NoCOUG Summer Conference, Aug. 19, 2004 4750 Walnut Street, Suite 106 James F. Koopmann Boulder, Colorado 80301 Director of Technical Consulting 303.938.8282 x115 720.219.3773 (mobile) jameskoopmann@confio.com www.confio.com Get a handle on DDL using Oracle 9iR2 Streams
68
Embed
NoCOUG Summer Conference, Aug. 19, 2004 4750 Walnut Street, Suite 106 James F. Koopmann Boulder, Colorado 80301Director of Technical Consulting 303.938.8282.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
NoCOUG Summer Conference, Aug. 19, 2004
4750 Walnut Street, Suite 106 James F. KoopmannBoulder, Colorado 80301 Director of Technical Consulting
•3 technical articles a month•Forum expert / moderator
•Database centric vendor reviews
•Database general interest issues
Where to Find Me on the NET
1. The DBA nightmare
2. How do you detect DDL changes
3. STREAMS overview
4. STREAMS environment
5. STREAMS setup
6. STREAMS sample DDL session
7. OEM
8. Sweat Dreams for the DBA9. Running STREAMS
Content
The DBA Nightmare
• Problems
• Are problems “problems” only if seen
• Is it a problem if no one sees a benefit
• Should I be working on something that doesn’t solve a problem
• Problems are caused by change
We need to be able to answer the hard questions
• What has happened• When did it happen• Will it happen again
• We have to fix Problems
• Reduce finger pointing
• Research viable sources…Test / Validate…Acceptance
• Be weary of “Your mileage may vary” type statements
Problem Detection & Viable Solutions
Your actions are reactive, not proactive, in nature
The events that trigger an investigation are often very specific to an incident, narrow in scope, and the solutions typically do not take the full health of a database into consideration.
Most of your time is spent in problem investigation & detection, not problem solving
Investigation & detection inherently wastes money
Customers / users drive the work flow of the database administrators
Database administration group is seen as ineffective
The DBA Nightmare
Lazy Detection Methodology Shortcomings
Has the issue been seen by others? Will someone benefit from you working
on the issue? Are you solving a real problem that is
causing pain for your users? Before solving the problem, are you be
able to determine what was the true root cause?
Will solving the problem eliminate it from recurring?
How we size the buffer cache1. Just issue the SQL to see the hit ratio2. Increase the buffer cache if hit ratio is low3. Done if hit ratio is high
The DBA Nightmare
SQL> SELECT b.name, b.block_size,
100*(1 - ((e.physical_reads-b.physical_reads) /
((e.db_block_gets-b.db_block_gets)+
(e.consistent_gets-b.consistent_gets)))) buffhit
FROM beg_buffer_pool_statistics b, end_buffer_pool_statistics e
WHERE b.name=e.name AND b.block_size=e.block_size;
How to diff1. Create table beg_buffer_pool_statistics as select * from v$buffer_pool_statistics2. Run workload through system3. Create table end_buffer_pool_statistics as select * from v$buffer_pool_statistics4. Just issue the SQL to see hit ratio5. Increase the buffer cache if hit ratio is low6. Done if hit ratio is high
Buffer Cache - diff’d Hit Ratio
The DBA Nightmare
SQL> SELECT name, block_size, size_for_estimate,
estd_physical_read_factor, estd_physical_reads
FROM V$DB_CACHE_ADVICE
WHERE advice_status = 'ON';
Buffer Cache - Advisory
How to get the statistics1. Set the db_cache_advice to ‘READY’2. Set the db_cache_advice to ‘ON’3. Run a valid workload through the system4. Just issue the SQL to see report
1. Don’t reduce the size of your cache if you are going to incur more physical reads
2. Don’t increase the size of your cache if you are not going to reduce the number of reads
Buffer Cache - Advisory Report
The DBA Nightmare
How Do You Detect DDL Changes
What kind of shop are you in
1. You run a tight / secure database shop. (does it need to be tighter)
2. Changes are kept to a minimal. (will one slip by)
3. The time window is not sufficiently large between DDL changes.
4. You need an inexpensive investigative approach for DDL extraction.
5. Do you hedge on the side of caution or paranoia
6. Do you need to be concerned with each and every DDL statement
7. Are you pro-active
8. Do you like systematic approaches to solving problems
9. Do you need to be notified of problems
Do I Care About Capturing DDL
The DBA Nightmare
Have you tried to set policies?
I don’t know how many shops I have worked in and tried to set policies about :• NO changes after 3:00pm on friday
• Can’t use production for Test / Stress / QA
• ALL changes must be approved through DBA
• ALL changes must be implemented by DBA
• ALL source must go through QA & Test
• and the list goes on and on and on and on
Bottom line
1. Startup fallacy 2. Everyone is in a rush 3. Most don’t want to do the job the right or proper way the first time
Your get a call at 4am on Sunday
You spend time with data integrity
This change MUST be in today
You don’t know what anyone is doing
Not one will tell you what they are doing
You have to pick up the pieces
Attempts at DDL stability through policy
How Do You Detect DDL Changes
1. Take a look at Change DATEs in DBA views
• SELECT Statements against DBA views
2. Compare Database Objects Over Time
• Take Snapshot at T1
• Take Snapshot at T2
• Compare T1 to T2
3. Use LogMiner
• Search and Extract DDL from Current Redo Logs & Archived Logs
4. Streams
• Set of database structures to capture and replicate changes from a source database to a destination database
Methodolgies
How Do You Detect DDL Changes
Object Date Field What
DBA_OBJECTS YES CREATED LAST_DDL_TIME
DBA_TABLES Yes LAST_ANALYZED
DBA_INDEXES Yes LAST_ANALYZED
DBA_TAB_COLUMNS NO
DBA_IND_COLUMNS NO
The Good - We know when object is added / changedThe Bad - No GranularityThe Ugly - We have to still validate a change through looking at the total object
Take a Look at Dates In DBA Views
How Do You Detect DDL Changes
1. T0 - CREATE TABLE time_dba_tables with DATE-TYPE column
2. T1 - INSERT INTO time_dba_tables (SELECT sysdate,* FROM dba_tables)
3. T2 - INSERT INTO time_dba_tables (SELECT sysdate,* FROM dba_tables)
4. TC - SELECT <changes> FROM time_dba_tables t1, time_dba_tables t2
OUTER JOINS - for what was added and deleted
DECODES - for proper column comparison
t1.column != t2.column - WHERE Cluase is “OR’d” for interested columns
The Good - Build Yourself. No reliance on 3rd party software or database vendorThe Bad - Complicated code. (you better be good)The Ugly - Oracle could change / add tables that are important to object change
Compare Database Objects Over time
How Do You Detect DDL Changes
1. Lets you look at the redo log activity in your database
2. You can look in the current redo logs or archived logs
3. New in 9i is the ability to look at DDL statements
LogMiner
How Do You Detect DDL Changes
1. Create a dictionary EXECUTE DBMS_LOGMNR_D.BUILD(
4. Query logminer views SELECT sql_redo FROM V$LOGMNR_CONTENTS WHERE operation=‘DDL’
The Good - Easy to setup and useThe Bad - Pinpointing the time when something might have happenedThe Ugly - Manual process of adding logs and searching through them
LogMiner
Oracle Streams
1. Is a new feature that was introduced in 9iR22. Is a more robust replication environment that allows for the sharing of
information and structures3. Improved performance of replication because Streams utilizes the extraction of
DML and DDL from the redo log files through separate background processes.4. Supports the hot mining of the current active redo log for change data at the
same time it is written to the log.5. Supports the mining of archived log files.6. No loss of data7. Directed Networks
OverviewOracle StreamsA feature within the Oracle database that allows for the replication of database structures and information between two separate databases
Oracle Streams
A directed network is defined as a networked system of hosts that allow for the passing of information to a destination database where the destination host is not directly accessible to the source host.
Two methods of information propagation.
1. Queue Forwarding, information is received by an intermediate database, and automatically forwarded to another intermediate or destination database.
2. Apply Forwarding, information is received by an intermediate database, applied, captured, and re-queued and ready for propagation to another destination database.
Directed Networks
CAPTURESET_UP_QUEUE
ADD_GLOBAL_RULES
HCMC(source)
PROPAGATIONADD_GLOBAL_PROPAGATION_RULES
APPLYSET_UP_QUEUE
CREATE TABLE history_ddl_lcrs
CREATE PROCEDURE history_ddl
ADD_GLOBAL_RULES
ALTER_APPLY SAIGON (destination)
User DDL Change
Oracle Streams
A feature within the Oracle database that allows for the replication of database structures and information between two separate databases
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Data Definition Language (DDL)
SQL statements that affect the structure of database objects, such as CREATE TABLE, ALTER TABLE, DROP TABLE, and RENAME TABLE.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Source database
The database that originates information to be shared within the Oracle Streams environment
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Capture A capture background process runs (cp01-cp99) on the source database for each of the capture queues defined. As the DDL is issued within the Oracle database, a set of capture rules governs what is important and extracts these statements from the redo logs. These statements are then stored (staged) in a capture queue on the source database.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Propagation A mechanism that, through a set of rules, picks up the captured statements from the source queue and then transports the statements across a network to a destination database queue.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Destination database
The database that receives information in the Oracle Streams environment.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY SAIGON (destination)
User DDL Change
Apply The apply process (one for each capture process that is propagating to this site) runs on the destination database and picks statements off the queue that were transported from the source database. A set of apply rules then acts on the statements, either applying the changes to the database, transforming the statements into another usable form, or rejecting them.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
Rules From the capture process, through the propagation process, and into the apply process, rules govern what data are of concern, where to send the data, and what to do with that data.
Streams Environment
CAPTURE
HCMC(source)
PROPAGATION
APPLY
SAIGON (destination)
User DDL Change
The Good - Able to report on every DDL statement issued without interventionThe Bad - Learning curve is a bit highThe Ugly - Intensive & cumbersome setup
Streams Environment
CAPTURESET_UP_QUEUE
ADD_GLOBAL_RULES
HCMC(source)
PROPAGATIONADD_GLOBAL_PROPAGATION_RULES
APPLYSET_UP_QUEUE
CREATE TABLE history_ddl_lcrs
CREATE PROCEDURE history_ddl
ADD_GLOBAL_RULES
ALTER_APPLY SAIGON (destination)
User DDL Change
Streams Environment
1. All databases using Oracle Streams must be upgraded to version 9.2.0.2 (patch #2632931)
2. Run catalog.sql & catproc.sql after you have upgraded to version 9.2.0.2
Software Requirements
Streams Environment
1. The source database must be running in ARCHIVELOG mode in order to capture changes.
i. No data loss
ii. No database hanging caused by LGWR waiting for capture process to complete
iii. Tune your redo log generation before implementing streams
iv. Do not use the NOLOGGING option
Archive Log Requirement
Streams Environment
Parameter Setting NotesCOMPATABLE 9.2.0 or higher. For Streams to work,
you must set this to at least 9.2.0.
JOB_QUEUE_PROCESSES 2 or higher. Should be at least 2 in the environment for each database, but should be equal to the number of dbms_jobs that can run concurrently plus 2.
LOG_PARALLELISM 1 If you’re running one or more captures on the database.
LOGMNR_MAX_PERSITENT_SESSIONS. equal to or higher than the number of capture processes.
If this parameter is set too low, capture processes will not run and may also inhibit you from dropping the capture processes.
OPEN_LINKS 4 or higher. This allows for the distributed transactions between source database and destination database to occur.
PARALLEL_MAX_SERVERS current value + (3 * capture processes) + (3 * apply processes)
This is just an estimate.Should be monitored to handle the number of parallel execution servers.
PROCESSES current value + ((capture processes + apply processes) * 10).
Again this is just an estimate. Normally I have noticed that most installations do not account for enough processes.
1. If you move LogMiner after you have captured or applied, you will lose the Streams directory changes that have been recorded.
2. The Streams data dictionary is not kept clean by Oracle which can also cause greater strains on the Streams dictionary and allow it to grow uncontrollably.
To Do
1. Simple, don’t forget to move LogMiner
2. Remove an object is not being used you can clean out the Streams dictionary by using DBMS_STREAMS_ADM.PURGE_SOURCE_CATALOG for a particular object.
LogMiner / Streams Issues
Streams Setup
Why1. For transporting the captured DDL from the source database to the destination
database
How
1. CONNECT ddlman/ddlman@hcmc
2. CREATE DATABASE LINK saigon CONNECT TO ddlman IDENTIFIED BY ddlman USING ‘saigon’
When the capture process mines information from the redo log, it reformats this information into LCRs. These LCRs are specific to the type of information captured and the completely defines the changed that has occurred.
SYS.ANYDATAThis is an overloaded object type that can be of any scalar (number, varchar,
char,date…) or user defined data type. It has defined with it methods that allows us to query what type of true data type it holds as well as methods to retrieve the values
$ORACLE_HOME/rdbms/admin/dbmsany.sql
Logical Change Records (LCRs)
Streams Setup
1. CONNECT ddlman/ddlman@saigon 2. CREATE PROCEDURE history_ddl(in_any IN SYS.ANYDATA) IS
Before we can start capturing, propagating, and applying within our Streams environment we must instantiate the destination database. This is nothing more than registering the source SCN with the destination database so it knows the point in time it can start applying captured information.
There are three methods for instantiating objects1. Instantiating and creating database objects by using Export/Import 2. Setting the instantiation SCN for an existing table, schema, or database manually,
by executing procedures in the DBMS_APPLY_ADM package at the destination database
3. Performing a metadata-only export and import for existing tables or schemas You can verify the instantiation by querying DBA_APPLY_INSTANTIATED_OBJECTS.
HCMC-SQL> ALTER TABLE DDL_CHECK_TABLE DROP COLUMN COL2;Table altered.
HCMC-SQL> DROP TABLE DDL_CHECK_TABLE;Table dropped.
Make Some Changes
Sample DDL Session
SAIGON-SQL> SELECT timestamp,substr(source_database_name,1,4) source, 2 logon_user,command_type,object_owner owner,object_name name, 3 object_type type 4 FROM ddl_history;
TIMESTAMP SOURCE LOGON_USER COMMAND_TYPE OWNER NAME TYPE--------- ------ ---------- -------------------- --------- --------------- -----11-OCT-03 HCMC SYS CREATE USER AA USER11-OCT-03 HCMC SYS CREATE TABLESPACE11-OCT-03 HCMC SYS CREATE USER TEMPUSER USER11-OCT-03 HCMC SYS ALTER USER SCOTT USER11-OCT-03 HCMC SCOTT CREATE TABLE SCOTT DDL_CHECK_TABLE TABLE11-OCT-03 HCMC SCOTT ALTER TABLE SCOTT DDL_CHECK_TABLE TABLE11-OCT-03 HCMC SCOTT TRUNCATE TABLE SCOTT DDL_CHECK_TABLE TABLE11-OCT-03 HCMC SCOTT ALTER TABLE SCOTT DDL_CHECK_TABLE TABLE11-OCT-03 HCMC SCOTT DROP TABLE SCOTT DDL_CHECK_TABLE TABLE
View From DDL_HISTORY Table
Sample DDL Session
SAIGON-SQL> SELECT ddl_text 2 FROM ddl_history;
DDL_TEXT---------------------------------------------------------------------------CREATE user aa identified by VALUES '1468620FBA6271E8'create temporary tablespace temp01create user tempuser identified by VALUES '2B4C9C62A2919AEF'alter user scott identified by VALUES 'A7E7E0150C6D5EF3'CREATE TABLE DDL_CHECK_TABLE (COL1 NUMBER)ALTER TABLE DDL_CHECK_TABLE ADD (COL2 VARCHAR2(500))TRUNCATE TABLE DDL_CHECK_TABLEALTER TABLE DDL_CHECK_TABLE DROP COLUMN COL2DROP TABLE DDL_CHECK_TABLE
View From DDL_HISTORY Table
The DBA Nightmare
Remember Me
Sweet Dreams for The DBA
SQL Statements Executed Over Time
Time
Wa
it T
ime
DROP INDEX CREATE INDEX
db file sequential reads
buffer busy waits
parse call
Running Streams
CREATE/ALTER/DROP Tables includes table comments
CREATE/ALTER/DROP Tablespace (requires global rules to be set)
CREATE/ALTER/DROP Indexes
CREATE/ALTER/DROP Triggers
CREATE/ALTER/DROP Views
CREATE/ALTER/DROP Synonyms
CREATE/ALTER/DROP Sequences
Creation of PL/SQL packages, procedures and functions
Changes to users/roles
GRANT or REVOKE on users/roles
COMMIT
ROLLBACK
AUDIT (can be done on user objects)
DDL Types Captured
Running Streams
CREATE , ALTER, or DROP MATERIALIZED VIEW LOG
CREATE , ALTER, or DROP MATERIALIZED VIEW
CREATE or ALTER TABLE for Index-organized tables
CREATE SCHEMA AUTHORIZATION
CREATE or DROP DATABASE LINK
RENAME (use ALTER TABLE instead)
CREATE TABLE ? AS SELECT for clustered tables
DDL Types Captured But NOT Applied
Running Streams
CREATE or ALTER DATABASE ALTER SESSION ALTER SYSTEM TRUNCATE CREATE/ALTER/DROP ROLLBACK CREATE/ALTER/DROP TYPE CREATE/ALTER/DROP PROFILE CREATE/ DROP LIBRARY CREATE/ DROP DIRECTORY SET ROLE SET TRANSACTION SET CONSTRAINT CREATE CONTROL FILE CREATE SPFILE CREATE PFILE ANALYZE EXPLAIN CALL PL/SQL Procedural calls Lock Table
CREATE , ALTER, or DROP MATERIALIZED VIEW LOG CREATE , ALTER, or DROP MATERIALIZED VIEW CREATE or ALTER TABLE for Index-organized tables CREATE SCHEMA AUTHORIZATION CREATE or DROP DATABASE LINK RENAME (use ALTER TABLE instead) CREATE TABLE ? AS SELECT for clustered tables
DDL Types NOT Captured
Running Streams
Setup / Running Streams1. Status of dba_capture & dba_apply where ABORTED2. ORA-01925: Maximum of 30 enabled roles exceeded3. What object is that?4. Mixed case global_name causing mismatch Streams and LogMiner
Remedy1. Stop the Capture and apply processes and start them again2. Increase current value for MAX_ENABLED_ROLES
• ALTER SYSTEM SET max_endabled_roles=1003. Avoid using system generated names
• Creation will work• ALTER / DROP statements will more than likely Fail
4. Change db_name & db_domain to uppercase
Problems You May Encounter
Running Streams
Bugs1. Analyze statement is not propagated on streams environment to target database
with using DBMS_STREAMS_ADM.ADD_SCHEMA_PROPAGATION_RULES. 2. DDL Issued through the EXECUTE IMMEDIATE are not applied at destination3. Checkpoints are skipped if there is any DDL activity in the redo logs causing the
capture process to reprocess logs during startup.4. Dropping the streams environment is difficult to cleanup.
Remedy1. Can use DBMS_STREAMS_ADM.ADD_TABLE_PROPAGATION_RULES. 2. Do not use dynamic DDL3. Ensure that no DDL has been issued around scheduled shutdowns.4. Get it right the first time or choose a version naming schema.
Problems You May Encounter
Running Streams
Just SomeDBA_QUEUESDBA_QUEUE_TABLESDBA_APPLYDBA_APPLY_PARAMETERSDBA_CAPTUREDBA_CAPTURE_PARAMETERSDBA_PROPAGATIONDBA_APPLY_ERRORDBA_RULESDBA_RULE_SETSDBA_RULE_SET_RULESDBA_JOBSDBA_QUEUE_SCHDULES
Monitoring
NoCOUG Summer Conference, Aug. 19, 2004
4750 Walnut Street, Suite 106 James F. KoopmannBoulder, Colorado 80301 Director of Technical Consulting