accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Elser (JIRA)" <>
Subject [jira] [Commented] (ACCUMULO-2372) Use POSIX logger for STDERR and STDOUT
Date Tue, 23 Jun 2015 21:43:42 GMT


Josh Elser commented on ACCUMULO-2372:

After looking into what these changes would entail, along with some explanation from Christopher,
I'm not a big fan of making this change as the new default.

As I understand it, the changes that Christopher is proposing would result in all "console
output" (stdout and stderr) being piped to the syslog and also copied to a single file. So,
we would have a single file that replaces the .out and .err files. I am hesitant to force
this change onto users as .out/.err files with a log file is extremely common across the Hadoop
"family". Making this change would be different compared to every other project.

While I agree that moving to "standard" process management tools provided at the operating
system level is a good decision, I don't think the community is ready to embrace this. If
we made this change. If there are no drawbacks to doing this, I would argue that this is something
Accumulo could help push "top-down" rather than be a "leaf" that does something different.

If we can get consensus from the community at large, I think we could consider this proposal
as on-by-default, but I definitely don't see this as ready for a 1.8.0 release. I have no
problem with syslog via {{logger}} support being added, controlled by an option that users
can enable for themselves.

> Use POSIX logger for STDERR and STDOUT
> --------------------------------------
>                 Key: ACCUMULO-2372
>                 URL:
>             Project: Accumulo
>          Issue Type: Improvement
>            Reporter: Christopher Tubbs
>              Labels: newbie
>             Fix For: 1.8.0
> Instead of piping STDOUT and STDERR to *.out and *.err files in our launch scripts, we
should pipe to the POSIX cli logger (/usr/bin/logger) to log to the system logging facility
at the desired log level.
> Incorporating the system logging facilities is useful for system administrators for troubleshooting
and/or auditing. More integration with external logging facilities also means that we reduce
our responsibility to maintain/manage the resulting logs to avoid problems (like filling up
the local file system with logs), deferring that to utilities dedicated to that responsibility.
> The command-line options used for /usr/bin/logger should be enough to associate the logs
with the particular process.
> (NOTE: Log4j can also be configured to use the system log facility, but that should be
done through log4j configuration, with the appropriate log appender.)

This message was sent by Atlassian JIRA

View raw message