Add a command line flag to specify the log destination

[guoshimin]

Currently prometheus logs to stderr. I could use shell redirection to make it log to a file, but that doesn't play well with the official prometheus docker image: I will have to change the entry point from prometheus to shell. It's better to make it a command line flag.

[fabxc]

Not supporting this was a conscious decision. There are just to many ways to approach this problem so the builtin solutions will inevitably blow up in complexity.

Maybe the other Docker users can chime in give advice how they are solving the problem.

@discordianfish @sdurrheimer

[discordianfish]

@guoshimin Why do you need to redirect it to a file? Docker keeps the logs for you right now and you can access them via docker logs my-prometheus-container.

If you need to, you can also run the docker container with: docker run -a prom/prometheus ... > log.txt 2>&1 to redirect the output.

[guoshimin]

I want to be able to rotate logs, so I don't want docker to keep the logs
for me

• In theory I could logrotate /var/lib/docker/containers# ls
  /var/lib/docker/containers//-json.log, but I don't want to apply the same
  log rotation policy to all container logs.

Yes there are other ways to approach this problem. Having the ability to
specify the log destination will make my run script slightly simpler. For
now I'll just continue to use shell as the entrypoint. I like to keep
everything inside the container or in volumes.

On Mon, Aug 17, 2015 at 11:00 AM, discordianfish notifications@github.com
wrote:

@guoshimin https://github.com/guoshimin Why do you need to redirect it
to a file? Docker keeps the logs for you right now and you can access them
via docker logs my-prometheus-container.

If you need to, you can also run the docker container with: docker run -a
prom/prometheus ... > log.txt 2>&1 to redirect the output.

—
Reply to this email directly or view it on GitHub
#997 (comment)
.

[discordianfish]

If you're using a recent version of Docker, you can use logging drivers to manage the logs: https://docs.docker.com/reference/logging/overview/

[swsnider]

We use cronolog for this, and it works pretty well.

[guoshimin]

Thanks for the suggestions.

@swsnider Do you guys run cronolog inside the same container as the main app? I'm trying to run one process per container wherever I can, what with shell not propagating signal to subprocesses.

[swsnider]

sorry, just realized I was mixed up -- we run our docker containers via upstart, and we use cronolog as part of the upstart script, in combination with docker run.

[guoshimin]

Gotcha. Thank you all for your help. I'll close the issue.

[SpoorthyB]

@guoshimin: I’m interested to know how you are handling this as of today as I have the same scenario as yours. Are you still using shell redirection or do you have an alternative ?

[guoshimin]

@SpoorthyB I no longer need to do this. I simply let docker capture the stdout/stderr and from there I also set up logstash to ship the logs to elasticsearch. If I needed to do it today I would run prometheus under supervisord: http://supervisord.org/configuration.html#program-x-section-example