Alertmanager FAQ Quick Reference for Interviews (2027)
36 Alertmanager questions answered: routing tree, group_wait, silences, inhibition, time intervals, receivers, clustering, amtool and troubleshooting.
Alertmanager receives alerts from Prometheus, removes duplicates, groups related alerts, applies silences and inhibition rules, and routes the result to receivers such as Slack, PagerDuty, email or a webhook. It does not evaluate alert rules; Prometheus does that. This page is a quick reference of 36 questions on configuration, routing, silences, notifications and clustering, checked against the Prometheus documentation in November 2026.
If you want longer, scenario-driven interview rounds, use our 70+ Alertmanager interview questions. This page is the short version you keep open while you write or debug a config. The current Alertmanager release at the time of writing is 0.35.0 (9 November 2026).
Key takeaways
- Prometheus evaluates rules and pushes firing alerts to Alertmanager. Alertmanager only decides who hears about them, when and how often.
- The routing tree starts at a root route with no matchers. Child routes use
matchers, andmatchandmatch_reare deprecated. - Defaults to remember:
group_wait30s,group_interval5m,repeat_interval4h,resolve_timeout5m. - Silences mute notifications for a set time. Inhibition rules mute alerts while a related alert is already firing.
- Alertmanager clusters over gossip on port 9094 by default, and Prometheus should list every Alertmanager instead of load balancing across them.
amtool config routes testshows which receiver a label set reaches before you deploy a config change.
How the pieces fit
| Component | Job | Where it is configured |
|---|---|---|
| Prometheus alerting rule | Evaluates a PromQL expression and creates alerts | Rule files loaded by Prometheus |
Prometheus alerting block | Tells Prometheus where Alertmanagers are | prometheus.yml |
| Route | Chooses receiver, grouping and timing | route in alertmanager.yml |
| Inhibit rule | Suppresses alerts while a source alert fires | inhibit_rules |
| Silence | Temporary mute by label matchers | Web UI, API or amtool |
| Receiver | Delivers the notification | receivers |
Monitoring context for this page sits in our guide on using Prometheus and Grafana together.
Basics
1. What does Alertmanager do?
It takes alerts that Prometheus (or any client that can call its API) sends, deduplicates them, groups them, mutes the ones that are silenced or inhibited, and sends notifications to receivers. It also repeats notifications for alerts that stay firing.
2. What does Prometheus do and what does Alertmanager do?
Prometheus evaluates alerting rules and decides whether an alert is pending or firing. Alertmanager never sees raw metrics. It only sees alerts with labels and annotations, so routing logic must use labels you attach in the rule.
3. What is the difference between a label and an annotation?
Labels identify an alert and drive routing, grouping and silences. Annotations carry human text such as a summary or a runbook link. Changing a label creates a different alert; changing an annotation does not.
4. What does an alerting rule look like?
A rule has alert, expr, optional for, optional keep_firing_for, plus labels and annotations. Label and annotation values can use templates such as {{ $labels.instance }} and {{ $value }}.
groups:
- name: example
rules:
- alert: InstanceDown
expr: up == 0
for: 5m
labels:
severity: page
annotations:
summary: "Instance {{ $labels.instance }} of job {{ $labels.job }} is down"
5. What is the for clause?
It makes an alert wait in the pending state until the condition has held for that long. An element that is active but not yet firing is pending. It filters brief spikes before Alertmanager is involved at all.
6. What does keep_firing_for do?
It keeps an alert firing for a set time after its condition last matched. Use it to stop flapping, where an alert resolves and fires again within minutes.
7. What is the ALERTS metric?
Prometheus stores a synthetic series called ALERTS for pending and firing alerts, labelled with alertname, alertstate and the alert's other labels. The value is 1 while the alert is in that state. You can graph it in Grafana to see alert history.
8. Which ports does Alertmanager use?
The web UI and API listen on 9093. The cluster listen address defaults to 0.0.0.0:9094 for gossip between peers. An empty --cluster.listen-address disables clustering.
Prometheus to Alertmanager wiring
9. How do you point Prometheus at Alertmanager?
Add an alerting block to prometheus.yml with the targets.
alerting:
alertmanagers:
- static_configs:
- targets:
- alertmanager1:9093
- alertmanager2:9093
- alertmanager3:9093
10. Should I put a load balancer in front of a cluster of Alertmanagers?
No. The Alertmanager documentation says Prometheus should be pointed at a list of all instances so that each one receives every alert. The instances then deduplicate through gossip.
11. What API does Alertmanager expose?
Version 2 lives under /api/v2, for example /api/v2/status. The management endpoints are /-/healthy (always 200), /-/ready and /-/reload (POST).
12. How do you reload the configuration without a restart?
Send a POST to /-/reload or send SIGHUP to the process. Check the Alertmanager log after a reload to confirm the new file was accepted.
Routing and grouping
13. What is a routing tree?
The top-level route matches every alert, so it cannot have matchers. Its child routes add matchers, and the first matching child wins unless continue: true is set. Children inherit grouping and timing values from the parent.
route:
receiver: default
group_by: [alertname, cluster]
group_wait: 30s
group_interval: 5m
repeat_interval: 4h
routes:
- matchers:
- severity="page"
receiver: oncall-pager
- matchers:
- team=~"db|storage"
receiver: data-team
continue: true
14. What are group_wait, group_interval and repeat_interval?
| Setting | Default | Meaning |
|---|---|---|
group_wait | 30s | How long to buffer the first alerts of a new group before the first notification |
group_interval | 5m | How long to wait before notifying about new alerts added to an already notified group |
repeat_interval | 4h | How long to wait before re-sending a notification for a group that is still firing |
15. What does group_by do?
Alerts that share the same values for the listed labels go into one notification. Grouping by [alertname, cluster] turns fifty failing pods of one alert into one message. Setting group_by: ['...'] disables aggregation, so every alert is sent on its own.
16. What does continue: true do?
It lets an alert keep matching later sibling routes after it matches this one. The default is false, so the first match ends the search. Use it to send the same alert to a team channel and to a central log webhook.
17. What is the matcher syntax?
A matcher is a label name, an operator and a value. = and != compare exactly. =~ and !~ use regular expressions anchored at both ends. Quote the whole matcher at the YAML level when the value has special characters.
18. Are match and match_re still valid?
They are deprecated in favour of matchers. New configs should use matchers.
19. Can a child route override timings?
Yes. group_wait, group_interval and repeat_interval are inherited from the parent unless the child sets its own. A paging route often uses a shorter group_wait than a ticket route.
20. How do you test routing before deploying?
Run amtool config routes test with a config file and a label set. With --verify.receivers it exits with code 1 when the receiver list does not match.
amtool config routes test --config.file=alertmanager.yml --tree \
--verify.receivers=oncall-pager severity=page service=database
Silences, inhibition and time intervals
21. What is a silence?
A silence mutes notifications for alerts whose labels match its matchers, for a set period. The alert still fires in Prometheus; Alertmanager just does not notify. Create silences in the web UI, through the API, or with amtool silence add.
amtool silence add alertname=Test_Alert
amtool silence query
amtool silence expire <silence-id>
22. What is an inhibition rule?
It mutes target alerts while a matching source alert is firing, for example muting service alerts while the whole cluster is unreachable. equal lists labels that must have the same value on source and target.
inhibit_rules:
- source_matchers:
- severity="critical"
target_matchers:
- severity="warning"
equal: [alertname, cluster]
23. What happens if equal is left out of an inhibit rule?
The rule applies whenever the source and target matchers match, with no label comparison between them. The documentation also notes that a missing label is treated like an empty one. That can mute far more than you intended, so set equal deliberately.
24. Silence or inhibition: which one should I use?
Use a silence for planned work, such as a maintenance window you know about. Use inhibition for relationships that always exist, such as warning versus critical for the same alert. Silences expire, rules do not.
25. How do time intervals work?
Define time_intervals at the top level with times, weekdays, days of month, months, years and a location. Reference them from a route with mute_time_intervals or active_time_intervals. The default location is UTC, so set an IANA name for local business hours. The root route cannot use these fields.
time_intervals:
- name: business-hours
time_intervals:
- weekdays: ['monday:friday']
times:
- start_time: '09:00'
end_time: '18:00'
location: Asia/Kolkata
26. Do mute intervals mean alerts are lost?
The alerts still exist in Prometheus and in Alertmanager; only notifications are muted during the interval. Do not rely on a mute interval to hide a real outage, and test what your receivers show when the interval ends.
Receivers and notification templates
27. Which receivers are supported?
The configuration reference lists Discord, email, Mattermost, Microsoft Teams (two variants), Jira, Opsgenie, PagerDuty, incident.io, Pushover, Rocket.Chat, Slack, Amazon SNS, Telegram, VictorOps, Webex, WeChat and generic webhooks. The older msteams_configs integration is flagged in the reference as deprecated by Microsoft, and msteamsv2_configs sits beside it.
28. How do you configure Slack?
Set api_url (or api_url_file) and a required channel. send_resolved defaults to false for Slack, so turn it on if you want resolved messages.
receivers:
- name: data-team
slack_configs:
- api_url_file: /etc/alertmanager/slack_url
channel: "#db-alerts"
send_resolved: true
title: '{{ .GroupLabels.alertname }}'
text: '{{ range .Alerts }}{{ .Annotations.summary }}{{ "\n" }}{{ end }}'
29. How do you configure PagerDuty?
Use routing_key for an Events API v2 integration or service_key for the Prometheus integration type. The two are mutually exclusive with their _file variants. severity defaults to error and is a template string.
30. How does the webhook receiver behave?
It POSTs a JSON body to url. send_resolved defaults to true, and max_alerts defaults to 0, which means all alerts in the group are included.
31. How do you set up email?
Set smtp_smarthost and smtp_from under global and give each receiver to in email_configs. The to field accepts a comma separated list. Keep SMTP credentials in a file or secret rather than in a repository.
32. What data can a template use?
Top-level fields are Receiver, Status, Alerts, GroupLabels, CommonLabels, CommonAnnotations and ExternalURL. Alerts.Firing and Alerts.Resolved filter the list. Each alert has Labels, Annotations, StartsAt, EndsAt, GeneratorURL and Fingerprint.
High availability and operations
33. How does Alertmanager clustering work?
Each instance receives alerts independently from Prometheus. Silences and notification state are replicated between peers over a gossip protocol, so a notification sent by one peer is not repeated by the others. Start the first node with defaults and add --cluster.peer=host:9094 on the others. Repeat the flag for each additional peer.
34. What limits the number of alerts per alertname?
The --alerts.per-alertname-limit flag caps active alerts per alertname, and alertmanager_alerts_limited_total counts what was dropped. It protects the service from a runaway rule.
35. What is resolve_timeout?
A global setting with a default of 5m. It applies only when an alert arrives without an EndsAt time, which is rare with Prometheus. Alertmanager then marks the alert resolved after that time if it is not refreshed.
36. What should you do when an expected notification did not arrive?
Work from the front of the pipeline. Check that the alert is firing in Prometheus and not stuck pending. Check that Alertmanager shows it. Run the label set through amtool config routes test. Look for an active silence or an inhibition rule. Finally check receiver credentials and the Alertmanager log for delivery errors.
Checklist for a new Alertmanager config
- Give every alert a
severitylabel and, where you route by owner, ateamlabel. - Choose
group_bylabels that put related alerts in one message. - Add an always-firing test alert so you can prove the whole path works end to end.
- Route critical alerts to a pager and warnings to a ticket or chat channel.
- Add inhibition for warning versus critical of the same alert.
- Run at least two clustered instances and list all of them in Prometheus.
- Test routing with
amtoolin CI before every config change.
Where to go next
For longer practice sets, read the scenario-based Alertmanager questions and the Prometheus alerting interview questions. For the surrounding stack, see the observability tools overview and 10 monitoring alerts every DevOps team should configure. If you prefer to practise with a mentor, our Prometheus and Grafana training runs in the Kharadi, Pune classroom and live online across India.
Frequently asked questions
What is Alertmanager used for?
It handles alerts sent by Prometheus: deduplication, grouping, silencing, inhibition and routing to receivers such as Slack, PagerDuty, email and webhooks.
What are the Alertmanager default timers?
group_wait is 30s, group_interval is 5m and repeat_interval is 4h. The global resolve_timeout is 5m.
How do I stop duplicate alert notifications?
Group alerts with group_by, run Alertmanager as a cluster with every instance listed in Prometheus, and avoid load balancing between instances.
What is the difference between a silence and an inhibition rule?
A silence is a temporary mute you create for matching labels. An inhibition rule is permanent configuration that mutes target alerts while a source alert is firing.
Which port does Alertmanager use?
9093 for the web UI and API, and 9094 for cluster gossip by default.
Are match and match_re deprecated?
Yes. Use matchers in routes and source_matchers and target_matchers in inhibition rules.
How can I test an Alertmanager routing change?
Use amtool config routes test with your config file and a label set. Add --verify.receivers to make it fail when the receiver does not match.
Does Alertmanager send resolved notifications?
It depends on the receiver. Slack defaults send_resolved to false, while the webhook receiver defaults to true.
References
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0