4.3.1 Managing shared disks
If servers share a disk (shared disk), HA Monitor uses OS commands to control connection with and switching of the shared disk. HA Monitor controls the switching of a shared disk in units of volume groups.
The following table shows the shared disk control methods and the formats of the commands used by HA Monitor.
Control method |
Format of command to be executed |
---|---|
Connection for referencing and updating |
vgchange -a y path-name-of-volume-group |
Disconnection |
vgchange -a n path-name-of-volume-group |
To enable use of the vgchange command with options for referencing and updating specified during connection, the vg_on_opt operand must have been specified in the server environment definition.
- Organization of this subsection
(1) Connecting
The kernel connects a shared disk.
When HA Monitor is connected to a shared disk, the kernel performs logical locking in units of volume groups so that the shared disk cannot be accessed from a standby server while the active server is using it.
The following figure shows the processing flow for connecting a shared disk.
(2) Switching
When a shared disk is switched, locking is performed in such a manner that the shared disk cannot be accessed from a server on a host resulting in a failure, but can be accessed from a server on a host that inherits job processing. When a volume group is disconnected, the fuser OS command is used to forcibly terminate all processes using the shared disk.
The processing flow for shared disk switching in the event of a server failure differs from that for host failures. The following figure shows the processing flow for shared disk switching in the event of a server failure.
The following figure shows the processing flow for shared disk switching in the event of a host failure.
(3) Log files
When volume groups are connected and disconnected, the execution results are output for each volume group with the file name volume-group-name.vglog under /opt/hitachi/HAmon/spool/. If an error occurs, check these log files to determine the cause of the error.
If the size of a log file exceeds 65,535 bytes when log data is acquired, HA Monitor creates a backup file named volume-group-name.vglog_old, clears the log file, and then collects log data again.