Most Asked PagerDuty Questions: On-Call Setup (2026)
Most asked PagerDuty interview questions on escalation policies, schedules, urgency and Events API v2 dedup_key, with limits from PagerDuty docs.
The PagerDuty questions asked most often in DevOps and SRE interviews are about on-call setup: how services, escalation policies and schedules fit together, how escalation timeouts and repeats work, how urgency decides whether someone is woken up, and how alerts are deduplicated through the Events API v2. This guide answers those with figures taken from PagerDuty's own documentation.
This post focuses on on-call design. The keeper post, PagerDuty engineer interview questions, covers the platform more broadly, and integrations are in 65+ PagerDuty interview questions on integrations. For the reliability ideas behind on-call, read the role of SREs in DevOps teams.
Key takeaways
- A service has exactly one escalation policy; one policy can serve many services.
- An escalation policy can have up to 20 rules. The default escalation timeout is 30 minutes, with a minimum of 1 minute for a single-target rule and 3 minutes for multiple targets.
- A policy can repeat up to nine times if nobody acknowledges.
- Schedules only notify people when they are added to an escalation policy that a service uses.
- High-urgency incidents escalate; low-urgency incidents use low-urgency notification rules and do not escalate.
- Events API v2 alerts go to
https://events.pagerduty.com/v2/enqueuewithevent_actionset to trigger, acknowledge or resolve, matched bydedup_key.
How the on-call objects fit together
| Object | What it is | Key limit or default |
|---|---|---|
| Service | Something you monitor and respond to, such as an API or a database | Linked to one escalation policy |
| Integration | How alerts reach a service, for example an Events API v2 integration key | Payloads up to 512 KB |
| Escalation policy | Ordered rules saying who is notified, and when the next level is tried | Up to 20 rules; default timeout 30 minutes |
| Schedule | Who is on call at a given time, built from rotating layers | Up to 60 layers; last layer created takes precedence |
| Urgency | High or low, controls notification rules and escalation | Can follow alert severity dynamically |
Beginner questions
1. What is PagerDuty used for?
PagerDuty takes alerts from monitoring tools, turns them into incidents on the right service, and notifies the person on call through the channels they choose, such as push, SMS, phone or email. It then tracks acknowledgement and resolution so the team has a record of the response.
2. What is the difference between an alert and an incident?
An alert is a signal from a tool, such as "CPU above 90%". An incident is the unit of work a responder acknowledges and resolves. Several alerts can be grouped into one incident, which keeps responders from getting one page per alert during an outage.
3. What states does an incident go through?
Triggered, acknowledged and resolved. Acknowledging stops further notifications and tells the team someone is working on it. Resolving closes it.
4. What is a service in PagerDuty?
A service represents something your team supports. Alerts arrive at a service through integrations, and the service's escalation policy decides who is notified. Good service design mirrors ownership: one team, one set of services.
5. What is an escalation policy?
An ordered set of rules. PagerDuty notifies the targets on the first rule and, if no one acknowledges within the timeout, moves to the next rule. Each rule needs at least one target, either a user or a schedule. If a rule has several targets, they are notified at the same time.
Escalation questions
6. What is the escalation timeout?
The time a responder has to acknowledge before PagerDuty moves to the next rule. PagerDuty's documentation gives a default of 30 minutes, a minimum of 1 minute for a rule with one target and 3 minutes for a rule with several. The first notification is always sent immediately; the timeout only delays escalation.
7. How many escalation rules can a policy have?
Up to 20. PagerDuty recommends at least two, so there is always a backup if the first person does not respond.
8. What happens if nobody acknowledges after the last rule?
If you enable "If no one acknowledges, repeat this policy", PagerDuty loops through the rules again, up to nine times. After the final loop, the incident stays assigned to the last user and notifications stop. That is why a manager or secondary schedule on the last rule is a common design.
9. If I change an escalation policy, does it affect open incidents?
No. PagerDuty's documentation says changes after an incident triggers do not affect open incidents. Fix the policy, then reassign open incidents manually if needed.
10. A rule targets a schedule with no one on call. What happens?
PagerDuty skips that rule and uses the next rule with an on-call responder. If no one is on call anywhere in the policy, an incident is not created, so gaps in coverage are worth checking regularly.
Schedule questions
11. How is a schedule built?
From layers. Each layer is a group of people rotating through the same shift, with a rotation type (daily, weekly or custom), a handoff time and optional restrictions. The final schedule shows the result after layers and restrictions are combined.
12. How do layers interact?
The most recently created layer takes precedence over earlier ones, and only one user per schedule is on call at a time. A common pattern is a base weekly rotation with a second layer that covers weekends.
13. What are restrictions used for?
Restrictions limit when a layer is on call, either by times of day or times of week. They are how you build follow-the-sun rotations, where an India team covers its daytime and another region covers the rest. Restrictions leave gaps, so another layer must fill them.
14. How do you handle a swap or leave?
Create an override for the shift. The override puts another person on call for that window without editing the rotation, so the original pattern continues afterwards.
15. Why can't you delete a layer?
PagerDuty keeps saved layers to preserve on-call history. You set an end date on a layer instead.
Urgency and noise questions
16. What is the difference between high and low urgency?
High-urgency incidents use each responder's high-urgency notification rules and escalate if no one acknowledges. Low-urgency incidents use low-urgency rules and do not escalate. A disk filling slowly overnight is usually low urgency; a checkout API returning errors is high.
17. How can urgency follow alert severity?
Set the service to "Dynamic notifications based on alert severity". Alerts with critical, error or unknown severity create high-urgency incidents, while warning or info create low-urgency ones. If a later alert raises the severity, the incident's urgency goes up; it never goes down.
18. What are support hours?
A service setting that uses one urgency during defined hours and another outside them. For example, an internal reporting tool might be high urgency during office hours and low urgency at night.
19. How do you reduce alert fatigue?
Alert on symptoms users feel rather than every cause, route warnings to low urgency, group related alerts into one incident, and review noisy alerts after each on-call week. Our post on tools to automate incident alerts covers the routing side.
Events API v2 questions
20. How does a monitoring tool send an alert to PagerDuty?
It sends a JSON event to https://events.pagerduty.com/v2/enqueue with the service's integration key as routing_key, an event_action, and a payload that includes the required summary, source and severity. Severity must be info, warning, error or critical.
curl -s -X POST https://events.pagerduty.com/v2/enqueue \
-H 'Content-Type: application/json' \
-d '{
"routing_key": "YOUR_INTEGRATION_KEY",
"event_action": "trigger",
"dedup_key": "checkout-api-5xx",
"payload": {
"summary": "Checkout API 5xx rate above 5%",
"source": "prometheus-prod",
"severity": "critical"
}
}'
21. What does dedup_key do?
It ties events to one alert. A trigger with the same key as an open alert adds to that alert instead of creating a new one. Acknowledge and resolve events with the key update the alert, as long as they use the same routing_key. If you do not set a key, PagerDuty generates one and returns it, so store it if you plan to resolve later. The maximum length is 255 characters.
22. What happens to a resolve event when no alert is open?
It is dropped. It does not create a new alert. After an alert is resolved, a new trigger with the same key creates a new alert.
23. What response codes should a sender handle?
202 means accepted. 400 means the JSON is invalid. 429 and 5xx responses should be retried, preferably with backoff. PagerDuty's rate limits adjust with traffic, so senders must handle 429 rather than assume a fixed limit.
Scenario questions
24. An engineer slept through a page and nobody else was notified. What do you fix?
The escalation policy probably has a single rule with no repeat. Add a second rule with a secondary schedule, enable repeats, and check that the engineer's high-urgency notification rules include a phone call, not only email.
25. One outage created 200 incidents. How do you prevent that next time?
Use stable dedup_key values per problem, group alerts into incidents, and alert on fewer, higher-level symptoms. Afterwards, review which monitors fired and remove duplicates.
26. Your team is spread across India and Europe. How would you design on-call?
Use one schedule with two layers restricted to each region's working hours, so each team covers its daytime. Add a backup rule on the escalation policy pointing to a secondary schedule, and set handoff times that leave a short overlap for handover notes.
27. How do you show on-call is improving?
Track how many incidents happen per shift, how many are actionable, time to acknowledge and time to resolve, and the share that happen at night. Review them in a weekly handover and after each postmortem. See time to restore service as an SRE metric.
28. How would you onboard a new engineer to the on-call rotation?
Have them set up their contact methods and notification rules first, with a phone call for high urgency, and send themselves a test notification. Then add them to the schedule as a shadow, or on a secondary rule of the escalation policy, for a week before they take primary. Walk through the runbooks for the services they will cover and the steps to acknowledge, reassign and escalate an incident. Our post on runbooks for incident response explains what those runbooks should contain.
29. Can change events be sent to any integration key?
No. PagerDuty's Events API v2 overview says change events cannot be sent to ruleset integration keys, which begin with R. Service integration keys accept any event type. Change events record deploys and config changes so responders can see what changed shortly before an incident, which often shortens diagnosis.
30. How large can an Events API v2 payload be?
PagerDuty documents a 512 KB limit. Keep payloads small anyway: put a clear summary in summary, add a few useful fields under custom_details, and link to dashboards or logs rather than pasting them. Responders read the summary on a phone first, so it should say what is broken and where.
31. Who should own a service's escalation policy?
The team that owns the service. Platform teams can provide templates and naming rules, but the owning team knows who can fix the service at 3 a.m. Review the policy whenever team membership changes.
How to prepare
In a PagerDuty account you can use for practice (check PagerDuty's pricing page for current plan options), build a service, an escalation policy with two rules and repeats, a two-layer schedule with restrictions, and an Events API v2 integration. Send trigger, acknowledge and resolve events with curl and watch the incident change. Being able to describe that setup in your own words answers most of the questions above. The scenario-based PagerDuty questions give you more practice.
Incident response and monitoring are part of our Prometheus and Grafana training, which includes routing alerts to on-call tools. Classes run at the Kharadi, Pune classroom and live online.
FAQ
How many rules can a PagerDuty escalation policy have?
Up to 20, according to PagerDuty's documentation.
What is the default escalation timeout in PagerDuty?
30 minutes. The minimum is 1 minute for a single-target rule and 3 minutes for a rule with several targets.
How many times can an escalation policy repeat?
Up to nine times when the repeat option is enabled.
Can one service have two escalation policies?
No. A service has one escalation policy, though one policy can be shared by many services.
Do low-urgency incidents escalate?
No. Low-urgency incidents use low-urgency notification rules and cannot escalate.
What is the Events API v2 endpoint?
https://events.pagerduty.com/v2/enqueue for alert events.
References
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0