You may want to do something like this:
· Anything at level INFO or above is intended to be seen by a normal user, and the log message should be written to reflect that. You may want to raise the threshold to WARN depending on your users. The output for normal users go to standard error (since this is messages, not program output which can be piped further on to other programs).
· Two log files are kept in parallel – one for anything at level INFO or above, and one for level DEBUG or above. The first one is used by the initial technical worker called in for triaging the situation, and the second one for the developer responsible for investigating. In an emergency situation the last one may even be generated for TRACE and above.
So to answer your question I would suggest you log at INFO or WARN and above depending on what the users need to see, and do so to standard error.
/Thorbjørn
From: logback-user-bounces@qos.ch [mailto:logback-user-bounces@qos.ch] On Behalf Of Steve Ramage
Sent: 21. oktober 2012 00:30
To: logback-user@qos.ch
Subject: [logback-user] Logging to stderr and stdout
I had a question about the conventions of logging. We have an application scientific in nature, and basically it runs some process for some amount of time, and then finishes with it's output. All output in the program is driven by logback, and we generally hide the complexity of logback from our users, and support very limited configuration as needed.
One issue is where error and warning messages should go for users.
One argument is that they should go to stderr. stdout should have INFO, DEBUG, and TRACE. There doesn't seem to be an easy way to do this in logback.
Another argument is that they should go to stderr. stdout should also have them, but also INFO, DEBUG, and TRACE. If the user is running directly then both get spammed together.
A final option is they should all go to stdout, as these are all log messages, and there is no need to split them
Now this isn't so much a religious argument, we are just more wondering what the conventions were, as we would like to follow them as much possible.
Thanks,
Steve Ramage