Re: truss shows polls. What does that mean?

  • From: Mladen Gogala <mladen@xxxxxxxxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Thu, 19 Feb 2004 15:15:48 -0500

Wolfgang, "poll" system call polls an open file for an event like 
"there is something to read" or "write will not block" and alike.  Polling
is  done by spinning, so that is why you are burning CPU, without any 
visible results. You are polling, because you are waiting for something 
to let you proceed with the I/O request. You can find out what is it that 
you are polling, from /proc/<the process PID>/fd (ls -l will reveal the
file names for each descriptor). If you are waiting on the network, then you
have something in your configuration that makes oracle go and peep thru
the network hole. If it is a disk, you may be having events, normally not
visible by the naked eye, like SCSI controller resets, because of the controller
failure, it may be a bad SCSI cable (I lost a whole night of my beauty sleep
and approximately 3.14% of my hair diagnosing a SCSI cable failure and had
to develop rather cunning methodology to do that in NYC in 1998. I still
remember saying to my boss, an English guy, "master, I have a cunning plan") 

On 02/19/2004 02:51:35 PM, Wolfgang Breitling wrote:
> We are having a production process that is sitting on a fairly benign 
> insert statement for the pas 2 hours. Last night it was processing for 9 
> hours plus, racking up LIO that make the federal deficit (US or Canada 
> doesn't matter) pale in comparison. I just ran the statement in sqlplus and 
> it does its thing in 11 seconds.
> 
> I turned on a 10046 level 12 trace , but there are no (oracle) waits. it is 
> just burning cpu.
> 
> Then I did a truss on the process and this is what it shows:
> 
> [dom1 ~/admin/FS75PRD/udump] truss -c -p 28279
> ^Csyscall              seconds   calls  errors
> poll                    1.00    7836
> yield                    .01     174
>                       -------  ------   ----
> sys totals:             1.01    8010      0
> usr time:               8.83
> elapsed:               13.62
> 
> 
> [dom1 ~/admin/FS75PRD/udump] uname -a
> SunOS dom1 5.8 Generic_108528-23 sun4u sparc SUNW,Sun-Fire
> 
> Oracle 8.1.7.2
> 
> Any explanation / suggestion?
> 
> Wolfgang Breitling
> Oracle7, 8, 8i, 9i OCP DBA
> Centrex Consulting Corporation
> http://www.centrexcc.com 
> 
> 
> ----------------------------------------------------------------
> Please see the official ORACLE-L FAQ: http://www.orafaq.com
> ----------------------------------------------------------------
> To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
> put 'unsubscribe' in the subject line.
> --
> Archives are at //www.freelists.org/archives/oracle-l/
> FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
> -----------------------------------------------------------------
> 
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Other related posts: