rsyslog_exporter

README

rsyslog_exporter

A prometheus exporter for rsyslog. It accepts rsyslog impstats metrics in JSON format over stdin via the rsyslog omprog plugin and transforms and exposes them for consumption by Prometheus.

Rsyslog Configuration

Configure rsyslog to push JSON formatted stats via omprog:

module(load="omprog")

module(
  load="impstats"
  interval="10"
  format="json"
  resetCounters="off"
  ruleset="process_stats"
)

ruleset(name="process_stats") {
  action(
    type="omprog"
    name="to_exporter"
    binary="/usr/local/bin/rsyslog_exporter [--tls.server-crt=/path/to/tls.crt --tls.server-key=/path/to/tls.key]"
  )
}

The exporter itself logs back via syslog, this cannot be configured at the moment.

Command Line Switches

• web.listen-address - default :9104 - port to listen to (NOTE: the leading : is required for http.ListenAndServe)
• web.telemetry-path - default /metrics - path from which to serve Prometheus metrics
• tls.server-crt - default "" - PEM encoded file containing the server certificate and the CA certificate for use with http.ListenAndServeTLS
• tls.server-key - default "" - PEM encoded file containing the unencrypted server key for use with tls.server-crt

If you want the exporter to listen for TLS (https) you must specify both tls.server-crt and tls.server-key.

Provided Metrics

The following metrics provided by the rsyslog impstats module are tracked by rsyslog_exporter:

Actions

Action objects describe what is to be done with a message, and are implemented via output modules. For each action object, the following metrics are provided:

• processed - messages processed by this action
• failed - number of messages this action failed to process
• suspended - number of times this action was suspended
• suspended_duration - amount of time this action has spent in a suspended state
• resumed - number of times this action has resumed from a suspended state

Inputs

Input objects describe message input sources. For each input object, the following metrics are provided:

• submitted - messages submitted to this input

Queues

Queues in rsyslog are used for the main message queue and for actions. Additionally, each ruleset in an rsyslog configuration may optionally have its own separate main queue. For each queue, the following metrics are provided:

• size - messages currently in queue
• enqueued - total messages enqueued during lifetime of queue
• full - number of times the queue was full
• discarded_full - number of times messages were discarded due to the queue being full
• discarded_not_full - number of times messages discarded but queue was not full
• max_queue_size - maximum size the queue reached during its lifetime

Resources

Rsyslog tracks how it uses system resources and provides the following metrics:

• utime - user time used in microseconds
• stime - system time used in microseconds
• maxrss - maximum resident set size
• minflt - total number of minor faults
• majflt - total number of major faults
• inblock - number of filesystem input operations
• oublock - number of filesystem output operations
• nvcsw - number of voluntary context switches
• nivcsw - number of involuntary context switches

Dynafile Cache

The omfile module can generate file names from a template. A cache of recent filehandles can be maintained, whose sizing can impact performance considerably. The module provides the following metrics:

• requests - number of requests made to obtain a dynafile
• level0 - number of requests for the current active file
• missed - number of cache misses
• evicted - number of times a file needed to be evicted from cache
• maxused - maximum number of cache entries ever used
• closetimeouts - number of times a file was closed due to timeout settings

Dynamic Stats

Rsyslog allows the user to define their own stats namespaces and increment counters within these buckets using Rainerscript function calls.

These are exported as counters with the metric name identifying the bucket, and a label value matching the name of the counter (the label name will always be "counter"). As well as custom metrics, a "global" dynstats namespace is also published with some additional bookeeping counters.

See the dyn_stats documentation for more information.

IMUDP Workerthread stats

The imudp module can be configured to run on multiple worker threads and the following metrics are returned:

• input_called_recvmmsg - Number of recvmmsg called
• input_called_recvmsg -Number of recvmmsg called
• input_received - Messages received