no CPU usage recorded in Java Mission Control Flight Recorder

Question

Today I had a look at Mission Control introduced in Java 7 update 45 and tested a bit with Flight Recorder. Everything seems to be working as expected, I get a lot of data about threads, memory and whatever from the application I tested. Only one thing seems strange to me: In the category "General" there's no CPU usage shown, underneath the icon for the CPU usage there's only the sentence "No Events in Recording". When I don't use Flight Recording but the JMX Console, CPU usage is shown in the same application. During creation of the Flight Recording I left most of the settings as their default values, but made sure that on the page "Event Details for Profiling" "Operating System/Processor/CPU Load" is enabled and has a configured Period of 1000 ms, which should be enough to get some data using a test period of some minutes.

Is there anything that I'm missing? Maybe a restriction due to performance reasons, something else I need to activate?

Thanks for your suggestions!

Answer 1

sorry for a bit of a late answer, but I think i know what might be your issue:

Are you running a localized version of Windows by any chance?

I managed to introduce a bug where I hardcoded a PDH Query path to use the "Process" PDH object string, like: "\Process(%s#%d)\%s". Of course, the "Process" string is not guaranteed on localized versions of Windows, it should have been resolved via the registry index instead.

I have fixed this in 9 and 8, i think there is backport in progress for 7 line as well.

Please see this bug: https://bugs.openjdk.java.net/browse/JDK-8043323

Thanks Markus

Answer 2

CPU load is enabled by default, in both templates that are shipped with the JVM (1000 ms).

Do you get the same problem if you don't modify settings in the advanced part of the wizard? If you just click Finish on a fresh template from the first page of the recording wizard?

Answer 3

Thanks to the help of Kire Haglin I knew what to search for and found the following bugs:

http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8019921

What I don't understand is that regarding the backport-Bugs the problem should have been fixed in u45:

http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8021038

But it still doesn't work for me and my applications like Eclipse, DbVisuliazer, Tomcat within Eclipse etc. Looks like I'll have a look if I'm able to reopen the bug.

My output of java -version:

Java HotSpot(TM) Client VM (build 24.45-b08, mixed mode, sharing)

suggests that "b53", as stated in the main bug, may be the next releases of JRE/JDK were the problem got fixed. One backport bug is targeted as JRE 7u60, too.