Cookbook

This page provides recipes to get the best out of sauna.

HAProxy health checks

When load balancing servers with HAProxy you might want to enable health checks. If one of your servers is running out of memory, overloaded or not behaving properly you can remove it from the pool of healthy servers.

To help you achieve that sauna has a special consumer that listens on a TCP port and returns the status of the server when getting an incoming connection. This consumer is the TCPServer consumer.

Enabling the TCP server

To enable the TCP server add it to the list of active consumers:

---
consumers:

  TCPServer:
    port: 5555

Let’s launch sauna and try to connect to the port 5555:

$ nc localhost 5555
OK

As the system is healthy sauna answers OK. Let’s try by switching a check to CRITICAL:

$ nc localhost 5555
CRITICAL

Configuring HAProxy

We will configure HAProxy to remove a server from the pool as soon as it is not in OK state. For that we will use tcp-check.

Assuming you have a load balancing frontend/backend already set up, activate checks:

backend webfarm
    mode http
    option tcp-check
    tcp-check connect port 5555
    tcp-check expect string OK
    server web01 10.0.0.1:80 check
    server web02 10.0.0.2:80 check
    server web03 10.0.0.3:80 check

• option tcp-check enables level 3 health checks
• tcp-check connect port 5555 tells HAProxy to check the port 5555 of servers in the pool
• tcp-check expect string OK consider the server down if it does not answer OK

Reusing Nagios plugins

Nagios plugins are still very popular, their simple API can be considered the de facto standard for monitoring checks. Sauna can run Nagios plugins through its Command plugin.

Here we will run the famous check_http for monitoring Google. Add a Command plugin to sauna.yml:

---
plugins:

  - type: Command
    checks:
      - type: command
        name: check_google
        command: /usr/lib/nagios/plugins/check_http -H www.google.com

Run sauna:

$ sauna
ServiceCheck(name='check_google', status=0, output='HTTP OK: HTTP/1.1 302 Found')

Note

Nagios plugins may be convenient but they rely on forking a process for each check. Consider using some of the lighter sauna core plugins if this is an issue.

Passive host checks

When it is not possible to check if a host is alive by sending a ping (for instance when the host is in a private network), Nagios and Shinken can use passive host checks submitted via NSCA.

Passive host checks work like normal service checks, except that they don’t carry a service name:

---
plugins:

  - type: Dummy
    checks:
      - type: dummy
        name: ""
        status: 0
        output: Host is up and running

Configure your monitoring server to consider your host down if no passive host check has been received for one minute:

define host {
    address                192.168.20.3
    host_name              test
    use                    generic-host
    check_command          check_dummy!2
    active_checks_enabled  0
    passive_checks_enabled 1
    check_freshness        1
    freshness_threshold    60
}