About Oracle RAC ASM Raw Devices
When Not Using ASMLib
Before starting the CTE implementation, investigate how the customer is using raw devices for their ASM configuration.
Devices using Raw Bindings
Typically, a device that uses a raw binding looks like the following to ASM:
/dev/raw/raw1
If the device is mapped this way, you must locate where the mapping is performed. Typically, you can find this in the following configuration file:
/etc/sysconfig/rawdevices
The underlying binding could be to either a standard device name or a multipath I/O device name. Either way, you must find where the bind commands are run so that you can modify them for SecVM.
Note
If raw bindings are in use, then typically no changes are needed for the asm_diskstring. Because the binding to the actual device is created through the bind command, locate where the binding occurs and change the binding to SecVM.
Multipath I/O Devices
Devices using multipath I/O are typically found with the name:
/dev/mapper/mpath1
Generally, when using multipath I/O, you create SecVM on the multipath device name.
Note
-
CTE: If you use multipath I/O devices in the ASM configuration to add its disk, you must modify the
asm_diskstringparameter to include the/dev/secvm/dev/*path. -
CTE-U: If you use multipath I/O devices in the ASM configuration to add its disk, you must modify the
asm_diskstringparameter to include the/dev/secvm/*path.
Standard Devices
In many cases the ASM configuration may be using plain device names, like the following:
/dev/sda1
Note
-
CTE: If you use standard device names in the ASM configuration to add a disk, you must modify the
ASM_DISKSTRINGparameter to include the/dev/secvm/dev/*path. -
CTE-U: If you use standard device names in the ASM configuration to add a disk, you must modify the
ASM_DISKSTRINGparameter to include the/dev/secvm/*path.
Consistent Naming of Devices across RAC Nodes
As previously stated, if the raw device mappings for the disk(s) are NOT identical across all nodes in the RAC, then you CANNOT use a Host Group and you MUST apply the GuardPoints to each Host individually. This is typically NOT optimal, as a Host Group is the most effective way to manage an Oracle RAC environment.
