I tried doing a trace and the results are a bit confusing. For the test I
put my master and snapshot on the same database (Scott/tiger user. I created
a snapshot emp_test which was a subset of emp). Unfortunately this is a
7.3.4 database so it is a bit out of date. I might try 8.1 a little later.
These are the steps I followed:
Insert one row into emp. Update one row of emp. (check mlog$_emp. The
entries exist)
turn on trace (level 12)
manually refresh the snapshot (execute
dbms_snapshot.refresh(list=>'emp_test')
turn off trace.
I wanted to see if the inserts into SNAP$_EMP_TEST and the deletes from
MLOG$_EMP were in the same transaction. These are some excerpts from the
trace file
=====================
PARSING IN CURSOR #7 len=144 dep=2 uid=0 oct=6 lid=0 tim=1659386
hv=2425594672 ad='80289d38'
update "SCOTT"."MLOG$_EMP" set snaptime$$ = :1 where snaptime$$ is
NULL or snaptime$$ > to_date('2100-01-01:00:00:00','YYYY-MM-DD:HH24:MI:SS')
END OF STMT
PARSE #7:c=0,e=0,p=0,cr=1,cu=0,mis=1,r=0,dep=2,og=0,tim=1659386
BINDS #7:
bind 0: dty=12 mxl=07(07) mal=00 scl=00 pre=00 oacflg=08
bfp=0088109c bln=07 avl=07 flg=0d
value="9/21/2001 11:41:12"
EXEC #7:c=0,e=0,p=0,cr=4,cu=4,mis=0,r=2,dep=2,og=4,tim=1659386
=====================
PARSING IN CURSOR #16 len=104 dep=1 uid=11 oct=2 lid=11 tim=1659401
hv=238238631 ad='8026f240'
insert into "SCOTT"."SNAP$_EMP_TEST" ( "EMPNO", "ENAME", "DEPTNO",
"M_ROW$$") values ( :1, :2, :3, :4)
END OF STMT
PARSE #16:c=0,e=0,p=0,cr=0,cu=0,mis=1,r=0,dep=1,og=0,tim=1659401
EXEC #16:c=0,e=1,p=0,cr=1,cu=3,mis=0,r=1,dep=1,og=4,tim=1659404
=====================
PARSING IN CURSOR #5 len=67 dep=1 uid=0 oct=6 lid=0 tim=1659404
hv=1979925439 ad='8026bf34'
update sys.snap$ set snaptime = :1 where vname = :2 and sowner = :3
END OF STMT
PARSE #5:c=0,e=0,p=0,cr=0,cu=0,mis=1,r=0,dep=1,og=0,tim=1659404
BINDS #5:
bind 0: dty=12 mxl=07(07) mal=00 scl=00 pre=00 oacflg=08
bfp=008991a4 bln=07 avl=07 flg=0d
value="9/21/2001 11:41:12"
bind 1: dty=1 mxl=32(08) mal=00 scl=00 pre=00 oacflg=08
bfp=0089918c bln=08 avl=08 flg=0d
value="EMP_TEST"
bind 2: dty=1 mxl=32(05) mal=00 scl=00 pre=00 oacflg=08
bfp=008997ec bln=05 avl=05 flg=0d
value="SCOTT"
EXEC #5:c=0,e=0,p=0,cr=1,cu=1,mis=0,r=1,dep=1,og=4,tim=1659404
XCTEND rlbk=0, rd_only=0
STAT #5 id=1 cnt=1 pid=0 pos=0 obj=120 op='INDEX UNIQUE SCAN '
=====================
PARSING IN CURSOR #5 len=82 dep=1 uid=11 oct=47 lid=11 tim=1659405
hv=1039323329 ad='8026ae9c'
begin sys.dbms_snapshot.wrap_up( :mowner, :master, :snapshot,
:snaptime); end;
END OF STMT
=====================
PARSING IN CURSOR #13 len=54 dep=2 uid=0 oct=7 lid=0 tim=1659414
hv=3795289681 ad='8025da64'
delete from "SCOTT"."MLOG$_EMP" where snaptime$$ <= :1
END OF STMT
PARSE #13:c=0,e=0,p=0,cr=0,cu=0,mis=1,r=0,dep=2,og=0,tim=1659414
BINDS #13:
bind 0: dty=12 mxl=07(07) mal=00 scl=00 pre=00 oacflg=08
bfp=008b127c bln=07 avl=07 flg=0d
value="9/21/2001 11:41:12"
EXEC
#13:c=0,e=0,p=0,cr=4,cu=4,mis=0,r=2,dep=2,og=4,tim=1659414==================
===
PARSING IN CURSOR #13 len=62 dep=2 uid=0 oct=6 lid=0 tim=1659414
hv=3644628630 ad='8025bfec'
update mlog$ set oldest = :1 where master = :3 and mowner = :2
END OF STMT
PARSE #13:c=0,e=0,p=0,cr=0,cu=0,mis=1,r=0,dep=2,og=0,tim=1659414
=====================
BINDS #13:
bind 0: dty=12 mxl=07(07) mal=00 scl=00 pre=00 oacflg=08
bfp=008b129c bln=07 avl=07 flg=0d
value="9/21/2001 11:41:12"
bind 1: dty=1 mxl=32(03) mal=00 scl=00 pre=00 oacflg=08
bfp=008adc34 bln=03 avl=03 flg=0d
value="EMP"
bind 2: dty=1 mxl=32(05) mal=00 scl=00 pre=00 oacflg=08
bfp=008b127c bln=05 avl=05 flg=0d
value="SCOTT"
EXEC #13:c=1,e=1,p=0,cr=8,cu=1,mis=0,r=1,dep=2,og=4,tim=1659415
STAT #13 id=1 cnt=1 pid=0 pos=0 obj=116 op='TABLE ACCESS CLUSTER
MLOG$ '
STAT #13 id=2 cnt=1 pid=1 pos=1 obj=115 op='INDEX UNIQUE SCAN '
STAT #10 id=1 cnt=1 pid=0 pos=0 obj=117 op='TABLE ACCESS CLUSTER
SLOG$ '
STAT #10 id=2 cnt=1 pid=1 pos=1 obj=115 op='INDEX UNIQUE SCAN '
EXEC #5:c=7,e=10,p=1,cr=89,cu=16,mis=0,r=1,dep=1,og=4,tim=1659415
XCTEND rlbk=0, rd_only=1
EXEC #1:c=58,e=69,p=87,cr=821,cu=44,mis=0,r=1,dep=0,og=4,tim=1659415
WAIT #1: nam='log file sync' ela= 1 p1=8 p2=0 p3=0
The main point is that it appears to insert into the snapshot table, end the
transaction (commit), delete from the master log table, end the transaction
(read only????). So it looks like you could actually update the snapshot
table without updating the master log. Also, why read only (XCTEND rlbk=0,
rd_only=1)?
Sorry for the confusing e-mail, but it's all I have time for right now.
Henry
-----Original Message-----
Sent: Thursday, September 20, 2001 2:41 PM
To: Multiple recipients of list ORACLE-L
Shailesh,
Thanks. My thoughts on #2 are that if the snapshot refresh and the delete
from the master log comprise a single transaction (2 phase commit), then I
shouldn't have a problem. A full refresh might be necessary at some point,
but I could deal with that with some specific, infrequently used code. Of
course I still need to confirm that this is a single transaction. Not quite
sure how to do this. Maybe try a trace, or limit available space on snapshot
so inserts will error. I'll see what happens.
Henry
-----Original Message-----
Sent: Thursday, September 20, 2001 1:56 PM
To: Multiple recipients of list ORACLE-L
Henry,
I think Anita's concerns are valid however in my experience of doing the
refreshes like you plan to do I have not encountered the problem# 2. For
problem# 1 what we do here is check the table you are going to populate the
changes into for existance of that primary key. If found then update the
record with new values else insert into the table.
The real catch here would be the deletions. Try to get the deletes before
you do the refresh from the MLOG$xxx at the master site and then use those
primary keys to perform the deletion.
We have about 100 snapshots that are used to perform the aforementioned
task for over a year without any problems. Let me know if you have any
problem.
HTH,
Shailesh
-----Original Message-----
Sent: Tuesday, September 18, 2001 11:52 AM
To: Multiple recipients of list ORACLE-L
Henry,
OK, I think I see what you're trying to accomplish.
By truncating the snapshot between refreshes, the
refresh is essentially populating the snapshot with
only the changed rows from the master site.
Probably not supported, but I can't see that it would
cause any real problems. The way the refresh
mechanism works, the missing rows on the snapshot site
shouldn't be a problem.
I see a couple of potential problems, however.
1. The refresh will pull over all changed rows on the
master site, in your case inserts AND updates. So if
a row does get updated on the master site, then it
will get refreshed to the snapshot site and your
counts will be off since you'll be treating the update
of an existing row as a newly inserted row.
It all depends on how critical the numbers are for the
developers, because I can assure you it's only a
matter of time before someone updates rows on a table
that should only have inserts ;)
2. If a fast refresh fails this requires that the
next refresh is a complete refresh, or the snapshot is
recreated, so you will not have a way of getting just
the set of changed rows. Your procedure will need to
be able to detect this and perform the joins against
the entire table again.
In the long run you're probably much better off
developing your own trigger to populate another table
or setting a flag, as you mentioned. Just because it
works today doesn't mean that it will work in a newer
release if they change the refresh mechanism.
HTH,
-- Anita
--- Henry Poras <[EMAIL PROTECTED]> wrote:
> Guess I wasn't too clear. Here's the scoop (there is
> probably a much easier
> way to do this, but I just can't get there from
> here). There is a 10million
> row table on a remote database. Each week, there is
> a change in about
> 5-10,000 rows (inserts). Currently this table is
> brought over to a local
> database. A convoluted query is then run to keep our
> developers happy and
> the result set is entered into another table. The
> query, though convoluted,
> does not use any aggregate functions, just a mess of
> joins. Right now this
> query is run on the complete 10 million row table
> brought over from the
> remote database. What I would like to do is to run
> the query on just the new
> rows, appending this result to the existing data.
> Since the snapshot log
> used for a fast refresh already keeps track of this
> for me I thought it
> might be a nice way to go. (I guess I could always
> flip a flag field, but
> why do extra work if Oracle already does it for me?)
>
>
> So my plan was to create a procedure which will
> manually take a fast
> refresh, run the query, check for success, truncate
> the table underlying the
> snapshot view. Since the underlying structure is
> just a table, and the
> master log data transfer is strictly in one
> direction, it looks like it
> should work, but are there any gothchas? It seems to
> be OK on a test system.
>
> Thanks again.
>
> Henry
>
>
> -----Original Message-----
> Sent: Sunday, September 16, 2001 9:38 AM
> To: [EMAIL PROTECTED]; Henry Poras
>
>
> Henry,
>
> I'm not sure what you're trying to accomplish by
> truncating the snapshot. The whole point of the
> refresh process is to keep the snapshot in sync with
> the master site.
>
> Presumably you are doing a CTAS off of the snapshot
> to
> do your manipulations, so what is the point in
> truncating the snapshot? I don't have a db handy to
> test whether Oracle detects that the snapshot has
> been
> truncated and does a complete refresh, but if the
> snapshot is very large it would probably be faster
> to
> drop and recreate it than to do a complete refresh.
>
> A more detailed explanation of the process would
> help.
>
> HTH,
>
> -- Anita
>
> --- Henry Poras <[EMAIL PROTECTED]> wrote:
> > !! Please do not post Off Topic to this List !!
> >
> > I need to move the changes in a table from one
> > database to another. It
> > seemed that fast snapshots would be a good way to
> do
> > this as the database
> > will automatically keep track of the changes for
> me.
> > The question is that
> > once I move the data to the snapshot side, I need
> to
> > manipulate it, move it
> > into another table, and start the cycle again.
> This
> > means truncating the
> > snapshot between refreshes. I tried this on a test
> > system and it seesm to
> > work, but I don't think Oracle supports it. I
> don't
> > want updatable snapshots
> > as this is strictly one way movement of data. Has
> > anyone else tried this?
> > Are there any potential problems? I think I am
> just
> > truncating a table
> > underlying the snapshot view so I don't see what
> > could go wrong.
> >
> > Thanks.
> >
> > Henry
__________________________________________________
Terrorist Attacks on U.S. - How can you help?
Donate cash, emergency relief information
http://dailynews.yahoo.com/fc/US/Emergency_Information/
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: A. Bardeen
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Yadav, Shailesh
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Henry Poras
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Henry Poras
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
