Captures all nodes metrics shown on the Overview page:
* File descriptors
* Socket descriptors
* Erlang processes
* Memory
* Disk
Not displaying any limits since they would make the variations
impossible to see. For example, when file descriptors go for 90 to 30,
if one of the metrics on the graph is 1048576 (Docker image default for
rabbitmq_node_sockets_total), it's impossible to see the metric change
from 90 to 30. The same problem is present in the current RabbitMQ Management
graphs on the node page, under Node statistics.
No thresholds have been set. Threshold values must be defined as
integers in Grafana 6, we can't reference metrics e.g.
rabbitmq_node_sockets_total. Templating the dashboard would be one way,
but the problem with that is keeping it in sync with limits. It's a more
difficult problem than meets the eye, deferring it for now.
Created on Grafana v6.1
[finishes #164374321]
Bumping all prometheus-related deps to latest stable. Defining them in
rabbitmq-components.mk, so that they can be promoted to all deps in
umbrella.
rabbitmq_management_agent is required for alarm-related metrics to be
available.
Added node label to most `rabbitmq_` metrics. I need help adding them to
mfa_totals - metrics_node_label_test test currently fails. The new unit
tests ensure that label/0 behaves as expected in all cases - made
refactoring easy. Run unit tests via:
gmake eunit EUNIT_MODS=prometheus_rabbitmq_core_metrics_collector
Updating to latest erlang.mk makes running eunit tests much faster: 2s
vs 10s. To do this, comment `ERLANG_MK_*` in Makefile and run `gmake
erlank-mk`.
Gerhard Lazu
Michael Klishin
Diana Corbacho