Oracle dataguard日志应用异步io问题

客户一数据库dg应用出问题，运行在hp 11.11 pa-risk平台，后台启动mrp进程时，alert日志显示如下错误：

引用

Attempt to start background Managed Standby Recovery process (zjv8)
MRP0 started with pid=22, OS id=3223
Tue Dec 14 09:04:11 2010
MRP0: Background Managed Standby Recovery process started (zjv8)
Managed Standby Recovery not using Real Time Apply
parallel recovery started with 5 processes
Tue Dec 14 09:04:16 2010
Errors in file /oracle/admin/zjv8/bdump/zjv8_ora_2114.trc:
Tue Dec 14 09:04:16 2010
Errors in file /oracle/admin/zjv8/bdump/zjv8_ora_2116.trc:
Tue Dec 14 09:04:16 2010
Errors in file /oracle/admin/zjv8/bdump/zjv8_ora_2118.trc:
Tue Dec 14 09:04:16 2010
Warning: OS async I/O limit 128 is lower than recovery batch 1024
Tue Dec 14 09:04:16 2010
Errors in file /oracle/admin/zjv8/bdump/zjv8_ora_2122.trc:
Tue Dec 14 09:04:16 2010
Warning: OS async I/O limit 128 is lower than recovery batch 1024
Tue Dec 14 09:04:16 2010
Warning: OS async I/O limit 128 is lower than recovery batch 1024
Tue Dec 14 09:04:16 2010
Errors in file /oracle/admin/zjv8/bdump/zjv8_ora_2120.trc:
Tue Dec 14 09:04:16 2010
Warning: OS async I/O limit 128 is lower than recovery batch 1024
Tue Dec 14 09:04:16 2010
Warning: OS async I/O limit 128 is lower than recovery batch 1024
Tue Dec 14 09:04:17 2010
Waiting for all non-current ORLs to be archived..

metalink给出相关解释：

引用

Applies to:
Oracle Server - Enterprise Edition - Version: 10.2.0.1 to 11.1.0.8
This problem can occur on any platform.
Oracle Server Enterprise Edition - Version: 10.2.0.1 to 11.1.0.8
Symptoms

A warning message in Alert log asynchronous I/O limit warning of the form: "Warning: OS async I/O limit 128 is lower than recovery batch 1024" may be signaled in the alert log by recovery slave or coordinator processes.This error is because of kernel has not been configured sufficiently for async IO operations.

Operations will still continue but IO operations will not be performed asynchronously which may impact recovery performance and will generate more trace files.

NOTE : HP-UX enables Oracle Database to perform I/O to (only) raw disk partitions using an asynchronous method
Cause
This warning is signaled on standby when an Oracle process cannot reserve enough async IO descriptors during instance or media recovery.
Solution

1. For more details about how to tune ASYNC I/O for Raw disk partitions on HP_UX kindly refer,

Oracle Database Administrator's Reference
11g Release 1 (11.1) for Linux and UNIX-Based Operating Systems
Part Number B32009-05
Appendix B 'Administering Oracle Database on HP-UX'
section 'Asynchronous Input-Output'

For Oracle 10G,

http://download.oracle.com/docs/cd/B19306_01/server.102/b15658/appb_hpux.htm#i637036

For Oracle 11G,

http://download.oracle.com/docs/cd/B28359_01/server.111/b32009/appb_hpux.htm#i637036


NOTE : The above document can be given as an 'example' of what to check for HP. For other platforms, check the relevant section in the same doc OR refer back to the OS/hardware vendors for further guidance.


2. For File system to improve the performance,

Set disk_asynch_io = FALSE and increase the DB writer process.

Based upon the number of CPUs and the number of processor groups, Oracle either selects an
appropriate default setting for DB_WRITER_PROCESSES or adjusts a user-specified setting.

NOTE : Make sure dbwr_io_slaves set to 0.

Example for 4 CPU machine,

SQL>alter system set dbwr_io_slaves = 0 scope = spfile ;
SQL>alter system set db_writer_processes = 4 scope = spfile ;

Sometime the false Warning message will be displayed in alert log eventhough the DISK_AYNC_IO is set to FALSE.

Inorder to stop this Warning message we have to apply this Patch 8475825 which is available for most of the OS.

Reference Bug 8475825 - OS ASYNC I/O LIMIT 128 IS LOWER THAN RECOVERY BATCH 1024