[
https://issues.apache.org/jira/browse/MESOS-3465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14944383#comment-14944383
]
Benjamin Mahler commented on MESOS-3465:
----------------------------------------
I'm curious if we also want to make it clear that the program is exiting in the
logs, e.g.
{noformat}
I1005 22:23:54.756332 37901 master.cpp:999] Lost leadership.
EXIT with status 1: Lost leadership
{noformat}
Is it critical that all of our output begin with the timestamp? I believe some
logging analysis tools rely on consistent log line formatting, hopefully they
ignore lines that don't fit. An alternative:
{noformat}
I1005 22:23:54.756332 37901 master.cpp:999] EXIT with status 1: Lost leadership.
{noformat}
> EXIT() should include timestamp
> -------------------------------
>
> Key: MESOS-3465
> URL: https://issues.apache.org/jira/browse/MESOS-3465
> Project: Mesos
> Issue Type: Improvement
> Reporter: Neil Conway
> Priority: Minor
> Labels: mesosphere, newbie
>
> EXIT() prints a log message but doesn't include a timestamp, and is otherwise
> inconsistent with the format of other log messages. EXIT() might be called
> during normal operational circumstances (e.g., during a network partition),
> so a timestamp would be useful to aid analysis.
> We could use LOG(FATAL), but apparently this dumps a backtrace, which is
> probably not warranted.
> How about EXIT() does LOG(ERROR), then calls exit() as it does now?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
