How to configure alerts in Prometheus for diskspace

Question

We have prometheus running on Win Server box, and WMI exporter on a separate box(client). Able to read client metrics in Prometheus. Now the requirement is the moment Diskspace =>90 % , send an email alert, so that we can run a job to clean up space using an automated job / manual job.

Could you please help on how to configure alert for diskspace >90

Solution 1

assuming you are using https://github.com/martinlindhe/wmi_exporter/blob/master/docs/collector.logical_disk.md you could use something along these lines for > 90 % use

  - alert: DiskSpaceUsage
expr: 100.0 - 100 * (wmi_logical_disk_free_bytes / wmi_logical_disk_size_bytes) > 90
for: 10m
labels:
  severity: high
annotations:
  summary: "Disk Space Usage (instance {{ $labels.instance }})"
  description: "Disk Space on Drive is used more than 90%\n  VALUE = {{ $value }}\n  LABELS: {{ $labels }}"

there are other examples on wmi_exporter repo for default node_exporter metrics ( not sure if available with windows ) it should be

- alert: DiskSpace10%Free
     expr: 100 - (100 * node_filesystem_avail_bytes / node_filesystem_size_bytes) > 90
     labels:
       severity: moderate
     annotations:
       summary: "Instance {{ $labels.instance }} is low on disk space"
       description: "diskspace on {{ $labels.instance }} is used over {{ $value }}% ."
´´´

Solution 2

To send email notification based on alert you need to setup alertmanager with prometheus. Here is the guide how to do that: https://github.com/prometheus/alertmanager

Also you can configure the alert rules. I am using node exporter to fetch node metrics and using the following rule

- alert: DiskSpace10%Free
     expr: node_exporter:node_filesystem_free:fs_used_percents >= 90
     labels:
       severity: moderate
     annotations:
       summary: "Instance {{ $labels.instance }} is low on disk space"
       description: "{{ $labels.instance }} has only {{ $value }}% free."

You can configure the above rule according to WMI exporter and you will be good to go. Hope this helps.

Solution 3

You might want to alert based on if it's going to fill up, not based on how full it is:

- name: node.rules
  rules:
  - alert: DiskWillFillIn4Hours
    expr: predict_linear(node_filesystem_free{job="node"}[1h], 4 * 3600) < 0
    for: 5m
    labels:
      severity: page

https://www.robustperception.io/reduce-noise-from-disk-space-alerts

Comments

[David]

Presumably you are using some sort of custom metric here node_exporter:node_filesystem_free:fs_used_percents is not a core metric. You give no explanation about that. Perhaps you could?

[Brom558]

Use of for: 5m good practice as this should avoid false positives where disk usage spikes but levels off i.e. backup clear down / creation