[dtrace-discuss] RES: Process in LCK / SLP (Please)

[Kleyson Rios]

Tanks everybody.

I will try to do what you guys said.
When I get new issues I post again.

Regards, 
Kleyson Rios.

-----Mensagem original-----
De: Jarod Jenson [mailto:jarod at aeysis.com] 
Enviada em: quarta-feira, 16 de abril de 2008 16:56
Para: Peter B. Kessler
Cc: Kleyson Rios; dtrace-discuss at opensolaris.org
Assunto: Re: [dtrace-discuss] Process in LCK / SLP (Please)



Peter B. Kessler's email at 4/16/08 2:01 PM, said:
> Michael Schuster wrote:
>> Kleyson Rios wrote:
>>> Hi,
>>>
>>> I really need help.
>>>
>>> How can i identify why my process are 100% in LCK and SLP ?
>>>
>>>    PID USERNAME USR SYS TRP TFL DFL LCK SLP LAT VCX ICX SCL SIG
PROCESS/LWPID
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0  30   0  18   0 java/28
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   2   0   1   0 java/94
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   3   0   1   0 java/93
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   1   0   0   0 java/100
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   1   0   1   0 java/18
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/49
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/47
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/48
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   0   0   0   0 java/32
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   0   0   0   0 java/31
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   0   0   0   0 java/30
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   0   0   0   0 java/29
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 0.0 100 0.0   0   0   0   0 java/27
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/26
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/25
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/24
>>>   6769 root     0.0 0.0 0.0 0.0 0.0 100 0.0 0.0   0   0   0   0 java/23
>> from the prstat man-page:
>>
>>       LCK
>>
>>           The percentage of time the process has spent waiting for
>>           user locks.
>>
>>       SLP
>>
>>           The percentage of time the process has spent sleeping.
>>
>> it looks like your processes have deadlocked themselves - not much one
can 
>> do without knowledge about what they're actually doing and access to the 
>> source code. I'd talk to the people developing this app.
>>
>> Michael
> 
> If it is a deadlock, and if it is a recent Java virtual machine, then
> you should be able to connect to it with jconsole (from the same directory
> where you get your java command), go to the "Threads" tab and use the
> "Detect Deadlock" button to get a thread stack dump showing where you
> have a cycle among your Java-level locks.
> 
> 			... peter
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
> 

*Highly* unlikely it is a deadlock. Insanely, Solaris charges time in 
condition wait ([pthread_]cond_wait or lwp_cond_wait) as lock time. This 
has never made sense to me, and I have yet to wield a big enough stick 
to get it fixed.

Take either a Java thread dump or a pstack(1) dump. If a Java dump, you 
will be in Object.wait (or similar) or sleep(). If a pstack(1), you will 
be in some form of cond_wait.

If it really is a deadlock (and I'll buy you a beer if it is because I 
am sure it isn't), Peter's suggestion is the way to go.

As for sleep, the threads are tired. Long day of garbage collection I 
suppose. Seriously though, they have no (or very little) work to do 
since the top thread is 100% sleep but made 18 system calls (rounding 
error).

Thanks,

Jarod