Re: Logback performance improvements

The append method isn’t synchronized but the writeBytes method acquires a lock. His code is actually a lot simpler than ours in that it just uses a BufferedOutputStream and he only obtains the lock when he is writing to it. Ralph > On Feb 6, 2017, at 5:23 PM, Matt Sicker

Re: Logback performance improvements

I used a ThreadLocal byte buffer and wrote to the file channel and if anything, it performed slightly worse. This is probably because I had to write after ever event, not when the buffer was full, otherwise the ordering of events in the output would get messed up. I decided to throw the

Re: Logback performance improvements

True, but I would still like to also see what difference it makes using the FileChannel instead of the OutputStream. Ralph > On Feb 7, 2017, at 9:45 AM, Remko Popma wrote: > > That is all doable but it may be a good idea to test if that is really where > the bottleneck

Jenkins build is back to normal : Log4j 2.x #2653

See - To unsubscribe, e-mail: log4j-dev-unsubscr...@logging.apache.org For additional commands, e-mail: log4j-dev-h...@logging.apache.org

Re: Logback performance improvements

We should try to override getBuffer()/drain()/flushBuffer() from OutputStreamManager in FileManager and RandomAccessFileManager (just like we do in MenoryMappedFileManager). On Tue, Feb 7, 2017 at 5:45 PM, Remko Popma wrote: > That is all doable but it may be a good idea

Re: Logback performance improvements

That is all doable but it may be a good idea to test if that is really where the bottleneck is. We could try whether we get better numbers if we remove the current synchronization (ignoring any scrambled output, just for testing purposes). On Wed, Feb 8, 2017 at 1:40 AM, Apache

[jira] [Resolved] (LOG4J2-1810) Upgrade Logback to version 1.1.10

[ https://issues.apache.org/jira/browse/LOG4J2-1810?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Ralph Goers resolved LOG4J2-1810. - Resolution: Fixed > Upgrade Logback to version 1.1.10 > - > >

[jira] [Closed] (LOG4J2-1810) Upgrade Logback to version 1.1.10

[ https://issues.apache.org/jira/browse/LOG4J2-1810?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Ralph Goers closed LOG4J2-1810. --- > Upgrade Logback to version 1.1.10 > - > > Key:

[jira] [Commented] (LOG4J2-1691) Setup build process for new Git repo logging-log4j-scala, including site

[ https://issues.apache.org/jira/browse/LOG4J2-1691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15856301#comment-15856301 ] Mikael Ståldal commented on LOG4J2-1691: I think we need to integrate the site for the modules of

Re: Logback performance improvements

In looking at FileManager and OutputStreamManager it does have a ByteBuffer but it requires synchronization. I am thinking it might make more sense to have a ThreadLocal ByteBuffer and then pass that to FileChannel.write() so that no synchronization is required. Ralph > On Feb 7, 2017, at

Re: Logback performance improvements

Can't we use the ByteBuffers introduced in the GC-free epic? I was under the impression that byte arrays being passed to appenders was created from a ByteBuffer at this point, though I haven't really taken a close look at this. On 7 February 2017 at 10:05, Apache

[jira] [Commented] (LOG4J2-1810) Upgrade Logback to version 1.1.10

[ https://issues.apache.org/jira/browse/LOG4J2-1810?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15856270#comment-15856270 ] ASF subversion and git services commented on LOG4J2-1810: - Commit

[jira] [Created] (LOG4J2-1810) Upgrade Logback to version 1.1.10

Ralph Goers created LOG4J2-1810: --- Summary: Upgrade Logback to version 1.1.10 Key: LOG4J2-1810 URL: https://issues.apache.org/jira/browse/LOG4J2-1810 Project: Log4j 2 Issue Type: Bug

Re: Logback performance improvements

A FileChannel is guaranteed to be thread safe. You can obtain a FileChannel from a FlieOutputStream, so that would seem to imply that FileOutputStream might be thread-safe, but you can’t really know that without looking at the source. The problem is that FileChannel.write() takes a ByteBuffer

[jira] [Commented] (LOG4J2-1691) Setup build process for new Git repo logging-log4j-scala, including site

[ https://issues.apache.org/jira/browse/LOG4J2-1691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15856220#comment-15856220 ] Matt Sicker commented on LOG4J2-1691: - The final step is the site integration. I'm not exactly sure

Re: Logback performance improvements

I looked at 1.2-SNAPSHOT and 1.1.10 and saw nothing special other than a lack of a synchronized keyword on the equivalent append method. Perhaps he figured out a simpler way to emulate locking? I've been working with async/non-blocking streaming APIs for long enough now that I can't even remember

[jira] [Commented] (LOG4J2-1691) Setup build process for new Git repo logging-log4j-scala, including site

[ https://issues.apache.org/jira/browse/LOG4J2-1691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15856075#comment-15856075 ] Mikael Ståldal commented on LOG4J2-1691: Are we done with this now? Can we now integrate the