Cluster Management¶

📖 Atlas Cluster Configuration - Cluster setup guide 📖 Atlas Cluster Tiers - Tier comparison

Cluster Tiers¶

Free and Shared Tiers¶

M0 (Free Tier): - 512 MB storage - Shared RAM and vCPU - No backup, no VPC peering, no Performance Advisor - Limited to 500 connections - Limited to 100 databases and 500 collections - Available in AWS, Azure, and GCP - Good for learning and prototyping only

M2 and M5 (Shared Tiers): - M2: 2 GB storage, M5: 5 GB storage - Shared infrastructure - Limited features compared to dedicated - No VPC peering or private endpoints - Basic metrics and alerting - Good for small applications and development

Dedicated Tiers (M10+)¶

📖 Dedicated Clusters - Dedicated cluster options

M10+ Features: - Cloud backup with snapshots and PIT restore - VPC peering and private endpoints - Performance Advisor and Query Profiler - Auto-scaling (compute and storage) - Analytics nodes - LDAP integration - Encryption at rest with customer-managed keys - Online Archive - Real-Time Performance Panel

Multi-Cloud Deployments¶

📖 Multi-Cloud Clusters - Multi-cloud configuration

Multi-Region Configuration¶

• Deploy replica set members across multiple regions
• Reduce read latency for geographically distributed users
• Provide disaster recovery across regions
• Configure node types per region:
• Electable nodes - Can become primary (participate in elections)
• Read-only nodes - Serve reads, never become primary
• Analytics nodes - Isolated analytical workloads

Node Types¶

Electable Nodes: - Participate in elections and can become primary - Majority must be in one region (prevents split-brain) - Minimum 3 electable nodes for HA - Can be 3, 5, or 7 electable nodes

Read-Only Nodes: - Replicate data but never become primary - Reduce read latency in remote regions - Up to 7 read-only nodes per cluster - Do not affect election outcomes

Analytics Nodes: - Isolated from operational workload - Target with readPreferenceTags=nodeType:ANALYTICS - Can be different tier than electable nodes - Up to 50 analytics nodes per cluster - Do not participate in elections

Global Clusters¶

📖 Global Clusters - Global write clusters

• Zone-based sharding for data locality
• Write data to specific regions based on zone key
• Requires M30+ tier
• Configure zones mapped to regions
• Shard key must include a zone key field (e.g., country, region)
• Ensures data residency compliance (GDPR, data sovereignty)

Auto-Scaling¶

📖 Auto-Scaling - Auto-scaling configuration

Compute Auto-Scaling¶

• Automatically scales cluster tier up or down
• Configure minimum and maximum tier
• Scaling based on CPU and memory utilization
• Cooldown period between scaling events (default: varies)
• Scaling is a rolling process (no downtime)
• Available for M10+ clusters

Configuration: - Minimum tier: Set the floor (e.g., M30) - Maximum tier: Set the ceiling (e.g., M60) - Scale down: Enabled/disabled independently - Operates within the same provider and region

Storage Auto-Scaling¶

• Enabled by default on all dedicated clusters
• Automatically increases disk space when utilization is high
• Scales up when disk utilization exceeds threshold
• Does not scale down (storage only increases)
• No downtime during storage scaling

Cluster Operations¶

Creating a Cluster¶

• Choose cloud provider (AWS, Azure, GCP)
• Select region(s) for deployment
• Choose cluster tier and configuration
• Configure replication and backup
• Set up network access and database users

Modifying a Cluster¶

• Tier changes are rolling (no downtime for most changes)
• Can change: tier, storage, IOPS, region configuration
• Cannot change: cloud provider (must create new cluster)
• Some changes require maintenance window

Pausing a Cluster¶

• Pauses a running cluster (M10+)
• Stops billing for compute (storage still billed)
• Data preserved during pause
• Auto-resumes after 30 days if not manually resumed
• Connections not accepted while paused

Terminating a Cluster¶

• Permanently deletes the cluster and all data
• Requires manual confirmation
• Backup snapshots retained per retention policy
• Cannot be undone

Atlas CLI for Cluster Management¶

📖 Atlas CLI - CLI documentation

# Create a cluster
atlas clusters create myCluster \
  --provider AWS \
  --region US_EAST_1 \
  --tier M30 \
  --mdbVersion 7.0

# List clusters
atlas clusters list

# Describe a cluster
atlas clusters describe myCluster

# Update a cluster (scale up)
atlas clusters update myCluster --tier M40

# Pause a cluster
atlas clusters pause myCluster

# Resume a cluster
atlas clusters start myCluster

# Delete a cluster
atlas clusters delete myCluster

# Load sample data
atlas clusters sampleData load myCluster

# Watch cluster status
atlas clusters watch myCluster

Cluster Metrics¶

Key Metrics to Monitor¶

Connection String¶

mongodb+srv://username:password@cluster0.xxxxx.mongodb.net/mydb?retryWrites=true&w=majority

• +srv format auto-discovers replica set members via DNS SRV records
• retryWrites=true enables automatic write retries
• w=majority configures majority write concern
• Connection string available in Atlas UI under "Connect"