7.5.5 Handling shared resource disconnection errors
This subsection explains the handling of shared resource disconnection errors. Normally, HA Monitor disconnects shared resources. In the event of a failure, the operator must disconnect the shared resources manually. The action to be taken depends on when the shared resource disconnection error occurred.
- Important
-
If abnormal or forced termination of a server or a shared resource inheritance timeout is detected during shared-resource control processing, the status of the shared resources becomes undefined. Therefore, you must check the status of the shared resources before restarting the server. For other notes about manipulating shared resources, see 7.2.3 Notes about maintaining shared resources. If you do not observe these notes, the shared resources might become corrupted or hot standby processing might fail.
- Organization of this subsection
(1) During hot standby processing
If a shared resource disconnection error occurs in the active system during hot standby processing, HA Monitor switches over to the standby system. If a host reset is to be performed, HA Monitor resets the host and then performs hot standby processing. If SCSI reservation for shared disk is to be used, the host on which the shared resource disconnection error occurred issues a kernel panic to terminate the OS and then performs hot standby processing. No operator intervention is needed.
If a shared resource disconnection error occurs in the active system during hot standby processing, contention might occur for shared resources between hosts, resulting in corruption of the shared resources and system shutdown. To avoid contention for shared resources between both hosts, if a host reset is to be performed, HA Monitor resets the host, and, if SCSI reservation for shared disk is to be used, the host on which the shared resource disconnection error occurred issues a kernel panic to terminate the OS.
(2) During server termination processing
If a shared resource disconnection error occurs during server termination processing, a message to that effect is issued. The operator must check the contents of the message and perform the disconnection processing that is appropriate to each shared resource.
-
Shared disk
Execute the following command:
vgchange -a n path-name-of-volume-group
-
File system
Perform the following procedure:
-
Use the umount OS command to unmount the file system.
-
If an attempt to unmount the file system fails, use the fuser OS command to forcibly terminate all processes that are using the file system, and then retry unmounting the file system.
-
-
LAN
To manually disconnect a LAN, you must correct the LAN status settings file and then execute it.
The file to be corrected is /opt/hitachi/HAmon/etc/server-alias-name.down.
You can also use the ifconfig OS command to disconnect a LAN.