Prune Execution Data to Prevent Database Bloat

By default n8n stores full input and output data for every execution, so the database can grow to tens of gigabytes and slow the interface. Enable pruning to delete old executions by age and count, and combine it with selective saving that keeps failures but drops successful-run data. Match the retention settings to your volume and compliance needs.

Why does the n8n executions table grow so large?

Every workflow execution stores its full input/output data in the database by default. After a few weeks of active use, the executions table can grow to tens of gigabytes, slowing down the entire n8n interface and eventually filling your disk.

Real-world example: A production n8n instance running 200 workflows became unusably slow after 3 months. Investigation revealed 2.8 million execution records consuming 47 GB of database storage.

How do you enable execution data pruning?

# .env configuration for execution pruning

# Enable automatic pruning

EXECUTIONS_DATA_PRUNE=true

# Keep successful executions for 7 days

EXECUTIONS_DATA_MAX_AGE=168

# Keep only the last 5000 executions regardless of age

EXECUTIONS_DATA_PRUNE_MAX_COUNT=5000

# Prune check interval in minutes (default: 15)

EXECUTIONS_DATA_PRUNE_TIMEOUT=15

How do you combine pruning with selective saving?

For high-volume deployments, combine pruning with selective saving:

# Don't save data for successful executions at all

EXECUTIONS_DATA_SAVE_ON_SUCCESS=none

# Always save data for failed executions (for debugging)

EXECUTIONS_DATA_SAVE_ON_ERROR=all

# Save manual executions (useful for development)

EXECUTIONS_DATA_SAVE_MANUAL_EXECUTIONS=true

Scenario	Recommended Settings
Development	Save all, prune after 30 days
Production (low volume)	Save all, prune after 7 days, max 10k
Production (high volume)	Save failures only, prune after 3 days
Compliance-required	Save all, prune after retention period, use external log export

Related: Flatten Deeply Nested API Responses · Use Docker Compose with Health Checks for n8n and PostgreSQL