JavaScript serverless functions with MongoDB access

Use for cache invalidation, notifications, and lightweight ETL without managing consumers.

Multi-Region and Global Clusters link

Global clusters (M30+) with zone sharding:

  US-East (default write) ←→ EU-West (local reads) ←→ AP-Southeast (local reads)
  

Data placed in geographic zones based on shard key — GDPR compliance and low-latency reads.

Atlas vs Self-Hosted Decision Matrix link

Common Mistakes link

• Using M0/M2 for production — no SLA, shared resources
• Allowing 0.0.0.0/0 in IP access list — open to internet
• Not enabling backup on M10+ — data loss risk
• Ignoring Performance Advisor suggestions for months
• Scaling tier without investigating query performance first
• Storing connection strings with credentials in source code

Troubleshooting link

Connection timeout

• Check IP access list includes your source IP
• Verify VPC peering routes if using private connectivity
• Test with mongosh "mongodb+srv://..."

Authentication failed

• Verify username/password in Database Access (not Organization users)
• Check authSource in connection string

Cluster auto-paused (M0)

• Free tier pauses after 60 days inactivity — resume in Atlas UI

Production Checklist link

• M10+ dedicated cluster in production region
• VPC peering or PrivateLink configured
• IP access list restricted (no 0.0.0.0/0)
• Least-privilege database users
• Cloud backup enabled with tested restore
• Alerts configured for lag, disk, connections
• Connection string in secrets manager
• Performance Advisor reviewed weekly

Best Practices link

1. Use separate Atlas projects for dev/staging/production
2. Tag clusters for cost allocation
3. Enable database auditing for compliance workloads
4. Use mongodb+srv:// connection strings for automatic failover
5. Plan cluster tier with 30% headroom for growth
6. Review Atlas billing alerts monthly

What Comes Next link

For workloads exceeding single replica set capacity, sharding deep dive covers shard key design, chunk management, and resharding strategies.