passenger-datadog-monitor

Send health metrics from Phusion Passenger to Datadog using the StatsD interface on the server agent.

Purpose

Graph and track Passenger threads and possibly detect misbehaving threads before they become a problem.

Tracked Metrics

Aggregated

• Processed requests: min,max,average,total
• Memory usage: min,max,average,total
• Thread uptime: min, max, average
• Request queue depth
• Threads in use vs Max thread configured & Threads used between runs

Discrete

• Process memory usage per Passenger Process PID
• OS thread count per Passenger Process PID
• Requests processed per Passenger Process PID
• Process Idle time per Passenger Process PID

Installation

Downloading from Github

The passenger-datadog-monitor binary can be downloaded from the releases area of this repository for Linux

Building the binary

You will first need to build the passenger-datadog-monitor executable using Go. You can download the source and dependencies, and build the binary by running:

go get -v github.com/Sjeanpierre/passenger-datadog-monitor

Once it completes, you should find your new passenger-datadog-monitor executable in your $GOROOT/bin directory.

Note that if you are building in a different environment from where you plan to deploy, you should configure your target operating system and architecture.

The Makefile in this repository will cross compile for Linux.

Installing the binary

After you've built the executable, you should install it on your server (e.g. in /usr/bin/).

Usage

passenger-datadog-monitor runs as a daemon with a 10 second sampling interval. Monit, God, SupervisorD, or any other daemon management tool should be used to manage the process.

Sample Monit config

check process passenger-datadog-monitor with pidfile /var/run/passenger-datadog-monitor.pid
start program = "/etc/init.d/passenger-datadog-monitor start"
stop  program = "/etc/init.d/passenger-datadog-monitor stop"

You should run passenger-datadog-monitor as root, since access to passenger-status requires root.

Flags

flag	type	description	example
-host	string	StatsD collector IP - useful when running with a Kubernetes DaemonSet or other external collector	-host=100.124.102.21
-port	string	StatsD collector UDP Port - useful when running in Docker or other custom environments	-port=81333
-print	bool	Enable debug and stats printing	-print

Full example:

passenger-datadog-monitor -host=$STATSD_IP -port=$STATSD_PORT

Testing

udp.rb can be run locally when you want to see what is being received on the server side.

Alternatively you can listen using netcat: nc -kulvw 0 8125