Watchdog Monitor

このページは日本語には対応しておりません。随時翻訳に取り組んでいます。翻訳に関してご質問やご意見ございましたら、お気軽にご連絡ください。

Overview

Watchdog is an algorithmic feature for APM, Infrastructure, and Logs. It automatically detects potential issues by continuously observing trends and patterns in metrics and logs, and looking for atypical behavior.

Monitor creation

To create a Watchdog monitor in Datadog, use the main navigation: Monitors –> New Monitor –> Watchdog.

Select alert type

In this section, choose between an APM, Infrastructure, or Logs alert:

An APM alert is created when Watchdog detects anomalous behavior on your system’s services or their child resources.

Select sources

Choose your primary tags, service, and resource from the dropdown menus.

After your selections are made, the graph at the top of the monitor creation page displays the matching Watchdog events over time, along with a list of events.

Infrastructure-wide alerts can include issues over the following integrations:

System: Host-level memory usage (memory leak), TCP retransmit rate, etc.
Redis
PostgreSQL
NGINX
Amazon Web Services: For the S3, ELB/ALB/NLB, CloudFront, and DynamoDB Amazon services.

After selecting Infrastructure, the graph at the top of the monitor creation page displays Watchdog events over time, along with a list of events.

Select sources

No selection is necessary. You are notified when Watchdog detects issues across your infrastructure.

A logs alert indicates that either a new pattern of error logs has been detected or that there has been an increase in an existing pattern of error logs.

The Watchdog monitor's edit page showing the alert category set to logs, alert type as log anomaly, env set to production, service set to ad-server, and the monitor's title is Anomaly Detected in Production Ad Server

Select sources

Choose your environment, service, log source, log status, and log anomaly type from the dropdown menus.

The first four parameters (environment, service, log source, log status) refer to attributes of the logs themselves. The monitor only triggers if Watchdog finds an anomaly on logs with attributes that match the ones specified.

The fifth parameter, log anomaly type, refers to whether the anomaly describes a new pattern of error logs or an increase in an existing pattern of error logs.

After the selections are made, the graph at the top of the monitor creation page displays the matching Watchdog events over time, along with a list of events.