Add an option to send logs to STDOUT

[toadjaune]

When using docker with a container orchestrator (kubernetes, in my case), most logging infrastructures will expect your containers to send their logs through STDOUT/STDERR.

In this image, we miss most interesting logs (typically, access logs) because they go directly to files. Moreover, the location where they go is a volume, which makes it way harder to symlink the logfiles to another location (such as /dev/stdout or /proc/1/fd/1).

It would be really convenient to have an environment variable that would allow all logs to be sent to STDOUT.

[toadjaune]

As a side note, if anyone has the same issue, I personally worked around it with the following supervisord service :

[program:logs]
priority=10
directory=/var/log
command=/logs-startup-script.sh
user=redmine
autostart=true
autorestart=true
stopsignal=QUIT
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=0
stderr_logfile=/dev/stdout
stderr_logfile_maxbytes=0

And the following script in /logs-startup-script.sh :

#!/bin/bash

# Ensure every log file actually exists
echo "Creating logfiles"
mkdir -p $REDMINE_LOG_DIR/nginx
mkdir -p $REDMINE_LOG_DIR/redmine
mkdir -p $REDMINE_LOG_DIR/supervisor
touch $REDMINE_LOG_DIR/nginx/access.log
touch $REDMINE_LOG_DIR/nginx/error.log
touch $REDMINE_LOG_DIR/nginx/redmine_access.log
touch $REDMINE_LOG_DIR/nginx/redmine_error.log
touch $REDMINE_LOG_DIR/redmine/cron_rake.log
touch $REDMINE_LOG_DIR/redmine/unicorn.stdout.log
touch $REDMINE_LOG_DIR/redmine/unicorn.stderr.log
touch $REDMINE_LOG_DIR/supervisor/cron.log
touch $REDMINE_LOG_DIR/supervisor/nginx.log
touch $REDMINE_LOG_DIR/supervisor/supervisord.log
touch $REDMINE_LOG_DIR/supervisor/unicorn.log

# Manually re-route all logs to stdout/stderr with a tail command
# NB : Some of these files are actually never logged to
tail -f \
  $REDMINE_LOG_DIR/nginx/access.log \
  $REDMINE_LOG_DIR/nginx/error.log \
  $REDMINE_LOG_DIR/nginx/redmine_access.log \
  $REDMINE_LOG_DIR/nginx/redmine_error.log \
  $REDMINE_LOG_DIR/redmine/cron_rake.log \
  $REDMINE_LOG_DIR/redmine/unicorn.stdout.log \
  $REDMINE_LOG_DIR/redmine/unicorn.stderr.log \
  $REDMINE_LOG_DIR/supervisor/cron.log \
  $REDMINE_LOG_DIR/supervisor/nginx.log \
  $REDMINE_LOG_DIR/supervisor/supervisord.log \
  $REDMINE_LOG_DIR/supervisor/unicorn.log

The resulting logs are a bit ugly, but it works. A proper setup probably would prefix log lines by the name of the service they are coming from, or something similar.

[jcormier]

I'd be open for a pull request.

…

On Mon, Feb 18, 2019, 10:56 AM Arnaud Venturi ***@***.*** wrote: When using docker with a container orchestrator (kubernetes, in my case), most logging infrastructures will expect your containers to send their logs through STDOUT/STDERR. In this image, we miss most interesting logs (typically, access logs) because they go directly to files. Moreover, the location where they go is a volume, which makes it way harder to symlink the logfiles to another location (such as /dev/stdout or /proc/1/fd/1). It would be really convenient to have an environment variable that would allow all logs to be sent to STDOUT. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#383>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFWomUGJNZnOchwdAgV-fjSopxGieV4Iks5vOs0wgaJpZM4bBKdA> .

[toadjaune]

I'll try to make one if I can find some time for that ;)