[
https://issues.apache.org/jira/browse/LOG4J2-1074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14614440#comment-14614440
]
Remko Popma commented on LOG4J2-1074:
-------------------------------------
The overhead of System.nanoTime() apparently varies greatly across platforms.
Aleksey Shipilёv (the author of JMH) in his [blog about
nanotime|http://shipilev.net/blog/2014/nanotrusting-nanotime/] mentions the
best we can hope for on modern hardware is around 30 nanoseconds, but there is
a scalability bottleneck if many threads call System.nanoTime() concurrently.
Some people [report|http://stackoverflow.com/a/19052481/1446916] that
System.nanoTime() may take microseconds to execute on some platforms (but no
further reference so we cannot verify this).
See also Stas's blog about [platform specific
differences|http://stas-blogspot.blogspot.jp/2012/02/what-is-behind-systemnanotime.html]
with System.nanoTime().
Gary raised a good point: the log event will carry an additional long, which
will change the serialized form of the LogEvent. And we need to add a method
getNanoTime() to the LogEvent interface.
Are there any concerns regarding this?
> Add a logformat token for nanotime
> ----------------------------------
>
> Key: LOG4J2-1074
> URL: https://issues.apache.org/jira/browse/LOG4J2-1074
> Project: Log4j 2
> Issue Type: New Feature
> Reporter: Will Hains
>
> Add a token in the log format specifier to output the current value of
> {{System.nanotime()}}.
> We are currently adding the nanotime value to our log messages manually,
> which we find very useful for troubleshooting unexpected latency, and as a
> quick-and-dirty data source to feed into performance analysis tools.
> It would be very nice if Log4j could do this for us. It would also be nice to
> have this in the log format specified, so we can turn it on/off at runtime by
> editing the Log4j config file.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
