Skip to main content

How to Send Prometheus Alertmanager Alerts

Fix Inventory constantly monitors your infrastructure, and can alert you to any detected issues. Prometheus Alertmanager is a tool to manage and escalate alerts. In this guide, we will configure Fix Inventory to send alerts to Alertmanager with a custom command.


This guide assumes that you have already installed and configured Fix Inventory to collect your cloud resources.

You will also need a running installation of Alertmanager.


  1. Open the fix.core.commands configuration by executing the following command in Fix Inventory Shell:

    > config edit fix.core.commands
  2. Add the following command definition to the configuration, replacing http://localhost:9093 with the URL of your Alertmanager instance:

    ​ commands:
    - info: 'Send result of a search to alertmanager'
    name: 'alertmanager'
    - name: 'duration'
    description: 'Alert duration'
    default: '3h'
    - name: 'description'
    description: 'Alert message'
    default: 'Fix Inventory Alert'
    - name: 'name'
    description: 'Globally unique alert name'
    default: null
    - name: 'alertmanager_url'
    description: 'Alertmanager URL'
    default: 'http://localhost:9093'
    template: 'aggregate sum(1) as count | jq --no-rewrite "if (.count // 0)==0 then [] else [.count | tostring] end" | flatten | jq --no-rewrite [{status: "firing", labels: {alertname: "{{name}}", issued_by: "Fix Inventory"}, annotations: {summary: ("Found "+.+ " violations!"), "description": "{{description}}"}{{#duration}}, startAt:"{{now}}", endsAt:"{{duration.from_now}}"{{/duration}}}] | http POST {{alertmanager_url}}/api/v1/alerts'
  3. Save and quit the editing interface.

    We now have a custom command alertmanager with the following parameters:

    ParameterDescriptionDefault Value
    nameGlobally unique alert namenull
    descriptionAlert descriptionFix Inventory Alert
    durationAlert duration3h
    alertmanager_urlAlertmanager URLhttp://localhost:9093

    The help command can also be used to access this information in the command-line interface:

    > help alertmanager
  4. Define search criteria that will trigger an alert. For example, let's say we want to send alerts whenever we find a Kubernetes Pod updated in the last hour with a restart count greater than 20:

    > search is(kubernetes_pod) and pod_status.container_statuses[*].restart_count > 20 and last_update<1h
    ​kind=kubernetes_pod, name=db-operator-mcd4g, restart_count=[42], age=2mo5d, last_update=23m, cloud=k8s, account=prod, region=kube-system
  5. Now that we've defined the alert trigger, we will simply pipe the result of the search query to the alertmanager command, replacing the name with your desired alert name:

    > search is(kubernetes_pod) and pod_status.container_statuses[*].restart_count > 20 and last_update<1h | alertmanager name=pod-restart-alert

    If the defined condition is currently true, you should see a new alert in Alertmanager:

    Example Alertmanager alert

  6. Finally, we want to automate checking of the defined alert trigger and send alerts to Alertmanager whenever the result is true. We can accomplish this by creating a job:

    > jobs add --id alert_on_pod_failure--wait-for-event post_collect 'search is(kubernetes_pod) and pod_status.container_statuses[*].restart_count > 20 and last_update<1h | alertmanager name=pod-restart-alert'

Further Reading