Materialized view being ignored

Hi all
Running Oracle 10.2.0.3 on Solaris 9.
I'm not sure how much info to provide at this stage but i'm having a problem with a materialized view.
The database contains tens of thousands of digitized books and an application to help manage all the data. Now, if I refresh the materialized view and run a full browse of all publications I get a response back in about 15 seconds. But then, within minutes this drops to a number of minutes. Again, as soon as I fresh the mat view it speeds up again.
Now, I know I have provided very little information here but I guess I'm wondering if the problem is something fundamental that other people may have come across/overcome.
If anyone can help shed light on this I would be very grateful.
Please let me know any other information that you require.
Many thanks
Farren Minns

user4662649 wrote:
Running Oracle 10.2.0.3 on Solaris 9.
The database contains tens of thousands of digitized books and an application to help manage all the data. Now, if I refresh the materialized view and run a full browse of all publications I get a response back in about 15 seconds. But then, within minutes this drops to a number of minutes. Again, as soon as I fresh the mat view it speeds up again.Farren,
one possible explanation is that by refreshing the materialized view you cache the data in your database and/or file system/SAN cache, so that immediately after the refresh only few physical I/Os are required. Once other activities have been performed that also use a significant part of your available cache, more physical I/O is required, so the overall response time increases. Without further details however this is just a speculation.
If you run your statement e.g. with AUTOTRACE enabled in SQL*Plus you could try if you can see a difference in the number of physical reads performed between a quick and slow execution. However this doesn't cover any caching performed outside the database buffers (e.g. file system cache), so tells you only half the truth.
If you trace your statement using extended SQL trace (including wait events) you could get more details, e.g. you could tell from the average wait times for the I/O requests whether they were served from an underlying cache or more likely real physical I/O.
Does your data contain LOB objects if you say "digitized" books, do you use one of Oracle's extensions to maintain that information (Oracle Text, Multimedia etc.)?
Regards,
Randolf
Oracle related stuff blog:
http://oracle-randolf.blogspot.com/
SQLTools++ for Oracle (Open source Oracle GUI for Windows):
http://www.sqltools-plusplus.org:7676/
http://sourceforge.net/projects/sqlt-pp/

Similar Messages

How to get Materialized View to ignore unused columns in source table

When updating a column in a source table, records are generated in the corresponding materialized view log table. This happens even if the column being updated is not used in any MV that references the source table. That could be OK, so long as those updates are ignored. However they are not ignored, so when the MV is fast refreshed, I find it can take over a minute, even though no changes are required or made. Is there some way of configuring the materialized view log such that the materialized view refresh ignores these updates ?
So for examle if I have table TEST:
CREATE table test (
d_id NUMBER(10) PRIMARY KEY,
d_name VARCHAR2(100),
d_desc VARCHAR2(256)
This has an MV log MLOG$_TEST:
CREATE MATERIALIZED VIEW LOG ON TEST with rowid, sequence, primary key;
CREATE MATERIALIZED VIEW test_mv
refresh fast on demand
as
select d_id, d_name
from test;
INSERT 200,000 records
exec dbms_mview.refresh('TEST_MV','f');
update test set d_desc = upper(d_desc) ;
exec dbms_mview.refresh('TEST_MV','f'); -- This takes 37 seconds, yet no changes are required.
Oracle 10g/11g

I would love to hear a positive answer to this question - I have the exact same issue :-)
In the "old" days (version 8 I think it was) populating the materialized view logs was done by Oracle auto-creating triggers on the base table. A "trick" could then make that trigger become "FOR UPDATE OF <used_column_list>". Now-a-days it has been internalized so such "triggers" are not visible and modifiable by us mere mortals.
I have not found a way to explicitly tell Oracle "only populate MV log for updates of these columns." I think the underlying reason is that the MV log potentially could be used for several different materialized views at possibly several different target databases. So to be safe that the MV log can be used for any MV created in the future - Oracle always populates MV log at any update (I think.)
One way around the problem is to migrate to STREAMS replication rather than materialized views - but it seems to me like swatting a fly with a bowling ball...
One thing to be aware of: Once the MV log has been "bloated" with a lot of unneccessary logging, you may perhaps see that all your FAST REFRESHes afterwards becomes slow - even after the one that checked all the 200000 unneccessary updates. We have seen that it can happen that Oracle decides on full table scanning the MV log when it does a fast refresh - which usually makes sense. But after a "bloat" has happened, the high water mark of the MV log is now unnaturally high, which can make the full table scan slow by scanning a lot of empty blocks.
We have a nightly job that checks each MV log if it is empty. If it is empty, it locks the MV log and locks the base table, checks for emptiness again, and truncates the MV log if it is still empty, before finally unlocking the tables. That way if an update during the day has happened to bloat the MV log, all the empty space in the MV log will be reclaimed at night.
But I hope someone can answer both you and me with a better solution ;-)

Moving data from Materialized View to downstream application

Database Version: 9i Release 2
I've created a Materialized View on a base table, the purpose of which is to capture data changes to send to a downstream application
(I've been advizsed to avoid using Streams for now until we upgrade to 10g)
I need to either:
a) Make that data available for another process to pick up and have knowledge of what it needs to pick up versus what it has already picked up
or
b) Send/Update a table or MV at a downstream MS SQL Server database (resides within our firewall)
Some notes/questions
I cannot change the structure of the underlying base table
On the MV I could capture a date-time which records each time a record is touched on the base table, the problem is that many of these updates are “meaningless” and I should not send such updates downstream
1.Is there a way I can timestamp/flag the records as they get inserted/updated on the MV? (potentially used by the pickup process to identify what it needs to pick up)
2. If I flag the records e.g. “needs pickup” – would it be ok to make the MV updateable so the process can then mark the records as “picked up”. Performance?
3. Can I create a trigger on the MV to write records to a secondary table as they are inserted/updated on the MV?
4. Could I use the MV log tables MLOG$ to help me in any way?
I’m guessing there is no way to do the following but:
5. Is there a way to include a column on a Materialized View but ignore it in terms of capturing changes. e.g. I have columns a,b,c,d,e on my base table, I want to include a,b,c in my Materialized View but I only want my MV to be updated if a or b is updated on base table. If only column c is updated do not send any update to MV
Many thanks for any insight

Yes I do have a unique identifier on each record in the base table and therefore also in the MV.
So now the Downstream application has to update some of its own database tables based on what is in the MV that I provide - just so I understand how would that work e.g.:
1. MV gets refreshed from my base table every minute
2. Some records are newly inserted and some simply updated
and this is the part I don't understand (I might be missing something obvious)....
3. Downstream application interrogates the MV every (say 1-2 minutes) and updates its own tables based on the data there.....but let's say MV contains 10,000 records and in the last minute 100 of those records have been updated and another 100 inserted - how will it know the 200 records it needs to pull?
Thanks again

Materialized views used by answers

Hi,
I had create some materialized views for some graphs created on answers of Analytics. The problem i have is that the graphs still take much time to show the results, the materialized views has summarized measures from they original table which is of 33 millions rows to 2000 rows.
Is there any option do i have to check on the Administration of ORACLE Business Intelligence for the materialized views being take by the queries?
thanks a lot

Tables were not created and imported into the physical Layer
- What was imported into the physical layer if not tables?
Do i have to add the materialized views in the physical Layer?
- If query rewrite is turned on then the materialized views don't have to be. Oracle can rewrite sql in the backgrounf
I haven't eneabled the QUERY REWRITE, so maybe with this can used it.
- Worth a shot
Where do i have to check the statistics? in OEM or Settings-Manage Sessions ?
- You can use OEM to analyze tables and materialized views (and their indexes)
- Speaking of indexes - Are the indexes being used or is table access done with full table scans
- Beach

Materialized view not being used in the report

I have had a materialized view (MV) for a particular report. Recently, the report definition has changed little bit, so the existing MV is not being used by the report anymore. I copied the code Discoverer's SQL Inspection and recreated the materialized view. But still it is not working. Obviously, I am missing something at the database level. Does anyone have any idea? The database is 9i and Discoverer is 9.0.2. Thank you!

The SQL that you see from the Inspect SQL option will provide you with the SQL that Discoverer sends to the database. The database then may do a query rewrite to point to an available materialized view if available. This rewrite won't be seen from the inspect sql option. You need to check at the database level to verify if a query rewrite did in fact take place.
You may want to check the section titled "Query Rewrite with Materialized Views" from Ch1 - "Introduction to the Optimizer" in the Oracle9i Database Performance Tuning Guide and Reference, Release 2 (9.2), Part Number A96533-02, available at http://download-uk.oracle.com/docs/cd/B10501_01/server.920/a96533/optimops.htm#37287
Regards
Abhinav Agarwal
Oracle Business Intelligence Product Management
http://www.oracle.com/bi
http://www.oracle.com/technology/documentation/bi_ee.html
http://www.oracle.com/technology/software/products/ias/htdocs/101320bi.html
http://oraclebi.blogspot.com/

Materialized View Fast Refresh been ignored For Complete refresh.

Hi,
I have a script that create a materialised view with a refresh method of 'FAST'.
When I refresh the view on one machine it refreshes using the 'FAST' method (i.e. LAST_REFRESH_TYPE = FAST) in 22 seconds, however, when I run the same script on a different machine the database chooses a 'COMPLETE' refresh method, despite me stating explicitly for it to use the FAST method. The complete refresh completes in about four hours.
Does anyone know the logic that Oracle uses in determining the type of refresh to use?
Thanks.

Assuming that the Materialized View definition is the same (and doesn't conflict with any FAST REFRESH requirements for that specific Oracle Version), check to see if the source table(s) has/have Materialized View Logs defined on them. MV Logs with PK/Rowid would be required.
Hemant K Chitale

Materialized View - Point in time refresh

I have a MV group which refreshes from a bunch of master tables. The master tables are constantly getting updated from an OLTP application. In this scenario, I want all the MVs to hold data which was available at the time the refresh started and not the changes after that.
Eg:
Master tables are T1, T2 ... T20 refresh to MVs T1_MV, T2_MV ...T20_MV
All the MVs T1_MV to T20_MV are in a refresh group.
At 3 PM: Execute an explicit refresh for the MV refresh group
At 3:05 PM: OLTP application inserts a record into table T5 & commits
At 3:06 PM (rather any time after 3:05 PM): Refresh of T5_MV happens. Will the latest insert get captured in T5_MV? I actually don't want this to happen for my requirement, because I want the data captured as of 3 PM into the MVs.
The Oracle Advanced Replication guide says the following about Materialized Views -
+"A materialized view is a replica of a target master from a single point in time."+
From the above line, it looks like by default this will not be captured in the MV. Can someone please confirm my understanding ?

MVs in a Refresh Group when refreshed as a Refresh Group are refreshed as of the same point in time.
So if the Refresh Group begins at 3:00pm and ends at 3:25pm, all the MVs in that group reflect the data as in the source tables as of 3:00pm, ignoring any updates to source tables after 3:00pm (i.e. while the Group is being refreshed).
However, if you refresh MVs individually, then each MV is refreshed at it's relative point in time -- so individual Refreshs of MVs would not be to the same point in time.
How have you defined the Refresh Group ?
How do you execute the Refresh for the whole Group ? What PLSQL procedure do you use ? What parameters do you use ?
Ensure that you do not execute an individual MV Refresh.
Hemant K Chitale
http://hemantoracledba.blogspot.com

Select query in materialized view with two dblinks

Hi All,
We have oracle 10g On windows.
We are trying to create materialized view. Scenario is we have base table on other database and we are creating mview on different database.
Basa database have two schema's and i am selecting records from that two schema's using two private db links.
But when i am tryin gto create mview its not getting created. After 15 hrs. its still showing creation command and not finished.
Query is :-
Is it good practice to have two db links in select query of materialized view.

Billy Verreynne wrote:
Chanchal Wankhade wrote:
Is it good practice to have two db links in select query of materialized view.Same db link being used twice, or two different db links?
If the former, you ideally want the local Oracle db to send the join to the remote database, and for the remote database to drive the join between those 2 tables. There is a hint (<i>driving_site</i>) that can be used - or the join query can be defined on the remote database as a view, and the local materialised view can then use that remote view.
If you have 2 different db links and joining across these - usually a bad idea to perform distributed database joins. There are lots of limitations as to how the tables can be joined. Worse case, full table scans of both remote tables, pulling all the rows from the 2 remote database tables to the local database, and joining these on the local database.
I have seen some severe performance issues in the past as a result of distributed joins. I'll rather use 2 materialised views for pulling both distributed tables's data locally, and then do the join on local data (using indexes, partition pruning, etc)Hi Billy,
My scenario is i have two database database A and database B. Database A is having two schema's SCOTT AND HR. SCOTT schema have select privileges on HR schema.
DB LINK is between Database B to Database A. name is db.link.B.A.oracle.com.
What if i priovide while creating materialized view, the schema name before the table name in database B for this particuler table so it will pick up the table from that schema using same DB LINK(db.link.B.A.oracle.com.) that i am using to fetch records from SCOTT schema.
Above schnario is like two base schema's and one db link using two schema.

Subquery in a Materialized View

I have a requirement where we have to build a summary view to show a customer's latest transaction. There can be several million customers so I am thinking if a Materialized view that gets refreshed once every day is a good option here. This is a Datawarehouse application where the records get appended once every day. I would like a summarized materialized view that is dropped and refreshed every day morning.
Here is the query that gets the latest order amount for each of the customer:
SELECT cu.cust_name,
od.order_amount
FROM customers cu,
orders od
WHERE cu.cust_id = od.cust_id
and od.order_date =
(select max(order_date) from orders od
where od.cust_id = cu.cust_id)
My questions are:
1) Is the materialized view a snapshot? Or does it go against the underlying table every time the MV is accessed, much like a regular view?
2) I would like a snapshot type of solution here. Since there could be 100 million orders for 1 million customers, a summary snapshot of 1 million records would perform much better IMO. Correct me if I got it wrong.
Thank you in advance for your help.

Users query will be explicitly against the Mview.
After doing some reading on the MView, I am more inclined to think that Mview is not an ideal solution for my requirement here.
SELECT cu.cust_name,
od.order_amount,
MAX(od.order_date) over (PARTITION BY od.cust_id) max_date
FROM customers cu,
orders od
WHERE cu.cust_id = od.cust_id
The orders is a 20+ million records table. I thought by using a Mview that holds the maximum order date will avoid going to the orders table when users want to see that is the latest order date for a customer. However, the report requirements are such that the users will want to see what is the maximum order date within a certain date range. If this being the case, there is no point in using an Mview.
Once again, pl correct me if my understanding is not correct.

Regd FAST refresh option in a Materialized view

Hi All,
I am using a pipeline function in which I am creating a table of records and a few cursors to fetch data from various tables.
Now this PL/SQL table is being used to construct a Materialized view.
Creation of Materialized view is happening fine but not with FAST refresh option. It gives an error " Cannot create a FAST refresh Materialized view from a complex query."
The query which I have used for the view creation is
CREATE MATERIALIZED VIEW CUSTOM.ABC
PCTFREE 0
BUILD IMMEDIATE
REFRESH FAST ON DEMAND
AS
SELECT A.Number,
A.Guarantors_Number,
A.Guarantors_Name,
A.Personal_Garantee_PCNT,
A.Company, LG.Source_System,
A.Type_of_Info,
A.File_Gen_Date,
A.Periodicity
FROM
TABLE(CUSTOM.CDM_LG_PACK_PF.CDM_LG_FUNC) A;
where CDM_LG_PACK_PF is the package and CDM_LG_FUNC is the pipeline function I have written to fetch all the records.
Please help me on how can I do a FAST refresh on this materialized view.
Thanks in advance,
Gaurav

Welcome to the forum!
FAST refresh doesn't mean that the operation is fast (time wise), it means it's an incremental refresh.
If you have a complex query, you can't use a FAST refresh - that's what the exception tells you.

Creating a Materialized View Log After the Data has been instered

Hi,
I am trying to create a method of replication from Oracle to MySQL using the Materialized View Log table.
This has been done before and works quite well, the only problem is that I am trying to impliment the log after the table has been created and populated and wish to push all the existing data through the log file...
Does anyone know if it is possible to refresh the Materialized View Log and not a Materialized View.
The way the replication is intended to work is:
Oracle> Data inserted into table
Oracle> writes the vector data to the MVL
Script> Monitors the MVL and can see the changes being made to the Oracle Table
Script> Updates MySQL with the data and removes the rows from the MVL
MySQL is then used with other unix systems
Currently we export the data from the table using Triggers and a cronjob running every x minute to check for changes in the TriggerTables
Many thanks for your time on this, I have been checking for almost a whole working day and not found the answer to this problem.

Thats what I thought, the MVL will only read data that has changed since it was created and wont have the option to load in all the data as though it was made before the table was created.
From what I have read, the MVL is quicker than a Trigger and I have some free code that prooved to work from a MVL using it as a reference to know what records to update. There is not that much to a MVL, a record ID and type of update, New, Update or Delete.
I think what I will have to do is work on a the same principle for the MVL but use a Trigger as this way we can do a full reload if required at any point.
Many thanks for your help.

Duplicates in Materialized view

Hi,
I have created a materialized view in Database B from 5 tables in DATABASE A. Each table has millions of records. The query for materialized view joins all these 5 tables and gets the data. If I do complete refresh it takes upto 4 hours. I want this materialized view refreshed every 3 hrs since that table will be used very often. So I decided to do fast refresh every 3 hrs. But when I created the materialized view first time there are no duplicates since it was a complete refresh. But after that each time the fast refresh happens I see different number of duplicates in the materialized view. Very rarely one or two times there is no duplicate. Why is this happens. If I add Unique index then there is an error message stating "Unique constraint Violated" (this is since the existing records are being inserted again instead of updated).
Any help is appreciated.
Thanks...

Yes. I have the materialized view logs for all the 5 tables in the remote database. First time when I created the materialized view it took around 4 hrs and there were no duplicates. I am doing the fast refresh every 6 hrs. After each fast refresh there are different number of duplicates. I don't have access to get the explain_mview to see the capabilities. I requested my DBAdmin to get that. In the meantime, if you have any other thoughts please let me know. I suspect whether the materialized view logs are not working as expected. But not sure.

Materialized view with join

In 10g release 2,I tried to create following materialized view with join:
test_link is a normal table
test_geom is a table contains a column in SDO_GEOMETRY
CREATE MATERIALIZED VIEW LOG ON test_link with rowid
CREATE MATERIALIZED VIEW LOG ON test_geom with rowid,primary key
CREATE MATERIALIZED VIEW MV_LINK USING INDEX REFRESH FAST ON DEMAND AS
SELECT li.rowid link_rowid,geom.rowid geom_rowid,li.link_id,geom.link
FROM test_link li, test_geom geom
WHERE li.link_id=geom.link_id
But I always got an error like:
ORA-12015: cannot create a fast refresh materialized view from a complex query
If I change the geometry table to another table, everything works fine.
Anyone have ideas?

Unfortunately, creating a fast refreshable materialized view on a join with one of the select columns being a user defined type (sdo_geometry is a user defined type) is not allowed. See 5303489 in the metalink bug database.
You could do like the workaround in the article suggests and create two materialized views and then create a regular view on top.
In our scenario, our materialized view also contains unions, so we would really like to have one physical object at the end of the day. One approach that we are currently investigating is to create the materialized view (MV1) without the geometry column, which makes it fast refreshable, and also create a materialized view (MV2) on the table containing the geometry column. MV2 is also fast refreshable. We then create a table (T3) that contains all of the columns from MV1, plus a geometry column. An insert, update, delete trigger on MV1 is created. The trigger is used to push all of the columns from MV1 to T3 and the geometry column from MV2 to T3. I have created the above in one of our test environments and haven't encountered any issues yet.
Let me know if you come up with a better approach.

Are Cube organized materialized view with Year to Date calculated measure eligible for Query Rewrite

Hi,
Will appreciate if someone can help me with a question regarding Cube organized MV (OLAP).
Does cube organized materialized view with calculated measures based on time series Year to date, inception to date eg.
SUM(FCT_POSITION.BASE_REALIZED_PNL) OVER (HIERARCHY DIM_CALENDAR.CALENDAR BETWEEN UNBOUNDED PRECEDING AND CURRENT MEMBER WITHIN ANCESTOR AT DIMENSION LEVEL DIM_CALENDAR."YEAR")
are eligible for query rewrites or these are considered advanced for query rewrite purposes.
I was hoping to find an example with YTD window function on physical fact dim tables with optimizer rewriting it to Cube Org. MV but not much success.
Thanks in advance

I dont think this is possible.
(My own reasoning)
Part of the reason query rewrite works for base measures only (not calc measures in olap like ytd would be) is due to the fact that the data is staged in olap but its lineage is understandable via the olap cube mappings. That dependency/source identification is lost when we build calculated measures in olap and i think its almost impossible for optimizer to understand the finer points relating to an olap calculation defined via olap calculation (olap dml or olap expression) and also match it with the equivalent calculation using relational sql expression. The difficulty may be because both the olap ytd as well as relational ytd defined via sum() over (partition by ... order by ...) have many non-standard variations of the same calculation/definition. E.g: You can choose to use or choose not to use the option relating to IGNORE NULLs within the sql analytic function. OLAP defn may use NASKIP or NASKIP2.
I tried to search for query rewrite solutions for Inventory stock based calculations (aggregation along time=last value along time) and see if olap cube with cube aggregation option set to "Last non-na hierarchical value" works as an alternative to relational calculation. My experience has been that its not possible. You can do it relationally or you can do it via olap but your application needs to be aware of each and make the appropriate backend sql/call. In such cases, you cannot make olap (aw/cubes/dimensions) appear magically behind the scenes to fulfill the query execution while appearing to work relationally.
HTH
Shankar

Create materialized view with specific column sizes

Hi all,
I'm trying to create a materialized view with a specific a column size. Something like
create materialized view test_mv
refresh force on demand
as
select id,
       cast(my_compound_field as nvarchar2(50))
from ( select id,
              field1 || field2 my_compound_field
       from   my_table);But Oracle seems to ignore the cast and takes the maximum size it finds for field1 || field2 in the select query. The resulting table has a column nvarchar2(44) instead of nvarchar2(50).
This can give a problem when the view is refreshed... there could be new data that exceeds the current size, i.e. where length(field1 || field2) > 44.
How can I override the column size of a field in a materialized view?
Edit: Some additional info to clarify my case:
field1 and field2 are defined as nvarchar2(25). field1 || field2 can theoretically have a length of 50, but there is currently no data in my table that results in that length, the max is 44. I am afraid that there will be data in the future that exceeds 44, resulting in an error when the MV is refreshed!
Edited by: Pleiadian on Jan 25, 2011 2:06 PM

Cannot reproduce what you are saying is happening.
SQL> create table t (a nvarchar2(50), b nvarchar2(50));
Table created.
SQL> create materialized view tmv as
2 select a, b, a || b c from t;
Materialized view created.
SQL> desc tmv
Name                                      Null?    Type
A                                                  NVARCHAR2(50)
B                                                  NVARCHAR2(50)
C                                                  NVARCHAR2(100)
SQL> drop materialized view tmv;
Materialized view dropped.
SQL> create materialized view tmv as
2 select a, b, substr(a || b, 1, 10) c from t;
Materialized view created.
SQL> desc tmv
Name                                      Null?    Type
A                                                  NVARCHAR2(50)
B                                                  NVARCHAR2(50)
C                                                  NVARCHAR2(10)
SQL> select * from v$version;
BANNER
Oracle Database 11g Enterprise Edition Release 11.1.0.7.0 - 64bit Production
PL/SQL Release 11.1.0.7.0 - Production
CORE    11.1.0.7.0      Production
TNS for Linux: Version 11.1.0.7.0 - Production
NLSRTL Version 11.1.0.7.0 - Production
SQL>Edited by: 3360 on Jan 25, 2011 8:10 AM
And with data
SQL> insert into t values ('3123423423143hhshgvcdcvw', 'ydgeew gdfwe   dfefde wfjjjjjjj');
1 row created.
SQL> commit;
Commit complete.
SQL> desc tmv
Name                                      Null?    Type
A                                                  NVARCHAR2(50)
B                                                  NVARCHAR2(50)
C                                                  NVARCHAR2(10)
SQL> select * from tmv;
A
B                                                  C
3123423423143hhshgvcdcvw
ydgeew gdfwe   dfefde wfjjjjjjj                      3123423423

Materialized view being ignored

Similar Messages

Maybe you are looking for