Cleaning Up Events

This guide covers using the cloacinactl admin cleanup-events command to manage the execution event table, which grows with every task execution and can impact query performance over time.

Prerequisites

• cloacinactl binary installed
• Access to the Cloacina database (PostgreSQL or SQLite)
• Database URL available via --database-url, DATABASE_URL env var, or ~/.cloacina/config.toml

Why Cleanup Matters

The execution events table grows with every task execution and can reach millions of rows in busy systems. Regular cleanup keeps the database responsive without losing data about recent executions.

Running Cleanup

Step 1: Preview with Dry Run

Always start with --dry-run to see how many events would be deleted:

cloacinactl admin cleanup-events \
  --database-url postgresql://cloacina:cloacina@localhost:5432/cloacina \
  --older-than 90d \
  --dry-run

Output:

[DRY RUN] Would delete 284,391 execution event(s) older than 2026-01-02T14:30:00Z

Step 2: Execute the Cleanup

Once satisfied with the preview, remove the --dry-run flag:

cloacinactl admin cleanup-events \
  --database-url postgresql://cloacina:cloacina@localhost:5432/cloacina \
  --older-than 90d

Output:

Deleted 284,391 execution event(s) older than 2026-01-02T14:30:00Z

If no events match the threshold:

No execution events found older than 2026-01-02T14:30:00Z

Duration Format

The --older-than flag accepts a human-readable duration string with the following units:

Unit	Meaning	Example
d	Days	90d = 90 days
h	Hours	24h = 24 hours
m	Minutes	30m = 30 minutes
s	Seconds	60s = 60 seconds

Units can be combined:

# 7 days and 12 hours
cloacinactl admin cleanup-events --older-than 7d12h --dry-run

# 1 day, 2 hours, 30 minutes
cloacinactl admin cleanup-events --older-than 1d2h30m --dry-run

The duration is case-insensitive (90D works the same as 90d). The default is 90d if --older-than is not specified.

Using the Config File for Database URL

To avoid passing --database-url every time, set it in the config file:

cloacinactl config set database_url "postgresql://cloacina:cloacina@localhost:5432/cloacina"

Then run cleanup without the flag:

cloacinactl admin cleanup-events --older-than 30d --dry-run

The command resolves the database URL from (in order): --database-url flag, DATABASE_URL environment variable, ~/.cloacina/config.toml.

Automating Cleanup

System Cron (Weekly)

Add a cron entry to run cleanup every Sunday at 3:00 AM:

crontab -e

0 3 * * 0 /usr/local/bin/cloacinactl admin cleanup-events --older-than 90d >> /var/log/cloacina-cleanup.log 2>&1

System Cron (Daily, Shorter Retention)

For high-volume systems where events are only needed for a week:

0 2 * * * /usr/local/bin/cloacinactl admin cleanup-events --older-than 7d >> /var/log/cloacina-cleanup.log 2>&1

Using Environment Variables

If the database URL is stored in an environment file:

0 3 * * 0 . /etc/cloacina/env && /usr/local/bin/cloacinactl admin cleanup-events --older-than 90d >> /var/log/cloacina-cleanup.log 2>&1

Where /etc/cloacina/env contains:

export DATABASE_URL="postgresql://cloacina:secret@db.example.com:5432/cloacina"

Considerations

Retention Policies

Choose a retention period that balances operational needs with storage costs:

Use Case	Suggested Retention	Rationale
Development	7d	Only recent runs matter for debugging
Staging	30d	Keep enough history for release validation
Production	90d	Quarterly retention for incident investigation
Compliance-heavy	365d or longer	Regulatory requirements may mandate longer retention

Before Cleanup

• Back up first: If you might need historical events for compliance or auditing, back up the database before running cleanup.
• Off-peak hours: Schedule cleanup during low-traffic periods to minimize lock contention on the events table.
• Dry run in production: Always use --dry-run first in production to verify the scope of deletion.

What Gets Deleted

The cleanup command deletes rows from the execution events table where created_at is older than the specified threshold. It does not delete:

• Pipeline execution records (the parent execution metadata)
• Task execution records
• Workflow packages or schedules
• API keys or tenant data

Execution records remain intact so you can still see that a pipeline ran, but the detailed per-task event log for old executions will be gone.