
- Planned maintenance for a subset of hosts or services.
- Scheduled deployments that temporarily change alerting behavior.
- Temporary suppression of noisy alerts while root cause is being investigated.
- Open the “Silences” tab and click “New silence”.
- Set the start time and duration (or explicit end time). Times are typically in UTC—confirm your UI timezone before creating long silences.
- Add label matchers to select which alerts to silence. Any alert that matches all provided matchers will be silenced.
- Use the UI control (often an “Add matcher” button or a blue plus sign) to add each matcher so you only silence the intended alerts.
- Enter the creator name and an optional comment describing why the silence exists.
- Click to create the silence.
Examples you might use in the UI:
- To silence all EC2 job alerts in production:
- To silence only critical EC2 alerts:
- To silence multiple jobs with a regex (if the UI supports regex):
Tip: Narrow your matchers as much as possible. Start by matching a single label (e.g.,
job="ec2") and expand only if necessary. You can verify the effect by viewing the Alerts list and checking that the intended alerts show as “Silenced”.- The Silences tab lists active and pending silences with their matchers, creators, comments, and remaining duration.
- You can edit a silence to change its end time, matchers, or comment.
- If maintenance ends early, you can expire (cancel) the silence immediately.
- Silences created via the UI are also available through the Alertmanager API, enabling automation.
- After creating a silence, confirm matching alerts are silenced in the Alerts view — they typically display a “Silenced” status or badge.
- If alerts you expected to be silenced are still notifying, double-check label keys/values for typos and ensure the matchers use exact or regex operators correctly.
- For programmatic validation, use the Alertmanager API to list silences and their matchers.
Use silences carefully. Broad or overly permissive matchers can suppress important alerts you still need to act on. Always scope silences to the minimal set of labels required.
- Did you scope matchers narrowly (job, instance, region, severity)?
- Did you set the correct start time, end time, or duration?
- Did you add a creator name and a clear comment explaining the reason?
- Did you verify the silence in the Alerts view and/or Alertmanager API?
- Alertmanager documentation: https://prometheus.io/docs/alerting/latest/alertmanager/