n8n execution data: set retention and pruning without losing diagnostics

SOURCE-BASED ANALYSIS · NO HANDS-ON RATING

Start from the evidence an operator needs

n8n can save different execution categories and prune completed execution data by age or count. The official self-hosting guidance recommends avoiding unnecessary saved data and enabling pruning. A retention policy should still preserve enough failed execution detail to diagnose errors and enough approved evidence for the workflow's operational requirements.

Build a retention matrix

For a fictional low-risk synchronization, decide separately whether to retain successful, failed and manual executions. Record the required investigation window and any external audit evidence before choosing values.

Understand both pruning thresholds

n8n documents age and count conditions for pruning finished executions. When either threshold applies, older eligible data can be removed. Running and waiting executions are not eligible, and annotated executions are preserved. Pruning uses a soft-delete and later hard-delete process, so an immediate database-size change should not be assumed.

1. Inventory execution volume and payload sensitivity.
2. Define the investigation and evidence windows.
3. Choose save rules for success, error and manual runs.
4. Set age and count pruning boundaries.
5. Create synthetic success and failure executions.
6. Verify operators can find the retained failure evidence.
7. Monitor database growth after the policy change.

Account for storage implementation

For default SQLite storage, pruned space may be reused without shrinking the database file unless a separate vacuum action is configured. Binary-data storage also needs its own review. Accept the policy when routine volume remains bounded and a controlled failure can still be explained and retried. This guide is not a sizing benchmark or a substitute for backup policy.