Deploying mtail


mtail is intended to run one per machine, and serve as monitoring glue for multiple applications running on that machine. It runs one or more programs in a 1:1 mapping to those client applications.

Configuration Overview

mtail is configured with commandline flags.

The --help flag will print a list of flags for configuring mtail.

(Flags may be prefixed with either - or --)

Basic flags necessary to start mtail:

mtail runs an HTTP server on port 3903, which can be changed with the --port flag.


Launching mtail

mtail --progs /etc/mtail --logs /var/log/syslog --logs /var/log/ntp/peerstats

mtail will start to read the specified logs from their current end-of-file, and read new updates appended to these logs as they arrive. It will attempt to correctly handle log files that have been rotated by renaming or symlink changes.

Getting the logs in

Use --logs multiple times to pass in glob patterns that match the logs you want to tail. This includes named pipes.

Polling the file system

mtail polls matched log files every --poll_log_interval, or 250ms by default, the supplied --logs patterns for newly created or deleted log pathnames.

Known and active logs are read until EOF every --poll_interval, or 250ms by default.


mtail --progs /etc/mtail --logs /var/log/syslog --poll_interval 250ms --poll_log_interval 250ms

Setting garbage collection intervals

mtail accumulates metrics and log files during its operation. By default, every hour both a garbage collection pass occurs looking for expired metrics, and stale log files.

An expired metric is any metric that hasn’t been updated in a time specified by a del after form in a program.

A stale log file is any log being watched that hasn’t been read from in 24 hours.

The interval between garbage collection runs can be changed on the commandline with the --expired_metrics_gc_interval and --stale_log_gc_interval flags, which accept a time duration string compatible with the Go time.ParseDuration function.

Runtime error log rate

If your programs deliberately fail to parse some log lines then you may end up generating lots of runtime errors which are normally logged at the standard INFO level, which can fill your disk.

You can disable this with --novm_logs_runtime_errors or --vm_logs_runtime_errors=false on the commandline, and then you will only be able to see the most recent runtime error in the HTTP status console.

Launching under Docker

mtail can be run as a sidecar process if you expose an application container’s logs with a volume.

docker run -d --name myapp -v /var/log/myapp myapp

for example exports a volume called /var/log/myapp (named the same as the hypothetical path where myapps logs are written.

Then launch the mtail docker image and pass in the volume:

docker run -dP \
   --name myapp-mtail \
   --volumes-from myapp \
   -v examples:/etc/mtail \
   mtail --logs /var/log/myapp --progs /etc/mtail

This example fetches the volumes from the myapp container, and mounts them in the mtail container (which we’ve called myapp-mtail). We also mount the examples directory as /etc/mtail in the container. We launch mtail with the logs and progs flags to point to our two mounted volumes.

The -P flag ensures mtail-myapp’s port 3903 is exposed for collection, refer to docker ps to find out where it’s mapped to on the host.

Writing the programme

Read the Programming Guide for instructions on how to write an mtail program.

Reloading programmes

mtail does not automatically reload programmes after it starts up. To ask mtail to scan for and reload programmes from the supplied --progs directory, send it a SIGHUP signal on UNIX-like systems.

For example, if configs are being delivered by a configuration management tool like Puppet, then program Puppet to send a SIGHUP when it has copied a new config file over.

exec { 'reload_mtail_programmes':
  command => "killall -HUP mtail",
  refreshonly = True,

file { mtail_programme:
  source => mtail_programme,
  notify => Exec['reload_mtail_programmes'],

Alternatively, if you’re using scp or some similiar method to copy the programme files without a receiver, then either follow it with a ssh $host 'killall -HUP mtail' or use a tool like inotifywait in a side process next to mtail to watch for changes and send the reload signal.

inotifywait -m /etc/mtail/progs | while read event; do killall -HUP mtail; done

Getting the Metrics Out

Pull based collection

Point your collection tool at localhost:3903/json for JSON format metrics.

Prometheus can be directed to the /metrics endpoint for Prometheus text-based format.

Push based collection

Use the collectd_socketpath or graphite_host_port flags to enable pushing to a collectd or graphite instance.

Configure collectd on the same machine to use the unixsock plugin, and set collectd_socketpath to that unix socket.

mtail --progs /etc/mtail --logs /var/log/syslog,/var/log/rsyncd.log --collectd_socketpath=/var/run/collectd-unixsock

Set graphite_host_port to be the host:port of the carbon server.

mtail --progs /etc/mtail --logs /var/log/syslog,/var/log/rsyncd.log --graphite_host_port=localhost:9999

Likewise, set statsd_hostport to the host:port of the statsd server.

Additionally, the flag metric_push_interval_seconds can be used to configure the push frequency. It defaults to 60, i.e. a push every minute.

Setting a default timezone

The --override_timezone flag sets the timezone that mtail uses for timestamp conversion. By default, mtail assumes timestamps are in UTC.

To use the machine’s local timezone, --override_timezone=Local can be used.


Lots of state is logged to the log file, by default in /tmp/mtail.INFO. See Troubleshooting for more information.

N.B. Oneshot mode (the one_shot flag on the commandline) can be used to check that a program is correctly reading metrics from a log, but with the following caveats:

This mode is useful for debugging the behaviour of mtail programs and possibly for permissions checking.