On the ceph node install collectd
- tested with collectd-5.7.0-4.el7ost.x86_64

Interval is set to 10 in collectd.conf

write_graphite plugin configured as follows;

LoadPlugin "write_graphite"
<Plugin write_graphite>
  <Node "local">
    Host "192.168.1.52"
    Port "2003"
    Protocol "tcp"
    LogSendErrors true
    Prefix "collectd."
    StoreRates true
    AlwaysAppendDS false
    EscapeCharacter "_"
    PreserveSeparator true
    SeparateInstances true
  </Node>
</Plugin>

5.7 introduces the PreserveSeparator parm, allowing the type instance name to
contain '.'. At the moment I used .'s in the metric name since the same plugin
provides all metrics.

Comments welcome!

This is what an entry looks like as seen in influx
collectd.obj-mon-1.storage.lab.cephmetrics.gauge.ceph.pools._rgw_root.num_bytes_recovered
        |                     |            \     \    \     \
        |                     |             \     \    \     \
        |                     |              |     \    \      \
prefix  | hostname            |  plugin name | type |ceph|metric| metric name
                                                     name  group

In the case of pools, the metric name is prefixed by the pool name

Added cephmetrics.conf to /etc/collectd.d dir

mkdir -p /usr/lib64/collectd/python-plugins

to the plugin dir, I copied
 - cephmetrics.py
 - collectors dir


Known Issues
1. Singlestat panels don't track the graph panel values 100% of the time




Container Configuration - pulled from docker.io registry
- grafana - grafana/grafana (official image)
- graphite - nickstenning/graphite - https://hub.docker.com/r/nickstenning/graphite/

