Autoscaling Volumes

This guide explains how to configure and manage autoscaling for Persistent Volumes

Volume autoscaling automatically expands persistent volumes when usage exceeds defined thresholds. This is particularly important for Druid historical nodes and other stateful components.

Configuration

volume-autoscaler:
  # Trigger threshold - when disk is this % full, consider scaling
  scale_above_percent: "75"
  
  # For Production environments:
  scale_up_percent: "50"                  # Increase by 50%
  scale_up_min_increment: "107374182400"  # Minimum 100GB increment
  scale_up_max_increment: ""              # No maximum increment (uses max size)
  scale_up_max_size: ""                   # Default 16TB on AWS EBS

Production Recommendations

1. Scaling Thresholds:

   • Set scale_above_percent: "75" to trigger before disk is critically full

   • Allow time for scaling operation to complete

   • Consider monitoring trends to scale proactively

2. Scaling Increments:

   • Use scale_up_percent: "50" for percentage-based growth

   • Set scale_up_min_increment to 100GB (107374182400 bytes)

   • Choose whichever is larger: 50% or 100GB

   • This balances cost with reducing scaling frequency

3. Size Considerations:

   • Do a proper sizing based on the data ingestion rate for services like Kafka, Druid

   • Consider data growth rate and retention period

   • Account for segment replication factor

   • Plan for peak ingestion periods

4. Provider Limits:

   • AWS EBS: Max 16TB per volume

   • Azure Disk: Varies by type (1TB-64TB)

   • GCP PD: Up to 64TB

⚠️ Important Notes:

• Volume expansion is one-way (cannot shrink)

• Some providers bill for allocated size, not used size

• Consider cost implications of aggressive scaling

• Monitor scaling events to optimize thresholds