forked-synapse/docs/metrics-howto.rst

69 lines
2.4 KiB
ReStructuredText
Raw Normal View History

2015-04-23 11:07:49 -04:00
How to monitor Synapse metrics using Prometheus
===============================================
2017-02-19 18:06:08 -05:00
1. Install prometheus:
2015-04-23 11:07:49 -04:00
2017-02-19 18:06:08 -05:00
Follow instructions at http://prometheus.io/docs/introduction/install/
2015-04-23 11:07:49 -04:00
2017-02-19 18:06:08 -05:00
2. Enable synapse metrics:
2015-04-23 11:07:49 -04:00
2017-02-19 18:06:08 -05:00
Simply setting a (local) port number will enable it. Pick a port.
prometheus itself defaults to 9090, so starting just above that for
locally monitored services seems reasonable. E.g. 9092:
2015-04-23 11:07:49 -04:00
2017-02-19 18:06:08 -05:00
Add to homeserver.yaml::
metrics_port: 9092
Also ensure that ``enable_metrics`` is set to ``True``.
2017-02-19 17:55:48 -05:00
2017-02-19 18:06:08 -05:00
Restart synapse.
2015-04-23 11:07:49 -04:00
2017-02-19 18:06:08 -05:00
3. Add a prometheus target for synapse.
It needs to set the ``metrics_path`` to a non-default value (under ``scrape_configs``)::
- job_name: "synapse"
metrics_path: "/_synapse/metrics"
static_configs:
- targets: ["my.server.here:9092"]
2016-10-31 11:06:52 -04:00
2017-02-19 18:06:08 -05:00
If your prometheus is older than 1.5.2, you will need to replace
``static_configs`` in the above with ``target_groups``.
2017-02-19 18:06:45 -05:00
Restart prometheus.
2017-02-19 18:06:08 -05:00
2016-10-31 11:06:52 -04:00
Standard Metric Names
---------------------
As of synapse version 0.18.2, the format of the process-wide metrics has been
changed to fit prometheus standard naming conventions. Additionally the units
have been changed to seconds, from miliseconds.
================================== =============================
New name Old name
---------------------------------- -----------------------------
process_cpu_user_seconds_total process_resource_utime / 1000
process_cpu_system_seconds_total process_resource_stime / 1000
process_open_fds (no 'type' label) process_fds
================================== =============================
The python-specific counts of garbage collector performance have been renamed.
=========================== ======================
New name Old name
--------------------------- ----------------------
python_gc_time reactor_gc_time
python_gc_unreachable_total reactor_gc_unreachable
python_gc_counts reactor_gc_counts
=========================== ======================
The twisted-specific reactor metrics have been renamed.
==================================== =====================
2016-10-31 11:06:52 -04:00
New name Old name
------------------------------------ ---------------------
python_twisted_reactor_pending_calls reactor_pending_calls
2016-10-31 11:06:52 -04:00
python_twisted_reactor_tick_time reactor_tick_time
==================================== =====================