Amey-Thakur On-call health
On-call is where reliability debt is paid, by people, at night. A rotation that burns people out loses the institutional knowledge that made the system operable, which is an expensive way to learn.
Method
- Track pages per shift and treat the number as a target. More than a couple of interrupting pages per shift is unsustainable and should drive work (see alerting-design).
- Delete alerts nobody acts on. An alert that is always acknowledged and ignored trains people to ignore all of them.
- Every page needs a runbook. Being woken without instructions is both slower and worse for the person (see runbook-writing).
- Fix the causes, prioritised. Follow-up work from incidents must be scheduled, or the same page recurs indefinitely (see incident-postmortem).
- Compensate and protect recovery. Time off after a bad night is not generosity, it is how you keep people.
- Share the rotation broadly. A small rotation is fragile and concentrates burnout; broadening it also spreads system knowledge.
- Hand over deliberately. A written handover of ongoing issues prevents the next shift rediscovering them (see oncall-handoff).
Boundaries
On-call health depends on system reliability, so process improvements cannot compensate for a fundamentally unstable system. Rotations need enough people, which is a staffing decision. Working-hours expectations and compensation for on-call are employment matters (see agent-people-ops-desk).