Learn some of the Oracle 11g gems that often gets unnoticed. Features that are not revolutionary in nature but can greatly simplify day-to-day life of any Oracle DBA.
The presentation is based on the material presented by Christo Kutrovsky at the Oracle Open World 2007 and later used by me to produce content for this presentation at the IOUG Collaborate 2008.
Pythian provides database services and consulting for Oracle, MySQL and SQL Server as well as EBS support and SA services. I’m been with the company for almost 3 years - in Sydney since September 2008 heading our operations in East Asia Pacific. We are a global company with main offices in Canada, the US, Europe and Australia. Over 12 years of operations, we have served 400+ clients and currently we have 100+ active clients world-wide.
Ask your questions right away - too many different topics
Not finished
Not finished
Not finished
Association - warehouse
Introduced in 10g and surprisingly stable
Based on old technology
INTRO - ASM and RDBMS - for ASM instances and for DB compatible init.ora parameter. ASM compat - metadata format. RDBMS compat - files content format.
Default 10.1 - no new features. Defined per disk group.
Case study - upgrade to 11g & downgrade, smooth upgrade path, transportable tablespaces.
select i.ksppinm, v.ksppstvl from x$ksppi i, x$ksppcv v where i.ksppinm in ('_rdbms_compatibility','_asm_compatibility') and i.indx=v.indx;
1. start rolling migration
2. upgrade all instances one by one
ASM provides limited services
3. stop rolling migration
AU - smallest chunk of space that Oracle can allocate; coarse stripe size
problem - inflexible physical layout + high overhead for VLDB
10g - underscore parameter - instance-wide + undocumented/unsupported
11g - official way to change AU size per diskgroup. AU 1-8 MB - compat 10.1 but 6-64 MB - compat 11.1.
Fine striping is not changed!
Fully automatic. Explain how it works.
Main purpose - reduce overhead on disk mount and file open for huge files, sga overhead for extent maps.
10g - disk failure -> BROKEN & rebalancing
11g FMR - disk failure -> OFFLINE -> wait for disk_repair_time (3.6h default) -> rebalancing if too log or ALTER DISKGROUP DISK ... ONLINE;
* can do manual ALTER DISKGROUP DISK ... OFFLINE;
FMR - case study SAN firmware update, FMR - case study extended clusters
FB - case study - rebalancing takes days and impacts performance but night can be offline
Case study - extended clusters - in addition to Fast Mirror Resync
cp - doesn’t work for controlfiles but works for datafiles and spfle; strange path copying to ASM
remap - ASM detects corruption on read and re-allocates extent. remap causes ASM to read blocks and repair if needed.
md_backup/md_restore - can re-create DG on disks and directory structure
Use case - recovery window of 7 days + monthly for one year
KEEP FOREVER - requires catalog
Deletion policy shipped is new in 11g
DB_NIQUE_NAME - useful for standby
ADVISE - human language; REPAIR .. PREVIEW - script
CHANGE - change priority, close
FROM ACTIVE DATABASE - no UNTIL clause - time is chosen automatically based on the last datafile copy completed
PASSWORD FILE - by default for standby and optional for normal duplicate.
Use case for Active Data Guard - use DR for reporting without RTO impact, management is happy - DR hardware is utilized!
RTO = Recovery Time Objective - how quick we can recover.
Use case for snapshot - testing
1. convert to snapshot
2. test on it (database diverges)
3. convert to physical standby
DB_UNIQUE_NAME - unsure that no standby has the same DB_UNIQUE_NAME as primary
Invisible indexes are maintained as usual by all DML’s but not used by CBO by default.
Old way - add NULL column, updated NULLs, change to NOT NULL
Doc says 1000000=infinity. To test - wait 11+ days
Can also do FK
Redesign physical data and keep referential integrity
I did a test in single process - no visible difference
Behind the scene does the same!
Laurent Schneider has interesting blog post about case-sensitive passwords