Notifiers

Notifiers are what communicate with your preferred alerting service. Currently Graphite-Pager supports the following notifiers:

• HipChat

• PagerDuty

• PushBullet

• Slack

PagerDuty requires an application key set in the environment as PAGERDUTY_KEY

HipChat requires an application key HIPCHAT_KEY and the room to notify HIPCHAT_ROOM

Slack requires an slack SLACK_URL.

PushBullet requires an application key PUSHBULLET_KEY and optionally comma separated list of devices in PUSHBULLET_DEVICES and/or comma separated list of contacts in PUSHBULLET_CONTACTS.

More notifiers are easy to write, file an issue if there is something you would like!

Documentation url

An attribute of docs\_uls in the configuration will add a link to the documentation of the alert. Currently this is in the format of {docs\_url}/{alert name}#{alert legend name}

Alert Format

Alerts have 4 required arguments and 2 optional arguments.

Required arguments:

• name - Name of thie alert group

• warning - Int for a warning value

• critical - Int for a critical value

• target - Graphtie metric to check, best if aliased

Graphite Pager understands the values for warning and critical in order to check < and >. If warning is less than critical, values above either will trigger an alert. If warning is greater than critical than lower values will trigger the alert.

Example:

    Warning: 1
    Critical: 2

    0 is fine, 3 will be critical

    Warning: 2
    Critical: 1

    0 is critical, 3 is fine.

Optional argument:

• from - The Graphite from parameter for how long to query for ex. -10min default -1min.

• exclude - A list of targets to include. This must match the full target so it is recommended that you use the Graphite function alias() to make these readable.

Ordering of Alerts

Alerts with the same name and target will only be checked once! This is useful if you want to have a subset of metrics with different check times and/or values

- name: Load
  target: aliasByNode(servers.worker-*.loadavg01,1)
  warning: 10
  critical: 20

- name: Load
  target: aliasByNode(servers.*.loadavg01,1)
  warning: 5
  critical: 10

In the above example, any worker-* nodes will alert for anything 10 or higher but the catchall will allow for the remaining metrics to be checked without alerting for worker nodes above 5