Re: DBMS_STATS runs for hours and creates no stats
From: Dan <daniel.ostertag_at_visaer.com>
Date: Fri, 30 Jul 2010 13:42:48 -0700 (PDT)
Message-ID: <aa2031dd-11fe-4047-8e49-a5f81ac4f34c_at_x21g2000yqa.googlegroups.com>
On Jul 30, 1:44 pm, joel garry <joel-ga..._at_home.com> wrote:
> On Jul 30, 7:19 am, Dan <daniel.oster..._at_visaer.com> wrote:
>
>
>
>
>
> > I'm running into an occasional issue with my stats gathering in Oracle
> > 10.2. I run the following commands:
>
> > DEFINE SCHEMA = "IBXREF1"
> > EXEC DBMS_STATS.GATHER_TABLE_STATS(ownname=> '&SCHEMA', tabname=>
> > 'HOUSEHOLD_DATA', estimate_percent=>DBMS_STATS.AUTO_SAMPLE_SIZE);
>
> > This is a large table, 135G, and the stats gathering runs for 5-6
> > hours on our server. It completes as normal with no errors, but when
> > you look at the table afterward there are no stats for it. The
> > USER_TABLE.LAST_ANALYZED is blank.
>
> > This command is from a script that I run every month, and most months
> > are fine, so it's not a syntax thing. This just seems to happen once
> > every 2 or 3 months on one of my tables.
>
> > Any ideas why this is happening and what I can do to fix it?
>
> > Dan
>
> I haven't looked recently, but I have a vague memory that there is a
> window in which things can be run (at least when you do this through
> EM), so I'm guessing sometimes it takes too long to fit in the window
> and dies. What is in the script and how is it run? How is it
> logged? Are you sure the default stats gathering is disabled?http://download.oracle.com/docs/cd/B19306_01/server.102/b14231/tasks.htm
>
> There are some number of bugs in the MOS bug database if you search
> for auto_sample_size, if it isn't a resource problem I speculate there
> is some instability you run into under odd conditions.
>
> jg
> --
> _at_home.com is bogus.http://www.sfgate.com/cgi-bin/blogs/scavenger/detail?entry_id=68992- Hide quoted text -
>
> - Show quoted text -
Date: Fri, 30 Jul 2010 13:42:48 -0700 (PDT)
Message-ID: <aa2031dd-11fe-4047-8e49-a5f81ac4f34c_at_x21g2000yqa.googlegroups.com>
On Jul 30, 1:44 pm, joel garry <joel-ga..._at_home.com> wrote:
> On Jul 30, 7:19 am, Dan <daniel.oster..._at_visaer.com> wrote:
>
>
>
>
>
> > I'm running into an occasional issue with my stats gathering in Oracle
> > 10.2. I run the following commands:
>
> > DEFINE SCHEMA = "IBXREF1"
> > EXEC DBMS_STATS.GATHER_TABLE_STATS(ownname=> '&SCHEMA', tabname=>
> > 'HOUSEHOLD_DATA', estimate_percent=>DBMS_STATS.AUTO_SAMPLE_SIZE);
>
> > This is a large table, 135G, and the stats gathering runs for 5-6
> > hours on our server. It completes as normal with no errors, but when
> > you look at the table afterward there are no stats for it. The
> > USER_TABLE.LAST_ANALYZED is blank.
>
> > This command is from a script that I run every month, and most months
> > are fine, so it's not a syntax thing. This just seems to happen once
> > every 2 or 3 months on one of my tables.
>
> > Any ideas why this is happening and what I can do to fix it?
>
> > Dan
>
> I haven't looked recently, but I have a vague memory that there is a
> window in which things can be run (at least when you do this through
> EM), so I'm guessing sometimes it takes too long to fit in the window
> and dies. What is in the script and how is it run? How is it
> logged? Are you sure the default stats gathering is disabled?http://download.oracle.com/docs/cd/B19306_01/server.102/b14231/tasks.htm
>
> There are some number of bugs in the MOS bug database if you search
> for auto_sample_size, if it isn't a resource problem I speculate there
> is some instability you run into under odd conditions.
>
> jg
> --
> _at_home.com is bogus.http://www.sfgate.com/cgi-bin/blogs/scavenger/detail?entry_id=68992- Hide quoted text -
>
> - Show quoted text -
Perhaps it runs too long. This takes 6 1/2 hours to run, maybe there is a bug in the process when it runs too long. Our server has been running slow this month so this was the longest time yet. I'm running this manually, not thru EM. We have no scheduled stats done for these tables so I don't think that would be the cause - but a valid suggestion to check on.
Regarding the ESTIMATE option, I've run this command probably 100 times exactly the same way with no problems, so I doubt syntax would be an issue. All valid suggestions though. I'm stumped. In the meantime I just have to run it again and hope...
Thanks,
Dan
Received on Fri Jul 30 2010 - 15:42:48 CDT