Alert: A bug is happening that is tailored for version 12.1 on AIX
Inscription. Some users using Oracle Database version 12.1 (which includes 12.1.0.1 and 12.1.0.2 initial releases) have again encountered a 'bug specifically tailored for AIX' that has a significant impact and is again of concern.
A while back, we posted an article titled. A bug customized for version 11.2.0.4 on AIX is running high , many of you replied having encountered this bug and started making fixes. Recently, as the number of 12c users has gradually increased, problems in this version have been presenting themselves.
Some users using Oracle Database version 12.1 (which includes 12.1.0.1 and 12.1.0.2 initial releases) have again encountered a 'bug specifically tailored for AIX' that has a significant impact and is again of concern.
The bug number is.21915719
On MOS was raised to『 warning 'Level.
The title of this bug is.Database hang or may fail to OPEN in 12c IBM AIX or HPUX Itanium
That is: the database may hang on a 12c IBM AIX platform or Cannot be opened by OPEN , which later joined the potentially affected Anthem platform.
At the time of the problem, the database error was.
Deadlock or ORA-600 [kcrfrgv_nextlwn_scn] ORA-600 [krr_process_read_error_2]
Doc ID 1957710.1 ALERT: Bug 21915719 Database hang or may fail to OPEN in 12c IBM AIX or HPUX Itanium - ORA-742, DEADLOCK or ORA-600 [kcrfrgv_nextlwn_scn] ORA-600 [krr_process_read_error_2]
The causes of this problem are.
As you know, in Oracle 12.1, the database evolved the original single LGWR process into a master-slave process structure. LGWR one master process, working with up to 100 slave SLAVE processes, can reduce the original typical Log File Sync wait and increase the transaction rate and processing power of the database.
This new feature is described in detail in the 12.2 architecture diagram for Cloud and Enmore
"Follow this public page and reply to.122ARCH Electronic version available'
It is the introduction of this new feature that brings the risk of such a database not starting or having internal errors.
How to hedge this risk.
To start with how to circumvent this, Oracle has set switches for almost every new feature, which can be controlled by the implicit parameter _use_single_log_writer, and setting this parameter to TRUE returns you to the 11g model of using a single LGWR:.
ALTER SYSTEM SET "_use_single_log_writer"=TRUE SID='*' SCOPE=SPFILE;
note : If the database is already experiencing this bug, it's too late to change the parameters. Oracle has provided a patch fix for this bug. This issue is fixed in the PSU in 12.1.0.2.160419, so if you already have this PSU applied, there is no need to worry about it, otherwise it should be addressed head on.
In this bug, Oracle describes how it is possible to step into this big hole this way.
This question arises only when all of the following conditions are encountered.
Taken together, the above information shows that.
The reason for the hang is cited: the LGWR Slave processes need to receive the logs in order (ordering), but on the Power processor, for some reason the logs are not delivered in order, causing all the Slave processes to wait for a log batch of a certain ordering number, which causes the hang.
It's so important to use new features, understand them, and keep an eye on industry developments for databases.
There are always new problems when using new versions, so feel free to leave a comment and share your tales of treading water with us.