aix hacmp: B机datavg concurrent起不来

wengxg

两台p550做oracle rac，昨天异常断电，重启后：
A机hacmp，rac都能够正常启动
B机hacmp启动时提示：
#smitty clstart
WARNING: The HACMP timestamp file for shared volume group: datavg is inconsistent
with the time stamp in the VGDA for the following nodes: p550b
#lspv 发现datavg 没有concurrent

不知道下一步该怎么处理？
请教各位了。。。

wengxg

#smitty hacmp
system Management (C-SPOC)
HACMP Concurrent Logical Volume Management
Concurrent Volume Groups
Import a Concurrent Volume Group

不知道在是不是这里面从A机import Concurrent vg的信息？
有做过这些操作的朋友吗？
请赐教！

树袋熊

我记得是用cluster verify 然后系统会提示你是否FIX it, 选YES就搞定了,以前我碰到过一次,这个网上也有介绍的:
ERROR: The HACMP timestamp file for shared volume group: datavg is inconsistent
with the time stamp in the VGDA for the following nodes: test2
Starting Corrective Action: cl_resource_modify_vg_time_stamp.
___________________________________________________________________________________
Would you like to re-import shared VG: odps_datavg in resource group testvg3 to obtain
the file system:  on node: test2  [Yes / No]:
<01> Volume group: odps_datavg will not be imported onto node: test2

doudalun

楼上的，自动fix的话，需要hacmp5.4之后吧
在verify的选项中把一个自动修正的选项选成yes吧
手工的话，就是重新exportvg/importvg

树袋熊

我建议你用自动fix,以前我自己做过,没有问题.手工做也可以:
下面是有人建议的手工流程,很经典的:
http://portal.explico.de/en/foru ... d=353&Itemid=26


Re:How to synchronize VG timestamps?? 4 Months, 1 Week ago     
Lets make things a little bit more easier:

- HACMP 5.4.1 => No timestamp file => timestamp is compared to the value in the ODM
- No enhanced Concuurent Mode VG => Terms active/passive do not apply

To determine the real state, execute the following command on both nodes:

lsattr -El <vgname>

The output contains a value named timestamp => this is the value which HACMP will compare to the actual value on the VG.

Now, on the node where the VG is currently varied on, execute the following command an any disk contained in this VG, e.g. hdisk4:

lqueryvg -p hdisk4 -T

This command will read the timestamp directly from the disk. The value should be the same as given by the lsattr command.
If HACMP warns you about a timestamp Problem. the lsattr Command will show a different value on at least one of the nodes. To correct this problem, you need to use a command which reads the timestamp from the disk and writes it to the ODM => this is what synclvodm odm does. importvg -L should work as well as it will call synclvodm (this would be the better choice anyway as importvg is high level command).

In order to be able to call importvg -L, the disks of the VG need to be accessible. In non-concurremt environments, access is usually blocked through SCSI3 disk reservation.

Assuming node 1 has currently the VG online, and node 2 needs to execute importvg -L doe the follwoing:

on node 1: varyonvg -b -u -n <vgname> # this will break the reservation

on node 2: importvg -L <vgname> # this should update the time stamp in your ODM
on node 2: lsattr -El <vgname> # verify if the value has really changed

one node 1: restore the reservation with the following command: varyonvg -n <vgname>

What C-SPOC calls are some pretty complicated functions - not so easy to track if you are not really familiar to it. The easiest way to check what C-SPOC finally did, is to look in the logfile /tmp/cspoc.log or /var/hacmp/log/cspoc.log on the initiating node.

A final word:

As long as no mount is issued, no consistency problem can appear.

Frank

树袋熊

现在多数AIX系统是5.3/6.1,HA就是5.4/5.5其他的版本IBM应该不再提供支持了,单签的除外.

RS9000

HA异常停止，启动HA之前推荐先做同步校验。。。。