From a6ba3d6f9db027ac8ef212d58dd18b7e4dde9f61 Mon Sep 17 00:00:00 2001
From: David Wragg <david@wragg.org>
Date: Mon, 4 Apr 2016 11:13:13 +0100
Subject: [PATCH] Update the Prometheus page

The integration is much more automatic these days, so there is less to
say.
---
 site/prometheus.md | 112 ++++++++-------------------------------------
 1 file changed, 19 insertions(+), 93 deletions(-)

diff --git a/site/prometheus.md b/site/prometheus.md
index eb13cd7..c97a0e3 100644
--- a/site/prometheus.md
+++ b/site/prometheus.md
@@ -3,102 +3,28 @@ layout: page
 title: Integrating with Prometheus
 ---
 
-Flux has a two-part integration with [Prometheus][prom-site]: firstly,
-the Flux daemon `fluxd` exposes metrics that Prometheus can scrape;
-and secondly, the web dashboard will query those metrics to populate
-its charts and gauges.
+Flux has a two-part integration with
+[Prometheus](https://github.com/prometheus/prometheus): firstly, the
+Flux daemon `fluxd` exposes metrics that Prometheus can scrape; and
+secondly, the web dashboard will query those metrics to populate its
+charts and gauges.
 
 ## Exposing stats to Prometheus from fluxd
 
-To tell `fluxd` to expose stats for Prometheus, supply the
-`--listen-prometheus` option, with a listening address (`:9000` is
-fine).
+Prometheus needs some way discover all of the hosts running fluxd, so
+that it can probe them for metrics.  The docker image
+`weaveworks/flux-prometheus-etcd` provides a Prometheus server that is
+customized to automatically discover `fluxd` instances via etcd.  You
+don't need to supply any options to `fluxd` to enable the integration
+with Prometheus, although you can customize the port number on which
+`fluxd` listens for connections from Prometheus with the
+`--listen-prometheus` option (it defaults to port 9000).
 
-The `bin/run-flux` script assumes this is what you want, and does it
-for you.
+Apart from the enhancements to support discovery via etcd,
+`weaveworks/flux-prometheus-etcd` is just a plain Prometheus server.
+If you already have a Prometheus server deployed, you can use that.
+Though you'll need to arrange to use one of Prometheus' service
+discovery mechanisms for it to scrape metrics from all hosts running
+`fluxd`.
 
-## Running Prometheus
 
-It's easy to run Prometheus under Docker; however, you will need some
-way of telling Prometheus about all of the hosts running the fluxd, so
-it knows to scrape them for stats. See below for some ways to do that.
-
-### Configuring Prometheus
-
-In the examples below, I've put the configuration in a file
-`prometheus.yml` in the current directory, and mounted it into the
-container filesystem in the place Prometheus expects it. I left the
-`global` section empty, which will work, though you may of course have
-your own configuration to add there.
-
-#### Using hostnames or IP addresses
-
-If you can supply a static list of hostnames or host IP addresses, you
-can just put them in a stanza in the configuration file. For example,
-using host names:
-
-```yaml
-global:
-
-scrape_configs:
-  - job_name: 'flux'
-    scrape_interval: 5s
-    scrape_timeout: 10s
-    target_groups:
-      - targets:
-        - host-one:9000
-        - host-two:9000
-        - host-three:9000
-```
-
-Note the port numbers, which match whatever you told fluxd to listen
-on with `--listen-prometheus` (or `9000` if you use `run-flux`).
-
-You can then give the Prometheus container the IP addresses (and the
-configuration, with a volume mount) when starting it:
-
-```bash
-docker run -d -p 9090:9090 \
-       -v $PWD/prometheus.yml:/etc/prometheus/prometheus.yml \
-       --add-host "host-one:192.168.99.101" \
-       --add-host "host-two:192.168.99.102" \
-       --add-host "host-three:192.168.99.103" \
-       prom/prometheus:master
-```
-
-#### Using Prometheus's service discovery configs
-
-Prometheus has a [handful of "service discovery" mechanisms][prom-sd],
-which let you put a record of the hosts somewhere, which Prometheus
-will poll.
-
-For example, if you happen to be running all your containers on a
-[Weave][weave-site] network, this can be as easy as making a DNS entry
-for each host,
-
-```bash
-weave dns-add $(weave expose) weave -h flux.weave.local
-```
-
-and adding a `dns_sd_configs` stanza to the Prometheus configuration:
-
-```yaml
-global:
-
-scrape_configs:
-  - job_name: 'flux'
-    scrape_interval: 5s
-    scrape_timeout: 10s
-    dns_sd_configs:
-        - port: 9000
-          type: A
-          names:
-            - 'flux.weave.local'
-```
-
-The `$(weave expose)` is needed to give the host an IP address on the
-Weave network, since fluxd runs in the host's network namespace.
-
-[prom-site]: https://github.com/prometheus/prometheus
-[prom-sd]: http://prometheus.io/docs/operating/configuration/#scrape-configurations-scrape_config
-[weave-site]: https://github.com/weaveworks/weave