HPUX 12.2.0.1 RAC异步IO问题处理

HPUX 12.2.0.1 RAC,数据库启动时实例恢复时报错：
2021-03-14T10:19:41.181674+08:00
Slave encountered ORA-27090 exception during crash recovery
2021-03-14T10:19:41.181741+08:00
Errors in file /oracle/app/oracle/diag/rdbms/orcl/orcl1/trace/orcl1_p006_6751.trc:
ORA-27090: Unable to reserve kernel resources for asynchronous disk I/O
Additional information: 2
设置系统层面的AIO后仍报错；
后来把数据库初始化参数：disk_asynch_io设置为FALSE不报错了。

fuser /dev/async，看到GRID的ASM进程使用了，如asm_o000_+ASM1、asm_gen0_+ASM1进程等；
fuser /dev/async也可以看到ORACLE用户也有oracl

参考：
https://docs.oracle.com/cd/B19306_01/server.102/b15658/appb_hpux.htm#UNXAR351
  HP-UX: ASM instance reports ORA-27090: Unable To Reserve Kernel Resources For Asynchronous Disk I/O (Doc ID 2307974.1)
  ORA-27090 Is Reported When PGA_AGGREGATE_LIMIT Is Set To 0 On HP-UX (Doc ID 2706572.1)
How to Disable Asynch_io on HP to Avoid Ioctl Async_config Error Errno = 1 or Errno = 11 (文档 ID 302801.1)
  
#########################################
1.数据库alert日志
2021-03-14T10:19:41.021942+08:00
Started redo scan
2021-03-14T10:19:41.174054+08:00
Completed redo scan
 read 503 KB redo, 6 data blocks need recovery
2021-03-14T10:19:41.181099+08:00
Slave encountered ORA-27090 exception during crash recovery
2021-03-14T10:19:41.181105+08:00
Slave encountered ORA-27090 exception during crash recovery
2021-03-14T10:19:41.181364+08:00
Recovery slave process is holding some recovery locks. Killing the instance now.
2021-03-14T10:19:41.181383+08:00
Recovery slave process is holding some recovery locks. Killing the instance now.
2021-03-14T10:19:41.181674+08:00
Slave encountered ORA-27090 exception during crash recovery
2021-03-14T10:19:41.181741+08:00
Errors in file /oracle/app/oracle/diag/rdbms/orcl/orcl1/trace/orcl1_p006_6751.trc:
ORA-27090: Unable to reserve kernel resources for asynchronous disk I/O
Additional information: 2
2021-03-14T10:19:41.181744+08:00
Errors in file /oracle/app/oracle/diag/rdbms/orcl/orcl1/trace/orcl1_p00c_6763.trc:
ORA-27090: Unable to reserve kernel resources for asynchronous disk I/O
--之后实例被终止terminating the instance due to error 27090

#########################################
#########################################
#########################################
2.数据库报错trace文件
*** 2021-03-14T15:55:35.392069+08:00
Ioctl ASYNC_ADDSEG error, errno = 7
Using largest hash table of 524288 entries (4Mb), htbl_size = 4194384, max_alloc
 = 1073729536
*** SESSION ID:(1706.35682) 2021-03-14T15:55:35.703699+08:00
*** SERVICE NAME:() 2021-03-14T15:55:35.703868+08:00
*** MODULE NAME:(oraagent.bin@L8LCDB1 (TNS V1-V3)) 2021-03-14T15:55:35.703887+08
:00
*** ACTION NAME:() 2021-03-14T15:55:35.703913+08:00
 
* kjb_start_recovery: pdb 0, passed in flags x5 (kjidomena 0) 
*                              CDB - CRASH RECOVERY 
2021-03-14 15:55:35.741 : Start recovery for domain=0, valid=0, flags=0x5, rcvin
st 1/32767/(1), endinc 2/0, pdb flags x204

*** 2021-03-14T15:55:38.626722+08:00
WARNING:1 Oracle process running out of OS kernelI/O resources request_aiolimit=
64 set_aiolimit=0 

*** 2021-03-14T15:55:38.646506+08:00
KCRP: blocks claimed = 1, eliminated = 0
----- Recovery Hash Table Statistics ---------
Hash table buckets = 524288
Longest hash chain = 1
Average hash chain = 1/1 = 1.0
Max compares per lookup = 0
Avg compares per lookup = 0/1 = 0.0
----------------------------------------------
Slave encountered ORA-27090 exception during crash recovery
Recovery slave process is holding some recovery locks. Killing the instance now.
ORA-27090: Unable to reserve kernel resources for asynchronous disk I/O
Additional information: 2

#########################################
#########################################
#########################################
3.HP-UX异步IO的配置检查
root@L8LCDB2[/]#cat /etc/privgroup
dba RTPRIO RTSCHED MLOCK
root@L8LCDB2[/]#/usr/sbin/setprivgrp -f /etc/privgroup
root@L8LCDB2[/]#/usr/bin/getprivgrp dba
dba: RTPRIO MLOCK RTSCHED 
--oinstall/asmadmin都进行了配置
root@L8LCDB2[/]#ls -al /dev/async
crw-rw----   1 grid       oinstall   101 0x000004 Mar 12 12:02 /dev/async


#########################################
#########################################
#########################################
4.检查fuser /dev/async查看进程
root@L8LCDB1[/]#fuser /dev/async
/dev/async:     3304o    3296o    3294o    3322o    3316o    3496o

e+ASM1_o000_orcl1、oracle+ASM1_tt00_orcl1这种连ASM的进程，但是没有数据库后台的DBWR LGWR等进程；