Re: Subject: Batch load freezes every couple of minutes

  • From: "Ranko Mosic" <mosicr@xxxxxxxxxx>
  • To: "sol beach" <sol.beach@xxxxxxxxx>
  • Date: Tue, 19 Apr 2005 18:49:59 -0400

It's not full, only 40% filled on average.

----- Original Message -----
From: "sol beach" <sol.beach@xxxxxxxxx>
To: "Ranko Mosic" <mosicr@xxxxxxxxxx>
Cc: <oracle-l@xxxxxxxxxxxxx>; <ryan_gaffuri@xxxxxxxxxxx>
Sent: Tuesday, April 19, 2005 2:21 PM
Subject: Re: Subject: Batch load freezes every couple of minutes


If the datablock is FULL, additional ITLs will NOT be available;
regardless of the value of MAXTRANS

On 4/18/05, Ranko Mosic <mosicr@xxxxxxxxxx> wrote:
> Isn't maxtrans of 255 ( default )sufficient ?
>
> ----- Original Message -----
> From: "sol beach" <sol.beach@xxxxxxxxx>
> To: <mosicr@xxxxxxxxxx>
> Cc: <oracle-l@xxxxxxxxxxxxx>; <ryan_gaffuri@xxxxxxxxxxx>
> Sent: Sunday, April 17, 2005 9:18 PM
> Subject: Re: Subject: Batch load freezes every couple of minutes
>
> >Instance runs 20 parallel load processes
>
> You might be encountering enqueue waits.
> Typically INITRANS=1 for tables & 2 for indexes (YMMV).
> Unless & until the session holding the Interested Transaction Lock
> releases it, other sessions will quietly get into line waiting for it
> to be released. This is NOT a deadlock & won't be reported as one.
>
> HTH & HAND!
>
> On 4/17/05, ryan_gaffuri@xxxxxxxxxxx <ryan_gaffuri@xxxxxxxxxxx> wrote:
> > Please provide your definition of 'hang'? To see if your processes are
> still running look in v$sess_io, if the values there are incrementing,
then
> processing is still going on.
> > you need to find out what the processes are doing at the point that they
> are freezing? Best way is to get a 10046 trace of one of the processes
> dumped to a file. Another alternative is to use statspack of before the
> 'hang' and then another snapshot afterward and look at what sql is
> processing.
> >
> > you mentioned that the times that you have problems are when you have
lock
> waits. you could have 1 process doing something for a long time and
aquiring
> a lock on a row. the rest of the processes require that row and have to
wait
> for a commit or rollback.
> >
> > the v$ tables(other than a quick look at v$sess_io) probably won't be
> helpful here. you need to know exactly what is going on when this occurs.
> >
> > --
> > //www.freelists.org/webpage/oracle-l
> >
>
>

--
//www.freelists.org/webpage/oracle-l

Other related posts: