Apache HTTP

Auto-Discovery

Health Check

Metrics

Logs

Agent uses a HTTP check if the service listen on port 80.

To enable metrics gathering, ensure Bleemeo agent could access Apache status on the URL http://server-address/server-status?auto. This should be done by default with Apache on Ubuntu or Debian. If not, this usually means to add the following to your default virtual host (e.g. the first file in /etc/apache2/sites-enabled/):

LoadModule status_module /usr/lib/apache2/modules/mod_status.so

<Location /server-status>
    SetHandler server-status
    # You may want to limit access only to Bleemeo agent
    # require ip IP-OF-BLEEMEO-AGENT/32
</Location>
ExtendedStatus On

If some auto-detected parameters are wrong, you can manually override them by adding the following to /etc/glouton/conf.d/99-local.conf:

service:
  [...]

  # For an Apache running outside a container
  - type: "apache"
    address: "127.0.0.1"
    port: 80   # HTTP listener, Agent don't support HTTPS here
    http_path: "/"   # This is the path used for availability check, not metrics gathering.
    http_host: "127.0.0.1:80"   # Host header sent. Like other option, you can omit them and default value will be used.

  # For an additional Apache running outside a container
  - type: "apache"
    instance: "NAME_ASSOCIATED_WITH_YOUR_ADDITIONAL_SERVICE"
    address: "127.0.0.1"
    port: 81   # HTTP listener, Agent don't support HTTPS here
    http_path: "/"   # This is the path used for availability check, not metrics gathering.
    http_host: "127.0.0.1:80"   # Host header sent. Like other option, you can omit them and default value will be used.

  # For an Apache running in a Docker container
  - type: "apache"
    instance: "CONTAINER_NAME"
    address: "172.17.0.2"
    port: 80   # HTTP listener, Agent don't support HTTPS here
    http_path: "/"   # This is the path used for availability check, not metrics gathering.
    http_host: "127.0.0.1:80"   # Host header sent. Like other option, you can omit them and default value will be used.

When using Docker, you may use Docker labels to set http_path and http_host:

docker run --label glouton.http_path="/readiness" --label glouton.http_host="my-host" [...]

Agent gathers the following metrics:

Metric	Description
service_status	Status of Apache
apache_busy_workers	Number of Apache worker busy
apache_busy_workers_perc	Apache worker busy in percent
apache_bytes	Network traffic sent by Apache in bytes per second
apache_connections	Number of client connections to Apache server
apache_idle_workers	Number of Apache workers waiting for an incoming request
apache_max_workers	Maximum number of Apache worker configured
apache_requests	Number of requests per second
apache_uptime	Time spent since Apache server start in seconds

Apache keeps track of server activity in a structure known as the scoreboard. There is a slot in the scoreboard for each worker and it contains the status of this worker. The size of the scoreboard is the maximum of concurrent users that Apache could server.

Metric	Description
apache_scoreboard_waiting	Number of workers waiting for an incoming request. It’s the same as apache_idle_workers
apache_scoreboard_starting	Number of workers starting up
apache_scoreboard_reading	Number of workers reading an incoming request
apache_scoreboard_sending	Number of workers processing a client request
apache_scoreboard_keepalive	Number of workers waiting for another request via keepalive
apache_scoreboard_dnslookup	Number of workers looking up a hostname
apache_scoreboard_closing	Number of workers closing their connection
apache_scoreboard_logging	Number of workers writing in log files
apache_scoreboard_finishing	Number of workers gracefully finishing a request
apache_scoreboard_idle_cleanup	Number of idle workers being killed
apache_scoreboard_open	Number of slots with no worker in the scoreboard

Apache do not start all workers when they are not needed (e.g. if there is enough workers waiting, Apache reuse them and don’t start a new one).

The sum of all scoreboard items is the maximum of concurrent requests Apache can serve. This sum is calculated and stored in apache_max_workers metric.