Re: RandomAccessFile and the native futex

Avinash,

I had a problem with Sonic MQ JMS server running Java 1.5 on Redhat ES  
4 on a four processor IBM x460 that appeared to result from  
inefficiencies in the Linux kernel's thread code. It may be completely  
unrelated. To better understand your situation could you please add  
some background information?

What does the application do?
Do you see a response time or a throughput that is less than you had  
expected?
Can you do a load test on Windows, OS X or Solaris x86 and compare  
throughput with NY
Why do your observations imply a problem

Is it running on a physical host or a VM?
Is it running on an alternate machine
How many CPUs/cores?
What can you say about the app under heavy load?
Do you expect the app to be IO bound or CPU bound?
What does sar/vmstatshow you when the application is under significant  
load?
What do threadumps show when the application is under significant load?

Peter Booth

On Dec 21, 2007, at 12:56 AM, Avinash Lakshman wrote:

> The behavior is the same with the IBM JVM. I was running Java 6. How  
> do I get around this problem? How can I make sure I am using the  
> correct glibc libraries? I can't seem to find any information as to  
> what the recommended version is for JVM to run.
>
> Thanks
> A
>
>> Date: Tue, 18 Dec 2007 19:52:21 -0500
>> From: [email protected]
>> Subject: Re: [ADVANCED-JAVA] RandomAccessFile and the native futex
>> To: [email protected]
>>
>> Can you try your benchmark with the IBM JVM?
>> On Dec 18, 2007, at 11:41 AM, Avinash Lakshman wrote:
>>
>>> It looks like this may have something to do with incompatibility
>>> between NPTL threads and Java threads on Fedora Core 4. Anyone run
>>> into this problem before? I am using JDK6 with FC4
>>> 2.6.12-1.1398_FC4smp.
>>>
>>> A
>>>
>>>> Date: Mon, 17 Dec 2007 19:03:42 -0600
>>>> From: [email protected]
>>>> Subject: [ADVANCED-JAVA] RandomAccessFile and the native futex
>>>> To: [email protected]
>>>>
>>>> Hi All
>>>>
>>>> I see 52 ms being spent in calls to readFully() method of the
>>>> RandomAccessFile. I ran strace against my process and figured that
>>>> these weren't reads going to disk. All reads were being served from
>>>> the buffer cache (verified by running iostat). It also looks like
>>>> the 52 ms is a futex timeout but strace doesn't report which system
>>>> call causes this. I was wondering if anyone has seen this problem
>>>> before and how do I go about addressing this? Do I have access to
>>>> the source code of the native library used by RandomAccessFile?
>>>> Here is the output from strace (the offending line is in bold):
>>>>
>>>>
>>>> 29673 16:36:47.747658
>>>> futex(0x2aac3f62a674, FUTEX_WAIT, 1, {0, 49972000} <unfinished
>>>> ...>
>>>> 29716 16:36:47.784379 <... futex
>>>> resumed> ) = -1 ETIMEDOUT (Connection timed out)
>>>> <0.044427>
>>>> 29716 16:36:47.784469
>>>> futex(0x2aac3f66e328, FUTEX_WAKE, 1) = 0
>>>> <0.000012>
>>>> 29716 16:36:47.784548
>>>> clock_gettime(CLOCK_REALTIME, {1197938207, 784565000}) = 0
>>>> <0.000012>
>>>> 29716 16:36:47.784610
>>>> futex(0x2aac3f670314, FUTEX_WAIT, 1, {0, 19970000} <unfinished
>>>> ...>
>>>> 29673 16:36:47.799354 <... futex
>>>> resumed> ) = -1 ETIMEDOUT (Connection timed out)
>>>> <0.051676>
>>>> 29673 16:36:47.799392
>>>> futex(0x2aac3f2a7228, FUTEX_WAKE, 1) = 0
>>>> <0.000011>
>>>> 29673 16:36:47.799434
>>>> clock_gettime(CLOCK_MONOTONIC, {6497274, 522227024}) = 0
>>>> <0.000007>
>>>> 29673 16:36:47.799481
>>>> clock_gettime(CLOCK_MONOTONIC, {6497274, 522275024}) = 0
>>>> <0.000007>
>>>>
>>>> Thanks
>>>> Avinash
>>>>
>>>> _________________________________________________________________
>>>> i’m is proud to present Cause Effect, a series about real people
>>>> making a difference.
>>>> http://im.live.com/Messenger/IM/MTV/?source=text_Cause_Effect
>>>> ===================================
>>>> This list is hosted by DevelopMentor®  http://www.develop.com
>>>>
>>>> View archives and manage your subscription(s) at http://discuss.develop.com
>>>
>>> _________________________________________________________________
>>> Share life as it happens with the new Windows Live.
>>> http://www.windowslive.com/share.html?ocid=TXT_TAGHM_Wave2_sharelife_122007
>>> ===================================
>>> This list is hosted by DevelopMentor®  http://www.develop.com
>>>
>>> View archives and manage your subscription(s) at http://discuss.develop.com
>>>
>>
>> ===================================
>> This list is hosted by DevelopMentor®  http://www.develop.com
>>
>> View archives and manage your subscription(s) at http://discuss.develop.com
>
> _________________________________________________________________
> The best games are on Xbox 360.  Click here for a special offer on  
> an Xbox 360 Console.
> http://www.xbox.com/en-US/hardware/wheretobuy/
> ===================================
> This list is hosted by DevelopMentor®  http://www.develop.com
>
> View archives and manage your subscription(s) at http://discuss.develop.com
>

===================================
This list is hosted by DevelopMentor®  http://www.develop.com

View archives and manage your subscription(s) at http://discuss.develop.com