An extremely good read for understanding how Oracle deals with clean/dirty buffer using UNDO tablespace. It’s pretty confusing though!
Originally posted on Oracle Scratchpad:
There is some confusion about the expression “clean” in Oracle circles, so I thought I’d write a short note to explain the different ways in which the word may be applied to Oracle blocks. There are five terms to consider:
- commit cleanout
- block cleanout
- delayed block cleanout
- delayed logging block cleanout
View original 1,187 more words
[ALERT] – Oracle products affected by Shellshock Bash Bug
This Security Alert addresses CVE-2014-7169 (initially identified as CVE-2014-6271), a publicly disclosed vulnerability affecting GNU Bash. GNU Bash is a popular open source command line shell incorporated into Linux and other widely used operating systems. This vulnerability affects multiple Oracle products. This vulnerability may be remotely exploitable without authentication, i.e. it may be exploited over a network without the need for a username and password. A remote user can exploit this vulnerability to execute arbitrary code on systems that are running affected versions of Bash.
Some of the affected products are;
- Oracle Audit Vault and Database Firewall
- All Oracle Engineered Systems
Systems that are public facing SHOULD be patched immediately.
Please refer to this link.
Oracle Database – How to create control files from scratch
Below is an example the “Create Controlfile” commands”.
CREATE CONTROLFILE REUSE DATABASE "TESTDB" RESETLOGS FORCE LOGGING ARCHIVELOG MAXLOGFILES 16 MAXLOGMEMBERS 3 MAXDATAFILES 100 MAXINSTANCES 8 MAXLOGHISTORY 292 LOGFILE GROUP 1 ('/u01/app/oracle/oradata/testdb/redo1a.log','/u01/app/oracle/oradata/testdb/redo1b.log') SIZE 5M, GROUP 2 ('/u01/app/oracle/oradata/testdb/redo2a.log','/u01/app/oracle/oradata/testdb/redo2b.log') SIZE 5M, GROUP 3 ('/u01/app/oracle/oradata/testdb/redo3a.log','/u01/app/oracle/oradata/testdb/redo3b.log') SIZE 5M DATAFILE '/u01/app/oracle/oradata/testdb/system01.dbf', '/u01/app/oracle/oradata/testdb/undotbs01.dbf', '/u01/app/oracle/oradata/testdb/sysaux01.dbf', '/u01/app/oracle/oradata/testdb/users01.dbf' CHARACTER SET WE8ISO8859P1;
Do note the following though;
- The database has lost the online redo logs
- If ASM is used, change the path to “+ASM/../..” accordingly
- The database have archive logging enabled
- Character set of WE8ISO8859P1
Hope this helps.
How to list all hidden/undocumented parameters in Oracle Database
Oracle database is extremely powerful due to many reasons. One of it is the level of tuning and optimisation that you can perform on it. You can optimise an Oracle database using parameters in PFile or SPFile. It has many parameters, both documented and undocumented for you to configure. Documented parameters are easy to configure due to the extremely clear documentation by Oracle. However, this does not apply to undocumented or hidden parameters. Of course, the parameters are hidden for a reason and it should NOT be changed unless you are extremely confident of what you are changing!
The query below will be able to list down all the hidden parameters for the database release.
SQL> select a.ksppinm, b.ksppstvl FROM x$ksppi a, x$ksppcv b WHERE a.indx=b.indx;
This is the output in Oracle Database 18.104.22.168. There are about than 2500 hidden parameters. I’m only showing a portion about it.
KSPPINM KSPPSTVL ---------------------------------------- --------------------------------------------- _appqos_qt 10 _ior_serialize_fault 0 _shutdown_completion_timeout_mins 60 _inject_startup_fault 0 _latch_recovery_alignment 65534 _spin_count 1 _latch_miss_stat_sid 0 _max_sleep_holding_latch 4 _max_exponential_sleep 0 _other_wait_threshold 0 _other_wait_event_exclusion 0
Hope this helps!
I was doing some researching on database migration from AIX and Solaris and I found some juicy stuff! Oracle has provided some free lab session for new Solaris administrators to get their hands dirty. If you are new to Unix and want to try them out, you can check out the links below!
- Explore the basics of Oracle Solaris 11 administration compared to IBM AIX administration
- Other Solaris hands on lab
Just a side note, I was recently exposed to Wintel environments, either Oracle/MSSQL on Windows Server 2008/2012. It was a utter pain to work with the permissions and services. Unix/Linux is still the number one enterprise operating systems, period.
Oracle DataGuard VS Storage Replication
Recently, I was involved in various projects on the design of the DR solution for Oracle databases. This question popped up at almost every time. I strongly believe that Oracle DataGuard is the correct solution for any Oracle DR solution.
Oracle DataGuard comes free with Oracle Enterprise Edition (EE)! The only reason why a company might NOT use Oracle DataGuard is because they are not on EE and the cost of upgrading to EE just to use DataGuard does not justify the price.
Lower Bandwidth Utilisation
An insert to a table in Oracle generates alot of IO activities.
- 8k insert into table
- update of leaf indexes
- 8k block of undo generated
- redo log generated
- archive log generated
All these unnecessary IOs are replicated to your DR site. If you are using DataGuard, only the stream of redo will be sent over to the DR :) Link
Lower Complexities and easier to manage.
During DR implementation.
The steps to create an Oracle DataGuard configuration are very well documented and straightforward. It is especially so if you are an Oracle DBA. If you find that the manual configuration (link) are too difficult, you can always create the physical standby using Oracle Enterprise Manager (OEM).
However, for storage replication, the LUN configuration must be extremely precise. All the LUNs used by the same database should be in the same consistency group. If it is a RAC database, then the OCR diskgroup must not be mirrored.
During DR exercise.
In Oracle DataGuard, to switchover to the standby database, you will only need to run a single command through the DataGuard broker.
DGMGRL> switchover to 'standby';
However, if we were to use storage replication, we will have to perform the following. Do note that all these steps are manual unless you can script them to run automatically.
- Break mirrored LUN
- Mount LUN onto standby database
- Startup database (Database will perform automatic instance recovery)
Data Block Corruption Detection
DataGuard is able to detect block corruption automatically as long as you have the parameters configured in the SPFile.
Storage replication does not have such capability because it does not have visibility in the Oracle database. To the storage layer, everything is in OS data block.
Oracle DataGuard Rolling Upgrade
Oracle DataGuard has a feature called rolling upgrade. It is part of the Oracle Maximum Availability Architecture (MAA). The purpose of rolling upgrade is to reduce/prevent planned downtime.
The concept of rolling upgrading is simple;
- Patch the standby database first
- Test if the patch is working
- Switchover to standby database
- Upgrade primary database
- Switchback to primary database
- DBA saves the day!!
If you are a more paranoid DBA, you could do the following prior to the steps above to further test the patch/upgrade;
- Snapshot the primary database LUNs
- Mount the snapshot as a test database
- Run a test patch on the test database
- Destroy the test database
I am an Oracle DBA by nature and an Oracle DBA likes full control. By using DataGuard, I have absolute control over the DR solution. I will not have to depend on the storage engineer and system engineer during the DR exercise. This is precisely why ASM technology was created, to allow DBA to have control over the database filesystems. Why would you want to have multiple stakeholders in the DR exercise when you can have a single person to do everything (hint:DBA!).
Of course, all the reasons above are pertaining to Oracle database. Storage replication is still a major advantage to have in your DR solution arsenal. It is perfect for non-clustered application.
Today, I found out that I made an extremely silly mistake.
This was a very simple project. A 2-Node Oracle RAC setup connected to a IBM v7000 storage. The deployment was smooth. I configured the ASM disk using udev and device-mapper. However, after the system was handover to the customer, they begin encountering reboots and the following error.
qla2xxx [0000:1b:00.0]-801c:1: Abort command issued nexus=1:1:0 -- 1 2002.
These were the steps I took to resolve the issue.
- Logging of service ticket to IBM, Red Hat and QLogic.
- Flashing of QLogic HBA firmware
- Patching of QLogic HBA driver
However, the issue was still not resolved. The root cause was infact that the udev was configured incorrectly! The SCSI timeout was supposed to be set to 120s according to IBM documentation. This was a mistake that made me angry with myself. These were the few reasons for my mistake
- Check and verify all configurations again. (This was the part I did not perform. I was VERY confident that I configured it correctly)
- Do not assume and bet on the root cause so quickly. (I was too quick to assume to root cause to be the firmware/driver issues)
- Do not be sway by the customer no matter how experienced they are. Have your own understanding and judgement!
Oracle RAC 11gR2 – ORA-01102: cannot mount database in exclusive mode
In one of the customer site, I was unable to start both of the RAC instance at the same time. When 1 of the RAC instance was up, I couldn’t start the other RAC instance.
#sqlplus / as sysdba SQL> startup; ERROR at line 1: ORA-01102: cannot mount database in EXCLUSIVE mode
ORA-01102 error occurs when 2 separate (Not RAC configured) Oracle instance tries to open the same database. In one of the RAC instance, one of the parameter was configured wrongly
CLUSTER_DATABASE = FALSE;
SQL>alter system set cluster_database=TRUE scope=spfile sid='SID1'; SQL>shutdown immediate; SQL>startup
Hope this helps.