Time-data statistics (InfluxDB, Grafana)

dhcperfcli is able to handle time-data statistics, which are collected and reported in real-time. This provides much more detail than the other available statistics (ongoing and end report) that are displayed on the console.

Time-data statistics are sent through HTTP (or HTTPS) to an Influx database. They can then be displayed in various ways through Grafana dashboards.

Multiple instances of dhcperfcli can send data simultaneously to the same Influx back-end.

Prerequisites

To work with InfluxDB, you need to build dhcperfcli with libcurl (and optionally json-c). Refer to the installation page for details.

You will also need to set InfluxDB and Grafana servers up. This is pretty straighforward, refer to their respective documentation.

Authentication

dhcperfcli must be able to perform REST requests serviced by an Influx server.

The following methods are supported:

basic authentication (with login and password)
bearer authentication (with token)

Configuration

Time-data must be specified to dhcperfcli through a configuration file. Refer to the sample configuration file provided which details all available parameters.

Configuration example:

conf/dhcperfcli-influx.conf

time-data test-dhcperfcli {
	destination = influx
	time_interval = 1.0
	max_backlog = 300
	influx {
		server = "10.11.12.84:8086"
		username = "admin"
		password = "*****"
		database = "dpc"
		uri = "http://${server}/write?db=${database}"
	}
}

The Influx database (here dpc) must be created beforehand. Optionally, you can create a retention policy which will take care of automatically removing old statistics (by default, data are retained indefinitely).

For example (using InfluxQL):

> CREATE DATABASE "dpc"
> CREATE RETENTION POLICY "one_day" ON "dpc" DURATION 1d REPLICATION 1 DEFAULT

TLS

TLS is automatically enabled when the configured URI scheme is "https". This is required if InfluxDB is configured with https-enabled = true (which is strongly recommended).

In addition, it is possible to handle server certificate chain validation. This allows the client to know that the server can be trusted. This is explained in the sample configuration file (see "tls" section).

For example:

time-data test-dhcperfcli {
	destination = influx
	influx {
		server = "10.11.12.84:8086"
		database = "dpc"
		uri = "https://${server}/write?db=${database}"
		tls {
			ca_file = "/etc/pki/tls/certs/ca-bundle.crt"
			check_cert = yes
			check_cert_cn = yes
		}
	}
}

Usage example

Execute dhcperfcli as usual, with the added time-data configuration.

For example:

echo "Client-Hardware-Address=\"%{ethaddr.rand}\"" | dhcperfcli --conf-file conf/dhcperfcli-influx.conf -T -L 60 -p 32 -r 1000 -g 10.11.12.1 10.11.12.42 discover

Statistics will be automatically collected, and written once every second (or the configured time_interval) to the specified destination.

If that destination is an Influx database, it must be available when the program starts. It can become unavailable later on, in which case a backlog of the time-data statistics is retained, so it can be written once Influx becomes available again. Reconnection is handled automatically.

An error is logged once when the Influx server becomes unavailable. Subsequent errors are suppressed until the connection is restored. If that never happens, when the program stops it will log a warning about discarded data points.

For example:

Error : Time-data: Failed to write to InfluxDB: Request failed: curl error (7) [Couldn't connect to server]
Info  : Time-data: Further errors will be suppressed until destination is available again
(...)
Warn  : Time-data: Discarded 22 data point(s) (of 63) due to destination unavailability