Skip to main content

Spend Log Cleanup

Automatically clean up old spend logs to manage database size and performance.

What causes the problem?​

LiteLLM accumulates spend logs for every request. Over time, these logs can grow to a large size, potentially impacting database performance and storage costs.

How the spend log cleanup feature works​

The spend log cleanup feature:

  • Runs periodically to delete old spend logs
  • Uses distributed locking to ensure only one instance runs cleanup at a time
  • Deletes logs in batches to prevent database overload
  • Configurable retention period and cleanup interval

How it works​

Stage 1. Instance acquires cleanup lock​

When cleanup is triggered, an instance will attempt to acquire a distributed lock:

  • Only one instance can acquire the lock at a time
  • The lock status is stored in Redis
  • If an instance acquires the lock:
    • It proceeds with cleanup
    • Other instances skip cleanup
  • If no instance has the lock:
    • Cleanup runs without distributed locking

Working of spend log deletions

Working of spend log deletions

Stage 2. Batch deletion of old logs​

The instance with the lock:

  • Calculates cutoff date based on retention period
  • Finds logs older than cutoff date
  • Deletes logs in batches of 1000
  • Adds small delays between batches to prevent database overload
  • Releases lock when complete

Batch deletion of old logs

Batch deletion of old logs

Usage​

Required components​

  • Redis (optional, for distributed locking)
  • Postgres

Setup on LiteLLM config​

You can configure spend log cleanup by setting the retention period and cleanup interval in the general_settings section of your proxy_config.yaml file:

litellm proxy_config.yaml
general_settings:
  maximum_spend_logs_retention_period: "7d" # Keep logs for 7 days, i.e the cutoff date, logs older than x should be deleted
  maximum_spend_logs_retention_interval: "1d" # Run cleanup every 1 day, i.e. the interval in which the job should run

litellm_settings:
  cache: True
  cache_params:
    type: redis

Configuration Settings​

Retention Period​

The maximum_spend_logs_retention_period determines how long to keep logs before deletion. Supports the following formats:

  • "7d" - 7 days
  • "24h" - 24 hours
  • "60m" - 60 minutes
  • "3600s" - 3600 seconds

Cleanup Interval​

The maximum_spend_logs_retention_interval controls how often the cleanup job runs. Uses the same time format as retention period:

  • "1d" - Run cleanup daily
  • "12h" - Run cleanup every 12 hours
  • "6h" - Run cleanup every 6 hours