11g XE beta: Oracle Database Express Edition

Added on September 2nd, 2011: the production release of Oracle Database Express Edition 11gR2 is now available for download.

As yesterday was April 1st, All Fool’s Day, I decided to wait a day or so before posting on the new free Oracle database edition: 11g XE beta. It is in fact out and downloading it took me less than 1 minute even via my slow wireless connection.

You can download it from http://www.oracle.com/technetwork/database/express-edition/11gxe-beta-download-302519.html and install it after accepting the license agreement:

You probably often click the “Accept” button and then “Next” but this time I decided to read the entire “Oracle Database Express Edition License Agreement”. You will also probably end up with some open questions. Just one example:

During installation, while reading the agreement you will see: “You may make one copy of the programs for backup purposes.” If you check the oracle.com license information, it says: “You may make a reasonable number of copies of the programs for backup purposes.”. But we should not forget that we are talking about a beta release of the software, so many things are to be fixed.

I would like to comment on two main things: restrictions and 11g upgrade.

1. 11g XE main restrictions and limitations:

- if Oracle Database XE is installed on a computer with more than one CPU (including dual-core CPUs), then it will consume, at most, processing resources equivalent to one CPU.

- only one installation of Oracle Database XE can be performed on a single computer which does not affect any existing installation or new installations of Oracle Database 11g Standard Edition, Oracle Database 11g Standard Edition One, or Oracle Database 11g Enterprise Edition. In addition, users can run only one instance of the Oracle Database XE database on each individual computer.

- the maximum amount of user data in an Oracle Database XE database cannot exceed 11 gigabytes compared to 4 gigabytes on 10g XE. If the user data grows beyond this limit, then an ORA-12592 error will appear. Note that 11G is just user data!

- the maximum amount of RAM that an Oracle Database XE database uses cannot exceed 1 gigabyte, even if more is available. The exact amount of RAM that Oracle Database XE uses is the sum of the System Global Area (SGA) and the aggregate Program Global Area (PGA).

- HTTPS is not supported natively with the HTTP listener built into Oracle Database XE.

- the list of all features not available with 11g XE is here. It is a rather long list and XE lacks lots of important stuff realted to HA, performance, securty, replication and DBAism.

2. Upgrade from 10g XE to 11g XE:

- copy the gen_inst.sql file from the upgrade directory of 11.2 XE home to some local directory
- connect to 10.2 XE database as SYS and run gen_inst.sql thus generating install.sql, gen_apps.sql and other .sql files in the local folder containing gen_inst.sql
- export the data from 10.2 XE database
- deinstall 10.2 XE if installation of 11.2 XE is planned on the same computer
- install the 11.2 XE database
- import the data into the 11.2 XE database
- connect to 11.2 XE database as SYS user and run the script install.sql, which was generated after running gen_inst.sql

The most important benefits of 11g XE are (as I see them):

A. Save from license costs by keeping small databases on top of XE. In case, you can live with the limitations of course.

B. On more global terms, you can use 11g XE for Database Sharding. 11 gigabytes of user data is quite sufficient for one shard! When it grows above say 10G, you will split it into at least two new shards.

P.S. Will 12g XE have the limitation of 12 gigabytes of user data ?

Oracle Database 11.2.0.3 new features

Oracle Database 11g Release 3 (11.2.0.3) was released on the 23rd of September 2011.

For now, 11.2.0.3 is available only for Linux x86 and Linux x86-64 and we have to download about 5G of data (7 files altogether).

11.2.0.3 is a full installation of the Oracle Database software meaning that you do not need to install 11.2.0.1 before installing 11.2.0.3.

According to the ReadMe for 10404530, the list of bugs fixed for Oracle Database 11g Release 2 (11.2.0.3) is in document 1348303.1: List of Bug Fixes by Problem Type available on the My Oracle Support Web site. Check it out please. It is not a short one. Not at all.

The complete list of new features of 11.2.0.3 can be found here. They are by far much less than the new features of 11.2.0.2, here is a summery of these new features:

Oracle ACFS Snapshot Enhancements
Oracle ACFS Security and Encryption Features
Support for ACFS Replication and Tagging on Windows
Oracle LogMiner Support for Binary XML
SQL Apply Support for Binary XML
Oracle LogMiner Support for Object Relational Model
SQL Apply Support for Object Relational Model
Deprecation of Obsolete Oracle XML DB Functions and Packages
Oracle Warehouse Builder Support for Partition DML
Enhanced Partitioning Support in Oracle Warehouse Builder
Oracle Warehouse Builder External Table Data Pump Support
Oracle Warehouse Builder External Table Preprocessor Support
Compressed Table and Partition Support in Oracle Warehouse Builder
Support for PL/SQL Native Compilation

Starting with Oracle Database 11g Release 2 (11.2.0.3) you can enter the Proxy Realm information when providing the details for downloading software updates. The proxy realm identifies the security database used for authentication. If you do not have a proxy realm, then you do not have to provide an entry for the Proxy Username, Proxy Password, and Proxy Realm fields. It is case-sensitive.

The following initialization parameter is new to Oracle Database 11g Release 2 (11.2.0.3): AWR_SNAPSHOT_TIME_OFFSET

Starting with Oracle Database 11g Release 3 (11.2.0.3), Oracle Universal Installer displays a new screen, Grid Installation Options, which helps in the installation of the examples software on an Oracle RAC database. The examples software is installed in the selected Oracle RAC home on all the nodes where it exists.

Nothing impressive but with a new patchset we do not actually expect new features, rather we hope that all (did I just say all) old bugs have been somehow fixed.

Unfortunately, I am waiting for the accessibility of document 1348303.1: List of Bug Fixes by Problem Type.

Exadata does not currently support Database 11.2.0.3.

When downgrading from release 11.2.0.3 to 11.2.0.2, the following error is raised when you run @catdwgrd.sql (reference Bug 11811073): ORA-20000: Upgrade from version 11.2.0.2.0 cannot be downgraded to version.

Data Pump Export operations do not work if the DMSYS schema is not removed as part of the upgrade to release 11.2.0.3 (reference Bug 10007411).

Oracle Database release 11.2.0.1 or 11.2.0.2 upgrade to Oracle Clusterware release 11.2.0.3 is not supported if the 11.2.0.1 or 11.2.0.2 release of Oracle Grid Infrastructure for a cluster is installed in a non-shared Oracle home and the 11.2.0.3 release of Oracle Grid Infrastructure for a cluster is installed in a shared Oracle home (reference Bug 10074804).

Starting with release 11.2.0.3 of Oracle Database, the Data Mining Java API is deprecated.

Starting in release 11.2.0.3, configuring HTTPS with Oracle XML DB requires that you first set up SSL_CIPHER_SUITES to include SSL_DH_anon (reference Bug 8403366).

After upgrading from 11.2.0.1 or 11.2.0.2 to 11.2.0.3, deinstallation of the Oracle home in the previous version may result in the deletion of the old Oracle base that was associated with it. This may also result in the deletion of data files, audit files, etc., that are stored under the old Oracle base.

Database links imported from an 11.2.0.3 database into a version prior to 11.2.0.3 (including 11.2.0.2) will not be usable in the import database. Any attempt to use a database link will cause the following ORA-600 error: ORA-00600 [kzdlk_zt2 err], [18446744073709551601]

For features not available or restricted in 11.2.0.3, click here.

For a complete list of all possible bugs and issues, check on the Open Bugs.

11.2.0.3 was called Oracle Database 11g Release 3 for a short period of time but Oracle fixed the documentation.

Restoring the Oracle database

If you are reading this, there is high chance that someone has been breathing behind your back while you were restoring a database or mildly harassing you over the phone with the only question: when do you think that the database will be open for WebLogic connections?

Indeed when? There is no way to say precisely. You can guess based on the time the backup had taken in average but that is just an estimate. After the restore, there might be a nasty long sequence of archivelog files that has to be applied (with the hope that none is missing) and then you can open with resetlogs. Unless you hit a “feature” or two.

There are 5 major reasons why RMAN restore performance may be poor:

1. Missing statistics: Before a restore even begins the the control file is queried, datafile headers are read, media managers are initialised, the catalog is resynced, PL/SQL is generated and compiled, RMAN metadata either in the control file or the catalog is queried. All this incurs I/O against the control file, the datafile headers and the recovery catalog.

Make sure that the catalog database is analyzed and reorganized on regular basis. It also requires regular maintenance: crosscheck backup, etc.

Generate statistics on the fixed objects, with the GATHER_FIXED_OBJECTS_STATS procedure. This should be done when the database has been running for awhile with generic workload, so the information in the fixed objects reflects a reasonable state of the database load.

Gather the statistics with the following command:

exec dbms_stats.gather_fixed_objects_stats;

Believe it or now, often adding this to the RMAN script might help for both faster backup and restore:

alter session set optimizer_mode=RULE;

2. The backup is hardware multiplexed: all channels write to a single tape device. If the media manager does not support multiplexing only one backuppiece can be returned at a time. Considering the fact that each backuppiece is multiplexed with the rest of the others, this restore would require as many scans as the number of channels allocated of at most the whole backup size each (with a tape rewind after each scan).

3. Individual files or tablespaces are being restored: A restore of a single file needs a scan of potentially the whole database backup depending on where the header of the file being restored is positioned on the tape. The worst situation is when single file is being restored so that the file’s block header is the last block written to the backuppiece.

4. A different number of channels is used for restore compared to backup: The number of channels to be allocated should be equal to the number of physical tape drives. Try to do a couple of things:

Set enough big large pool: set LARGE_POOL_SIZE to at least 256M
Enable the backup_tape_io_slaves by setting BACKUP_TAPE_IO_SLAVES = true

However, some media managers will allow hardware multiplexing to a single tape and are able to parallelize the restore (in the media manager layer) such that the backuppieces are returned from a single tape to multiple channels in parallel. In such a situation it is possible to use an unpublished SET PARALLELMEDIARESTORE OFF command to make RMAN ignore this check.

SET PARALLELMEDIARESTORE OFF;
      RUN { ...
        RESTORE DATABASE;
        RECOVER DATABASE; ...
      }

5. The RMAN relation with bugs is not a new one. The “mend-it-or-end-it” principle does not play any role here. 5 years ago, Jaffar posted on his blog Known RMAN Performance Problems. It is worth checking again the MOS note 247611.1

Yet, here are some Oracle or MML bugs not fixed by January 1st, 2012:

Bug 12363733: restoring backups are very slow (reproduced both with asm and filesystem)
Bug 7573468: RMAN duplicate until time very slow
Bug 11835641: 11gr2 – RMAN restore took more time – duplicate or restore is slow
Bug 10210264: RMAN restore is slower following upgrade to 11.2
Bug 11827990: select name from v$datafile wait on “rdbms ipc reply” on rman restore
Bug 12543119: EM not using correct ASM diskgroup to restore datafiles when creating clone
Bug 11892765: please allow RMAN restore to pre-allocate file extents
Bug 9556740: RMAN job failed when using sudo
Bug 9724316: dbclone not choosing the correct controlfile for the clone
Bug 9845029: OEM clone 11gR2 ASM database fails during recovery
Bug 5381095: RMAN: poor restore performance in nocatalog mode
Bug 10093810: RMAN backup need to maintain an index for file header block for each file
Bug 9918138: RMAN restoring a compressed backup running very slow
Bug 6964596: RMAN single channel restore performance is faster than 6 channels
Bug 5462916: RMAN: recover database skip forever tbs with ORA-01157
Bug 2608118: restore of autobackup controlfile take a long time to complete
Bug 1533686: RMAN hangs when restoring via adsm with an ops target db
Bug 2911857: bad performance of RMAN duplication with large number of files (> 1500)
Bug 13349761: RMAN RESYNC IS SLOW
Bug 6770620: RMAN recover of copy database command consumes CPU
–
Bug 13109129: after upgrade to 11.2 compressed RMAN backups are slow
Bug 13066081: RMAN unregister database is slow/hangs
Bug 12896388: RMAN backup slower with – check logical – option on compressed tablespace
Bug 13500084: RMAN switch command is taking too long
Bug 13064833: RMAN process hangs after finished backup
Bug 13520050: RMAN session does not terminate- loops -spinning on cpu when destination is full

Some of the bugs will never be probably fixed as you can see that they may get, and are closed (or suspended) because of vendor OS problems, problem cannot be replicated, closed as duplicate, information not avaiable, etc.

Or how about those bugs with status 92: closed, not a bug. For example bug 11835641 is actually not a bug? Right.

Additional information: 11.2 Oracle Database Backup and Recovery User’s Guide. Check Part VI, Tuning and Troubleshooting.