今天在测试rac 11g时发现vip不能自己飘过来,于是查了下些资料,做如下整理:
10.2.0.1 / 10.2.0.2 : There is not permanent network monitoring, so it is not possible knowing when the public interface is back again, therefore we need to do a manual fail back
More in detail, it was the instance ( every 10 minutes ) the one who tries to move the VIP back to the preferred node (original node) doing this:
Stop the VIP
Start up the VIP on the original node
So in case the network wasn't fully available this operation will fail causing a new fail-over
During this time there's no service and one instance should exists at least in the node to perform the check
You may have noticed this new resource:
This resource monitors the network constantly, so its not the instance anymore
In case the network fails, there's a new feature called "offline resource monitoring", so it keeps pinging every 60 seconds ( default ) the network to check if its back again
10.2.0.1 / 10.2.0.2 : There is not permanent network monitoring, so it is not possible knowing when the public interface is back again, therefore we need to do a manual fail back
crs_relocate10.2.0.3 : The VIP automatically fails back when the network is up again, however the implementation wasn't totally fine and there are some side-effects introduced.
More in detail, it was the instance ( every 10 minutes ) the one who tries to move the VIP back to the preferred node (original node) doing this:
Stop the VIP
Start up the VIP on the original node
So in case the network wasn't fully available this operation will fail causing a new fail-over
During this time there's no service and one instance should exists at least in the node to perform the check
10.2.0.4: There's no VIP automatic fail back anymore, however it is possible to emulate the 10.2.0.3 behavior uncommenting the line
ORA_RACG_VIP_FAILBACK=1 && export ORA_RACG_VIP_FAILBACK Location: racgwrap script @ ORACLE_HOME11gR2 Starting with 11.2, the failover is made trough the monitoring of a newly introduced network resource.
You may have noticed this new resource:
crsctl status resource -t
ora.net1.network ONLINE ONLINE rac1 ONLINE ONLINE rac2
This resource monitors the network constantly, so its not the instance anymore
In case the network fails, there's a new feature called "offline resource monitoring", so it keeps pinging every 60 seconds ( default ) the network to check if its back again
crsctl stat type ora.network.type -p ATTRIBUTE=OFFLINE_CHECK_INTERVAL DEFAULT_VALUE=60Finally, it is also possible to disable this feature by setting the OFFLINE_CHECK_INTERVAL in the network resource to zero by doing this
crsctl modify type ora.network.type -attr "ATTRIBUTE=OFFLINE_CHECK_INTERVAL ,TYPE='int',DEFAULT_VALUE=0"Please note that is not supported to edit the ora.* resources
来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/22969361/viewspace-1126726/,如需转载,请注明出处,否则将追究法律责任。
转载于:http://blog.itpub.net/22969361/viewspace-1126726/