Disk and disk adapter tunable parameters

The AIX operating system includes disk and disk adapter kernel tunable parameters.

num_cmd_elem: Disk Adapter Outstanding-Requests Limit
Item Description
Purpose Specifies the maximum number of requests on a Small Computer System Interface (SCSI) bus that is applicable only to the SCSI-2 Fast or Wide adapter.
Values The values range from 40 to 128. The default value is 40.
Display To display the value of the num_cmd_elem tunable parameter, enter the following command:
lsattr -E -l scsin -a num_cmd_elems
Change To change the value of the num_cmd_elem tunable parameter, enter the following command:
chdev -l scsin -a num_cmd_elems=NewValue

The change is effective immediately and is permanent. If the -T flag of the chdev command is used, the change is immediate and lasts until the next boot. If the -P flag of the chdev command is used, the change is deferred until the next boot and is permanent.

Diagnosis Applications that perform large writes to striped raw logical volumes do not achieve the required throughput rate.
Tuning The value must be equal to the number of physical drives, which are multiplied by the queue depth of the individual drives. The number of physical drives must include the drivers in the disk arrays on the SCSI bus.
num_cmd_elems: Fibre Channel Adapter Outstanding-Requests Limit
Item Description
Purpose Specifies the maximum number of requests on a Small Computer System Interface (SCSI) bus that is applicable only to the SCSI-2 Fast or Wide adapter.
Values The values range from 40 to 128. The default value is 40.
Display To display the value of the num_cmd_elem tunable parameter, enter the following command:
lsattr -E -l scsin -a num_cmd_elems
Change To change the value of the num_cmd_elem tunable parameter, enter the following command:
chdev -l scsin -a num_cmd_elems=NewValue

The change is effective immediately and is permanent. If the -T flag of the chdev command is used, the change is immediate and lasts until the next boot. If the -P flag of the chdev command is used, the change is deferred until the next boot and is permanent.

Diagnosis Applications that perform large writes to striped raw logical volumes do not achieve the required throughput rate.
Tuning The value must be equal to the number of physical drives, which are multiplied by the queue depth of the individual drives. The number of physical drives must include the drivers in the disk arrays on the SCSI bus.
queue_depth: Disk Drive Queue Depth
Item Description
Purpose Specifies the maximum number of requests the disk device can store in its queue.
Values The manufacturer sets the range of values. The default value of IBM disks is 3 and non-IBM disks is 0.
Display To display the value of the queue_depth tunable parameter, enter the following command:
lsattr -E -l hdiskn
Change To change the value of the queue_depth tunable parameter, enter the following command:
chdev -l hdiskn -a q_type=simple -a queue_depth=NewValue

The change is effective immediately and is permanent. If the -T flag of the chdev command is used, the change is immediate and lasts until the next boot. If the -P flag of the chdev command is used, the change is deferred until the next boot and is permanent.

Diagnosis Not applicable.
Tuning If the non-IBM disk drive is capable of request-queuing, perform the change to ensure that the operating system takes advantage of the capability.

For more information, see Setting SCSI-Adapter and disk-device queue limits.

fast_lnk_recov: To control quick I/O failure for a Fibre Channel
Item Description
Purpose Specifies the maximum number of pending requests in a Fibre Channel adapter.
Values The values range from 200 to 4096. The default value is 1024.
Display To display the value of the num_cmd_elems tunable parameter, enter the following command:
lsattr -E -l fcsn -a num_cmd_elems
Change To change the value of the num_cmd_elems tunable parameter, enter the following command:
chdev -l fcsn -a num_cmd_elems=NewValue

To change this attribute, the fcsn adapter, where n is a whole number, must be in a defined state. Otherwise, the -P flag of the chdev command is used to change the attribute. The -P flag defers the change until the next boot operation and the change is permanent.

For Fibre Channel adapters with data rates of 16 Gbps or more, the fcsn adapter does not need to be in a defined state. You can update the num_cmd_elems attribute by using the following command:
chdev -U -l fcsn -a num_cmd_elems=NewValue
Note: The default value and the range value vary for each Fibre Channel device. For some Fibre Channel and Fibre Channel over Ethernet (FC/FCoE) adapters, the maximum value of the num_cmd_elems parameter that can be set might be less than the maximum range mentioned in the Object Data Manager (ODM). If the specified value for the num_cmd_elems parameter of the chdev command is larger than the value supported by the FC/FCoE adapters, an error message is logged for these adapters.
Tuning To get optimum performance, set the value of the num_cmd_elems parameter to the maximum supported range.
queue_depth: Disk Drive Queue Depth
Item Description
Purpose Specifies the maximum number of requests the disk device can store in its queue.
Values The manufacturer sets the range of values. The default value of IBM disks is 3 and non-IBM® disks is 0.
Display To display the value of the queue_depth tunable parameter, enter the following command:
lsattr -E -l hdiskn
Change To change the value of the queue_depth tunable parameter, enter the following command:
chdev -l hdiskn -a q_type=simple -a queue_depth=NewValue

The change is effective immediately and is permanent. If the -T flag of the chdev cpmmand is used, the change is immediate and lasts until the next boot. If the -P flag of the chdev command is used, the change is deferred until the next boot and is permanent.

Diagnosis Not applicable.
Tuning If the non-IBM disk drive is capable of request-queuing, perform the change to ensure that the operating system takes advantage of the capability.

For more information, see Setting SCSI-Adapter and disk-device queue limits.

fast_lnk_recov: To control quick I/O failure for a Fibre Channel
Item Description
Purpose Controls quick I/O failure for a Fibre Channel link failure case, for Fibre Channel adapters with data rates equal to or more than 16 Gbps.
Values The following values are supported:
Yes
The Fibre Channel SCSI protocol driver quickly fails over I/O operations to the SCSI device driver when the host or storage link is down. This failover operation helps the SCSI device driver to switch I/O operations to alternative paths. This value is the default value.
No
Disables the quick I/O failure for FC link failure attribute.
Display To display the value of the fast_lnk_recov tunable parameter, enter the following command:
lsattr -E -l fscsin -a fast_lnk_recov
Change To change the value of the fast_lnk_recov tunable parameter, enter the following command:
chdev -a fast_lnk_recov=yes -l fscsin -U
Tuning This tunable parameter is used only for a Multi Path I/O (MPIO) environment where another path to fail over exists. As a best practice, you must have an alternative path through another Fibre Channel Host Bus Adapter (HBA) to handle HBA hardware failure scenarios. In a non-MPIO environment, it is recommended not to change the default settings.
io_64k_dma_page: 64 KB DMA page size for storage I/O
Item Description
Purpose Enables and disables the use of 64 KB direct memory access (DMA) page size by the storage I/O.
Values The following values are supported:
disabled
Storage I/O does not use the 64 KB DMA page size. This value is the default value.
enabled
Storage I/O uses the 64 KB DMA page size.
Note: Use the io_64k_dma_page tunable parameter for all the secondary devices of a Peripheral Component Interconnect (PCI) device. If this tunable parameter is disabled, the storage I/O does not request the 64 KB DMA page size. However, if the storage I/O requests the 64 KB DMA page size when this tunable parameter is disabled, the PCI bus driver tries to allocate the 64 KB DMA page size to the storage I/O.
Display To display the value of the io_64k_dma_page tunable parameter, enter the following command:
lsattr -EHl pcin
where n is a whole number.
Change To change the value of the io_64k_dma_page tunable parameter, enter the following command:
chdev -1 pcin -a io_64k_dma_page=disabled|enabled

To change the value before the next boot operation, the PCI device must be in the Defined state. To change the io_64k_dma_page tunable parameter when the PCI device is not in the Defined state, use the -P flag of the chdev command. The -P flag does not change the io_64k_dma_page tunable parameter value until the next boot operation. The change in the io_64k_dma_page tunable parameter after the boot operation is permanent.

Tuning The io_64k_dma_page tunable parameter is disabled by default. If the total 4 KB page size DMA requirement for all the devices that are included in a PCI device is less than 2 GB, then you can enable the io_64k_dma_page tunable parameter.
start of changemt_qk_io_recov: To control quick I/O failover for a Fibre Channelend of change
start of change
Item Description
Purpose

Controls quick SCSI I/O failover to redundant Fibre Channel (FC) Host Bus Adapter (HBA) during maintenance operations and error recovery scenarios for the FC adapters with data rates equal to or more than 16 Gbps.

Values The following values are supported:
Yes

The SCSI disk driver quickly fails over I/O operations to alternative Multi Path I/O (MPIO) paths during Firmware update operation, when PCIe bus level errors are encountered on FC HBA. This value is the default value.

No

Disables the quick SCSI I/O failover to alternative MPIO paths during mentioned scenarios.

Change
To change the value of the mt_qk_io_recov tunable parameter, enter the following command:
chdev -a mt_qk_io_recov=no -l fscsin -U
Tuning

This tunable parameter is used only for a MPIO environment where another path to fail over from alternate FC HBA exists. As a best practice, you must have an alternative path through another FC HBA to handle HBA hardware failure scenarios.

end of change