varnishprom

Varnish Prometheus Exporter

This Varnish prometheus exporter exposes metrics gathered from the Varnish shared memory log AND from Varnish internal counters (varnishstat) depending on the flags used.

Use case

Most prometheus exporters for varnish will just parse and export the varnishstat metrics. This exporter also parses the varnishlog and exports the metrics from the log. This is useful for monitoring exactly what you want from within VCL by adding std.log("<keyword>=<metricname> label1=<value> label2=<value>") to the part of the VCL you like to create a counter for

The default keyword is prom, but this can be changed with the -k flag.

Examples

Count the number of retries above 1 in the backend response:

sub vcl_backend_response {
    if (bereq.retries > 0) {
      std.log("prom=backend_retries retries=" + bereq.retries);
    }
}

A full working vcl example to count hits/misses for each backend is here

Usage

./varnishprom --help
Usage of ./varnishprom:
❯ ./varnishprom --help                                     
Usage of ./varnishprom:
  -a string
        Varnish admin interface (default "127.0.0.1:42717")
  -h string
        Hostname to use in metrics, defaults to hostname -S (default "airmone")
  -i string
        Listen interface for metrics endpoint (default "127.0.0.1:7083")
  -k string
        logkey to look for promethus metrics (default "prom")
  -l    Start varnishlog parser
  -p string
        Path for metrics endpoint (default "/metrics")
  -s    Start varnshstats parser

It needs to be started with the -l and/or -s flags to start the varnishlog and/or varnishstat parsers.

systemd file

Example systemd file to run the exporter as a service.

Bugs

The varnishstat parses is trying to be smart, but might be stupid. It will try to parse the output of varnishstat -1