cronner
is a command line utility to that wraps periodic (cron) jobs for statistics gathering and success monitoring. The amount of time the command took to ran, as well as the return code, are emitted as vanilla statsd metrics to port 8125. It also implements file-level locking for very simple, and dumb, job semaphore.
The utility also supports emitting DogStatsD Events under the following occasions:
- job start and job finish
- job finish if the job failed
- if the job is taking too long to finish running
If your statsd agent isn't DogStatsD-compliant, I'm not sure what the behavior will be if you an emit an event to it.
For the finish DogStatsD event, the return code and output of the command are provided in the event body. If the output is too long, it is truncated. This output can optionally be saved to disk only if the job fails for later inspection.
Cronner is released under the BSD 3-Clause License. See the LICENSE
file for
the full contents of the license.
Usage:
cronner [OPTIONS] -- command [arguments]...
Application Options:
-d, --lock-dir= the directory where lock files will be placed (default: /var/lock)
-e, --event emit a start and end datadog event
-E, --event-fail only emit an event on failure
-F, --log-fail when a command fails, log its full output (stdout/stderr) to the log directory using the UUID as the filename
-g, --group=<group> emit a cronner_group:<group> tag with statsd metrics
-G, --event-group=<group> emit a cronner_group:<group> tag with Datadog events, does not get sent with statsd metrics
-k, --lock lock based on label so that multiple commands with the same label can not run concurrently
-l, --label= name for cron job to be used in statsd emissions and DogStatsd events. alphanumeric only; cronner will lowercase it
--log-path= where to place the log files for command output (path for -F/--log-fail output) (default: /var/log/cronner)
-L, --log-level= set the level at which to log at [none|error|info|debug] (default: error)
-N, --namespace= namespace for statsd emissions, value is prepended to metric name by statsd client (default: cronner)
-p, --passthru passthru stdout/stderr to controlling tty
-P, --use-parent if cronner invocation is runner under cronner, emit the parental values as tags
-s, --sensitive specify whether command output may contain sensitive details, this only avoids it being printed to stderr
-V, --version print the version string and exit
-w, --warn-after=N emit a warning event every N seconds if the job hasn't finished, set to 0 to disable (default: 0)
-W, --wait-secs= how long to wait for the file lock for (default: 0)
Help Options:
-h, --help Show this help message
The label (-l
, --label
) flag is required.
To run the command /bin/sleep 10
and emit the stats as cronner.sleeptyime.time
and cronner.sleepytime.exit_code
you would run:
$ cronner -l sleepytime -- /bin/sleep 10
To note, --
in the command line arguments tells cronner to stop parsing CLi flags. It then grabs the rest of the arguments as the command to execute.
The cronner
process sets a few environment variables for subprocesses to consume if they wish.
The CRONNER_PARENT_UUID
environment variable is the canonical way for determining whether or not we are running under cronner
.
Variable | Description |
---|---|
CRONNER_PARENT_UUID |
this is the UUID being used by the parent cronner process for its events; use this being set to determine if running under cronner |
CRONNER_PARENT_EVENT_GROUP |
this is the event group used by the parent process for its events |
CRONNER_PARENT_GROUP |
this is the group used by the parent process for its metrics |
CRONNER_PARENT_NAMESPACE |
this is the namespace used by the parent process for its metrics |
CRONNER_PARENT_LABEL |
this is the label used by the parent process for its metrics |
If you invoke the cronner
command with the -P/--use-parent
flag it will look for these variables and tag the events and metrics emissions
with their values. It lowercases the variable name before emitting the tag, so CRONNER_PARENT_GROUP
becomes cronner_parent_group
.
If you were to have a UDP listener on port 8125 on localhost, the statsd emissions would look something like this:
cronner.sleepytime.time:10005.834649|ms
cronner.sleepytime.exit_code:0|g
It emits a timing metric for how long it took for the command to run, as well as the command's exit code.
If you want to run /bin/sleep 5
as sleepytime2
and emit a DogStatsD for when the job starts and finishes:
$ cronner -e -l sleepytime2 -- /bin/sleep 5
The UDP datagrams emitted would then look like this:
_e{35,12}:Cron sleepytime2 starting on rinzler|job starting|k:ab31f2f6-498e-468a-b572-ab990065e8d3|s:cronner|t:info
cronner.sleepytime2.time:5005.649979|ms
cronner.sleepytime2.exit_code:0|g
_e{55,22}:Cron sleepytime2 succeeded in 5.00565 seconds on rinzler|exit code: 0\\noutput:(none)|k:ab31f2f6-498e-468a-b572-ab990065e8d3|s:cronner|t:success
To make cronner
easier to install and use, there is a
cronner Chef cookbook
available for use. Not only does it allow you to install cronner
, but it makes
it easy to replace your cron_d
resources with ones that will wrap the jobs
with cronner
.
- Tim Heckman
- Thomas Dziedzic
-
set up your workspace as per the instructions for standard Go development
-
clone the cronner repository
git clone [email protected]:theckman/cronner.git
-
make your changes to the codebase, including adding relevant test cases
-
run your tests to ensure all pass
go test -v ./... -check.vv
-
confirm that building cronner works
go build