Service Level Objectives

This page describes our Service Level Objectives (SLOs) for the Runlater service. These are targets we aim to meet, not contractual guarantees. For a binding SLA, please contact us.

Uptime

99.95% monthly uptime target

Equivalent to less than 22 minutes of downtime per month.

Uptime is measured as the availability of the Runlater API and scheduler. Planned maintenance windows are excluded and announced at least 24 hours in advance.

Current service status is available at runlater.eu/status.

Task Execution Timing

< 15s p95 execution start latency

95% of tasks begin executing within 15 seconds of their target time under normal load.

Scheduled tasks (cron and one-time): the scheduler creates executions ahead of time via a lookahead window. Workers claim pending executions as soon as they become due. Under low load, latency is typically 1–3 seconds.

Queued tasks (immediate): bypass the scheduler entirely and are picked up by the next available worker, typically within 1–2 seconds.

Under load: when queue depth is high, worker availability and per-organization fairness limits may add latency. The 15-second target accounts for these scenarios.

Webhook Delivery

Timeout: Configurable per task (1 second to 5 minutes, default 30 seconds)
Retries: One-time tasks retry up to 10 times with exponential backoff. Cron jobs do not retry (next scheduled run serves as the retry).
Payload size: Request bodies up to 256KB. Response bodies stored up to 256KB.

API Response Time

< 200ms p95 API response time target

95th percentile response time for API requests under normal load.

Data Durability

Database: PostgreSQL with regular automated backups
Retention: Execution history retained for 7 days (Free) or 30 days (Pro)
Location: All data stored in netcup, Nuremberg, Germany (EU)

Notifications

Failure alerts: Email and/or webhook notification sent within 1 minute of a task failure
Recovery alerts: Notification sent when a previously failing task succeeds
Limit warnings: Email sent when approaching monthly execution limits

Maintenance

We deploy updates with zero-downtime deployments. In rare cases where maintenance requires brief interruption:

Planned maintenance is announced at least 24 hours in advance
Maintenance windows are scheduled outside peak hours (typically 02:00-05:00 UTC)
Queued tasks are not lost during maintenance; they execute once the service resumes

Support

Email support: support@runlater.eu
Response time target: Within 1 business day
Critical issues: Best-effort same-day response