An on-call schedule assigns responsibility for responding to critical incidents during particular shifts. The on-call person must be ready to react quickly, usually within 30 to 60 minutes. The duty rotates among team members—for example, one person may cover a week before handing it to the next. section 16.1
In a small company, incidents may initially be handled by whoever is available. As the system grows, that can blur responsibility: everyone and no one is responding. Formalize a rotation when that pattern appears, and make escalation explicit for situations that do not fit the usual accountability arrangements. section 16.1
When planning a rotation, consider:
- Who responds first, and who can take over a problem requiring component-specific expertise? Larger operations may need a dedicated on-call team with multiple tiers. section 16.1 section 16.5
- How will the team balance efficiency and redundancy? Maintain enough capacity for peak loads without working people to exhaustion. section 16.6
- Are responsibilities explicit, and is system documentation current and available to the people who need it? section 16.6
A two-level rotation
For an image search system with API handlers, an index, ranking models, and internal tools, a senior engineer is responsible for each component and has visibility into the others. A primary on-call engineer responds first and can escalate specific problems to component on-call engineers. section 16.5
