Re: missing Java method information in operf profile

On 08/25/2014 07:18 PM, Brian Hall wrote:
> On 08/25/2014, Maynard Johnson wrote:
>> On 08/25/2014 01:01 PM, Brian Hall wrote:
>>> I'm running the operf from AT 7.0-5 on a POWER8 system:
>>>   # /opt/at7.0/bin/operf -v
>>>   /opt/at7.0/bin/operf: oprofile 1.0.0git compiled on Jul 24 2014 13:49:08
>>> using the command (as root):
>>>   /opt/at7.0/bin/operf --events PM_RUN_CYC:10290000 --separate-thread --separate-cpu --system-wide
>>>
>>> There are 14 active JVM processes running on the system that I am trying to profile.  For some reason, the profiles only contain full JIT compiler method information for 2 or 3 of the processes (it varies across different runs).  I have checked that the agentlib option is being correctly passed to the JVMs, and that the JIT dump files for all 14 processes are available in /tmp/.oprofile/jitdump at the time that operf is run.
>>>
>>> A casual look at verbose output from operf didn't indicate any problems in accessing the JIT dump files, it just didn't try to process all of them.  I tried adding --lazy-conversion to the operf options, and that run did show an issue at the end of the verbose output:
>>>   add sym: name=Ljava/utilchild received signal 9
>>Hi, Brian,
>>This sounds similar to a problem that was reported on this list on July 30 by Andrew Haley regarding an ill-advised hard-coded timeout.  Andrew's problem has been fixed upstream now, but if you were running into that exact problem, using --lazy-conversion would have resolved it.  Looking at the code now, I see there is a second hard-coded timeout.  The one I got rid of was where the main operf thread is waiting for the oper_read_pid to finish.  There's another one, where the operf_read_pid is waiting for yet another process to finish . . .  which was started via exec to do the JIT dump conversion.  The timeout for the actual JIT dump conversion is 2 seconds. It appears you've hit that limit.
>>
>>I'm attaching a patch that should resolve this, if you could please test it.
>>
> 
> Hi Maynard,
> The patch worked perfectly, with or without --lazy-conversion.  Thanks!
> Brian
Brian, thanks for reporting the problem and testing the fix.  I pushed the patch upstream, and I'll ensure it's included in 1.0.0 GA.

-Maynard
> 
>>Thanks.
>>-Maynard
>>>
>>> Brian
> 


------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/