On-Call Fatigue: Causes, Warning Signs and Fixes for SRE Teams
On-Call Fatigue: Causes, Warning Signs and Fixes for SRE Teams

On-call fatigue is the physical and mental exhaustion that builds from interrupted sleep and the constant readiness to respond. It is driven less by the number of pages than by how often they break sleep and how much context an engineer must rebuild at 3 a.m. The most effective fixes cut avoidable pages, shorten the time to understand an incident, and protect recovery after a rough night. Opsphere’s read-only investigation layer works on the second of those: it pulls AWS, Kubernetes, observability and CI/CD context into one interface, so the responder starts from evidence instead of ten open tabs.
TL;DR
- Problem: On-call work disrupts sleep and recovery, and even quiet shifts keep the mind partly at work. Over time that shows up as mood, focus and error-rate problems.
- Key takeaway: Attack three levers: fewer pages that truly need a human, faster investigations when a page does fire, and guaranteed recovery after disruption.
- Practical next step: Audit last quarter’s pages, mark which ones needed a human at night, and fix or silence the rest first.
- Tooling note: Unifying context without adding another telemetry pipeline keeps the fix from becoming more sprawl.
Table of Contents
- What on-call fatigue actually is
- Why quiet shifts still wear people down
- What broken sleep does to performance
- Warning signs your team should watch
- Protecting sleep and recovery as an individual
- Team practices that reduce on-call fatigue
- Cutting investigation time without adding tools
- How Opsphere supports a lighter on-call load
- A 30-day plan for your team
- FAQ
- Sources
What on-call fatigue actually is
Tiredness goes away after a good night’s sleep. On-call fatigue does not, because the cause is repeated: broken sleep, unpredictable demands and the lack of true time off the mental clock.
It overlaps with burnout, but they are not the same. The World Health Organization’s ICD-11 defines burnout as an occupational phenomenon with three dimensions: energy depletion or exhaustion, mental distance from the job or cynicism, and reduced professional efficacy. On-call fatigue is one of the routes into that state. Think of fatigue as the early, physical layer and burnout as what follows when it goes unaddressed.
The pattern also differs from alert fatigue. Alert fatigue is desensitization to noisy notifications. On-call fatigue is the broader cost to the person, and noisy alerts are one of its largest inputs.
Why quiet shifts still wear people down
Many engineers assume that if the pager stays silent, the shift was free. The research says otherwise.
A 2016 study by Dettmers et al. found that simply having to stay available during non-work hours limited recovery and was associated with next-morning mood and cortisol changes. An Australian cross-sectional study of 228 on-call workers found that 56% reported their sleep was affected even when no calls came in. The top reasons were interruption to family and leisure time (70%), fear of missing a call (69%) and planning ahead in case of a call (69%).

A Dutch study of on-call workers points the same way: 57.5% felt at least somewhat restricted during on-call duties, and 30.6% found having no control over whether they would be called at least somewhat stressful.
The lesson for rotation design: the on-call week itself is a load, whatever the page count. Teams that measure only pages per shift miss part of the picture.
What broken sleep does to performance
The fastest way to see the cost is to look at what happens in the first minutes after a page wakes someone.
Sleep inertia
Sleep inertia is the groggy transition between sleep and full wakefulness. According to a review by Gupta et al., its most severe symptoms typically last 15 to 30 minutes, and full recovery can take over an hour. During that window, memory, logical reasoning, reaction time and problem solving all suffer. The same review reports that 70% of on-call workers had difficulty getting back to sleep after a call.
For an SRE, those 15 to 30 minutes are exactly when the first diagnostic decisions are made: which dashboard, which deploy, which rollback.
Mood, motivation and vigilance
A 2024 JAMA Network Open cohort study of 96 medical interns over 8 weeks compared 24-hour overnight call with regular day shifts. Call schedules were associated with 13% poorer mood, 21% poorer motivation, 29% higher sleepiness ratings and 21 ms slower vigilance. A float schedule with consecutive 12-hour night shifts did not show these associations. This is a medical population, so treat it as directional evidence for how schedule shape matters, not as a measured SRE figure.
Interruption cost
Each page also costs focus. Ryn Daniels notes in Increment that a single interruption can waste at least 20 minutes through loss of focus and context switching. Multiply that by the pages an engineer handles in a week and the daytime productivity loss becomes visible.
Household and long-term effects
The disruption is not confined to the engineer. A 2026 IT Support Group summary of sleep research notes on-call work can also disturb the sleep of partners and family members. A 2025 British Journal of Nursing article links chronic fatigue to higher risk of burnout and cardiovascular disease, and notes a fatigue risk for people working more than 13 hours in one stretch.
Warning signs your team should watch
Individuals rarely announce fatigue. Managers and teammates usually see it first as a pattern. The OffShift guide to on-call burnout lists these signs:
- Persistent tiredness that days off do not relieve
- Emotional numbness or detachment
- Rising cynicism about the work
- Trouble concentrating on routine tasks
- Frequent minor illnesses
- Dread at the start of every shift
In SRE terms, add the operational tells: slower acknowledgment times on later pages in a week, more mistakes in change reviews, engineers swapping out of rotations, and post-incident reviews that keep citing tired responders.
A 2025 Catchpoint report, as summarized by IT Support Group, found nearly 70% of SREs said on-call stress affected burnout and attrition. That makes fatigue a retention problem as well as a wellbeing one.
When to see a doctor. Per the OffShift guide, seek medical advice for low mood lasting more than two weeks, thoughts of self-harm, physical symptoms such as chest pain or unexplained weight loss, or regular use of alcohol or other substances to cope. If you are in the US and in crisis, call or text 988, the Suicide & Crisis Lifeline.
Protecting sleep and recovery as an individual
Individual habits do not replace good rotation design, but they matter at 3 a.m.
- After a page: Gupta et al. suggest sleep inertia countermeasures such as bright light, caffeine and exercise to raise alertness, and reducing bright light or using mindfulness afterward to help sleep return. They caution that alertness aids can affect later sleep, so use them deliberately.
- Nap when you can: In the JAMA study, naps during night shifts were associated with 16 ms faster vigilance than nights without naps. A 2026 Sleep Health systematic review of rotating-shift healthcare workers found scheduled 30-minute naps reduced sleepiness and improved cognition.
- Slow down the first hour: Treat the minutes after waking as a high-risk window. Read the evidence before running a destructive command, and ask a second person to confirm risky changes.
- Detach on purpose: Because the mental clock keeps running, plan something absorbing for off-hours of an on-call week and make clear to teammates when you are the primary and when you are not.
See a clinician if sleep trouble persists after your rotation ends. Shift work sleep problems have recognized treatments, and the American Academy of Sleep Medicine has published a 2026 review on managing shift work disorder.
Team practices that reduce on-call fatigue
The Google SRE Workbook states the aim plainly: cover critical services without achieving reliability at the expense of an engineer’s health. These practices put that into effect.
1. Audit what deserves to wake someone
Ryn Daniels recommends auditing alerts to find which truly need a human off-hours. Pull a quarter of pages and sort each into one of four buckets:
| Bucket | Question | Action |
|---|---|---|
| Actionable, urgent | Did a human need to act now? | Keep, and improve the runbook |
| Actionable, not urgent | Could it wait until morning? | Move to a ticket or business-hours alert |
| Self-resolving | Did it clear before anyone acted? | Add delay or remove |
| Duplicate | Did several alerts fire for one cause? | Correlate into one page |
The goal is a smaller set of pages that each justify the interruption. Our guide to reducing false positive alerts covers the noise side in more depth.
2. Shape shifts around recovery
The Sleep Health review found that limiting shifts to 16 hours or less cut attentional failures by 50% among rotating-shift healthcare workers. Not every setting translates directly, but the principle holds: long stretches carry a cost. Where team size allows, consider follow-the-sun coverage so nobody carries a night pager, and give engineers an explicit day off or late start after a rough night. Daniels and Barteneva both recommend a culture where sleeping in after a bad night is normal, and Barteneva recommends an on-call day off.
3. Build the culture, not only the schedule
Daria Barteneva surveyed and interviewed about 240 engineers and found in her USENIX article that team culture, autonomy and the power to change things shape on-call satisfaction beyond alert counts. Give on-call engineers real authority to fix the causes of pages, and protect time for it.
4. Treat runbooks as first-class work
A tired responder cannot reconstruct tribal knowledge. Keep runbooks short, current and tied to the alert that triggers them, and review them in every post-incident review.
5. Plan for the sleep inertia window
Kovac et al. propose a preliminary framework with three levels: remove the chance of sleep inertia, reduce its severity, and manage the risk of errors during it. For SRE, the third level means guardrails: read-only first steps, peer confirmation for risky changes and clear rollback paths. The authors describe the framework as preliminary, so adapt it and watch the results.
Cutting investigation time without adding tools
Much of the fatigue in a page is the minutes after it: which account, which cluster, which deploy, which dependency. Engineers hop between a metrics tool, a log tool, the cloud console, the CI system and chat while sleep inertia is at its worst. Adding yet another dashboard makes it worse. The better path is a layer that reads from what you already run.
Three practical criteria for any approach:
- It leaves your tools as the source of truth. No duplicated telemetry to maintain.
- It is read-only by default. A groggy responder should gather evidence safely before anything changes. Our piece on safe automation from read-only agents to scoped remediation lays out that progression.
- It shows the evidence. A responder should see why a hypothesis is suggested, not just a conclusion.
A small-team example: a Kubernetes pod crash-loops after a deploy. Without unified context, the responder checks the alert, opens the cluster, finds the pod, searches logs, then checks CI for a recent release. With a unified view, those same checks run side by side in one query. Our walkthrough of checking CloudTrail before chasing logs shows the same idea for AWS incidents, and the guide to PagerDuty incident triage covers the first minutes of a page.
For teams that stay small, the small SRE teams use case describes how to get operational clarity without more tool sprawl.
How Opsphere supports a lighter on-call load

Opsphere is an operational intelligence platform that unifies context across AWS, Kubernetes, observability, CI/CD, security and engineering tools. It combines AI agents, Model Context Protocol (MCP) and more than 300 read-only operational tools, so responders can investigate from one interface, on the web or in their IDE. It connects to existing tools without requiring replacement and does not duplicate telemetry. The AI SRE platform runs structured, evidence-backed investigations with parallel hypothesis testing and context memory.
Plans start with a Community tier at 0 € per month, a Developer plan at 19 € per month, a Team plan at 500 € per month and an Enterprise plan at 450 € per month, with Custom Solutions available. See the pricing page for what each includes.

If your next step is to cut investigation time on the pages that do fire, start free on the Community plan or book a 20-minute demo from opsphere.io.
A 30-day plan for your team
- Week 1: Export a quarter of pages. Classify them using the table above. Count night pages per engineer.
- Week 2: Fix or remove the top five noisiest alerts. Move non-urgent ones to business hours.
- Week 3: Add an explicit recovery rule: a late start or day off after a night with a disruptive page.
- Week 4: Update runbooks for the five most frequent pages and review investigation time with the responders.
Track three measures: night pages per engineer per week, time from page to first useful evidence, and rotation swaps. Falling numbers and fewer swaps are the signal that the load is easing.
FAQ
Is on-call fatigue the same as burnout?
No. Fatigue is the physical and mental exhaustion from disrupted sleep and constant readiness. Burnout, as the WHO’s ICD-11 defines it, adds cynicism and reduced efficacy. Unaddressed fatigue is one common path to burnout.
Can a quiet on-call week still cause fatigue?
Yes. In an Australian study of 228 on-call workers, 56% said sleep was affected even when no calls came. Anticipating a page keeps the mind partly at work.
How long does it take to think clearly after being woken?
Severe sleep inertia symptoms typically last 15 to 30 minutes, and full recovery can take over an hour, according to Gupta et al. Use read-only checks and a second reviewer for risky changes in that window.
What is the fastest way to reduce on-call fatigue?
Audit your pages. Remove or defer those that do not need a human at night, correlate duplicates, then set a recovery rule for engineers after disruptive nights.
Sources
- Burnout | PSNet (AHRQ)
- Burnout and the On-call Pattern | OffShift
- On-call work and sleep: the importance of switching on (Gupta et al.)
- Sleep, Well-Being, and Cognition in Medical Interns (JAMA Network Open)
- Associations between sleep, cognitive function, and work (Sleep Health)
- Identifying and managing fatigue in nurses completing on-call work (British Journal of Nursing)
- The Relationship of On-Call Work with Fatigue, Work-Home Interference
- What Factors Influence the Sleep of On-call Workers?
- On-Call Is Burning You Out | IT Support Group
- The Science of On-Call Anxiety
- Crafting sustainable on-call rotations | Increment
- Google SRE Workbook: On-Call
- Improving On-call Fatigue | USENIX
- A preliminary framework for managing sleep inertia
- Management of shift work disorder (AASM)
Recommended
- PagerDuty Incident Triage: A Practical Guide for SRE Teams
- Safe Automation for SREs: From Read Only Agents to Scoped Remediation
This article is provided for general informational purposes only and does not constitute professional, legal, security, or compliance advice. Please evaluate recommendations against your organization’s specific environment and requirements.
