Skip to content

Latest commit

 

History

History
88 lines (79 loc) · 3.56 KB

README.md

File metadata and controls

88 lines (79 loc) · 3.56 KB
Raw

Overview

This implements a rsyslog configuration that logs directly into Influxdb.

I tested the suggested rsyslog integration with Telegraf and got a lot of error logs like

action 'action 1' resumed (module 'builtin:omfwd')

I wasn't able to fix the root cause and decided to use a different approach. My setup uses a log template and the Influx UDP Input to deliver the data directly into a database. It runs smooth without reducing functionality.

Setup

1.) Enable UDP Input for Influxdb

Find the [[udp]] section to enable the input. Modify the database parameter to write it into your db, or it will go into the standard "udp" database.

/etc/influxdb/influxdb.conf

[[udp]]
  enabled = true
  # bind-address = ":8089"
  # database = "udp"

Restart influx and make sure to adjust your UDP buffer size!

2.) Add a configuration file to rsyslog

Create a new file in rsyslogs config folder.

/etc/rsyslog.d/50-influx.conf

template(
  name="Influx"
  type="string"
  string="syslog,hostname=%HOSTNAME%,appname=%PROGRAMNAME%,facility=%SYSLOGFACILITY-TEXT%,severity=%SYSLOGSEVERITY-TEXT% severity_code=%SYSLOGSEVERITY%i,facility_code=%SYSLOGFACILITY%i,procid=\"%PROCID%\",message=\"%$!msg%\"\n"
)
set $!msg = replace($msg, "\"", "'");
action(type="omfwd" Protocol="udp" Target="<your influx server>" Port="8089"  Template="Influx")

This logging template writes into a measurement called syslog. Rename it if you like. Restart rsyslog if you are done.

rsyslog configuration for the Influx host itself

If you want to collect data on the host running the InfluxDB you need a slightly different configuration to avoid a logging feedback loop. Here it is necessary to exclude influxd logs from the forwarding.

template(
  SAME AS ABOVE
)
set $!msg = replace($msg, "\"", "'");
if $programname != 'influxd' then {
  action(type="omfwd" Protocol="udp" Target="localhost" Port="8089"  Template="Influx")
}

Using the measurement

A simple test to check if your setup works is to ask the database directly.

root@collector ~ # influx
Connected to http://localhost:8086 version 1.7.1
InfluxDB shell version: 1.7.1
Enter an InfluxQL query
> use udp   (or your DB name)
Using database udp
> select * from syslog limit 3
name: syslog
time                appname facility facility_code hostname message                                                          procid severity severity_code
----                ------- -------- ------------- -------- -------                                                          ------ -------- -------------
1550679421543669928 CRON    authpriv 10            colector  pam_unix(cron:session): session opened for user root by (uid=0) 3317   info     6
1550679421545719847 CRON    cron     9             colector  (root) CMD (   cd / && run-parts --report /etc/cron.hourly)     3318   info     6
1550679421546549914 CRON    authpriv 10            colector  pam_unix(cron:session): session closed for user root            3317   info     6

The written data fields and tags are based on Telegrafs syslog measurement. Dashboards and queries should work as before.

> show field keys from syslog
name: syslog
fieldKey      fieldType
--------      ---------
facility_code integer
message       string
procid        string
severity_code integer

> show tag keys from syslog
name: syslog
tagKey
------
appname
facility
hostname
severity