[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-serviceability-dev
Subject:    Re: JEP 167
From:       Staffan Larsen <staffan.larsen () oracle ! com>
Date:       2014-05-16 10:22:13
Message-ID: 0D2E67AC-3D74-42E4-AC39-980517F43FBB () oracle ! com
[Download RAW message or body]


On 16 maj 2014, at 10:59, Mario Torre <neugens@redhat.com> wrote:

> On Fri, 2014-05-16 at 09:12 +0200, Staffan Larsen wrote:
> 
> Hello Staffan,
> 
> Thanks for the quick reply.
> 
> > JEP 167 is done and should be in the completed state. I’ll see if I
> > can get that updated. The relevant bug (with change set information)
> > is: https://bugs.openjdk.java.net/browse/JDK-8005849.
> 
> Great, yeah, updating the page seems a good idea :)

The mercurial text file has been updated, but the publishing script is broken so the \
web-page won’t be updated (I don’t expect this to be fixed until JEP 2.0 is final).

> 
> > For seeing what instrumentation has been added, take a look at
> > src/share/vm/trace/trace.xml which lists all instrumentation points.
> > To enable trace output in OpenJDK, run with q (although running Java
> > Flight Recorder [1] from Java Mission Control [2] provides a much
> > better user experience).
> 
> I see, thanks. It would be nice to also have some documentation how to
> get those traces, for instance, assuming I want to turn on/off a
> specific trace, and I want to attach my own agent, how would I do this?
> 
> I guess I can figure it out somehow from the code, but it's not obvious
> if those traces can be used other than via the target vm printing stuff.
> 
> I understand Flight Recorded is able to access this information, but
> what if I want to write my own agent? Is this accessed via JVMTI, or
> JMX, or is there another way (perhaps less expensive)?

There is currently no way to access the information outside of Flight Recorder (or \
-XX:+EnableTracing in OpenJDK).

> 
> A final question, do you know the impact when instrumentation is turned
> on (I can do my one testing, but maybe there are some official numbers
> around already :)?

For Flight Recorder we see a performance overhead of around 2-3%.

> 
> > Regarding the broken links, it looks like the
> > http://opengrok.neojava.org is no longer what it used to be. In
> > general, page you refer to has not been updated for several years...
> 
> Hmm, I believe this should be updated as well, it's a pretty important
> page and I think the first place anyone interested in serviceability
> would look at, what do you think?

That would be good. I don’t know who can edit the page.

/Staffan

> 
> Cheers,
> Mario
> 


[prev in list] [next in list] [prev in thread] [next in thread]