Configuring Fcas With The Oracle San Driver Stack - HP XP20000/XP24000 Configuration Manual

Table 20 Max throttle (queue depth) requirements for the devices (Solaris)
Queue depth option
Option 1
Option 2 (preferred)
NOTE: You can adjust the queue depth for the devices later as needed (within the specified
range) to optimize the I/O performance.
The required I/O time-out value (TOV) for devices is 60 seconds (default TOV=60). If the I/O TOV
has been changed from the default, change it back to 60 seconds by editing the sd_io_time
or ssd_io_time parameter in the /etc/system file.
You may also need to set several other parameters (such as FC fibre support). See the user
documentation that came with your FCA to determine whether other options are necessary to meet
your operational requirements.
NOTE: Use the same settings and device parameters for all systems. For Fibre Channel, the
settings in the system file apply to the entire system, not to just the FCAs.
To set the I/O TOV and queue depth for the devices:
1. Make a backup of the /etc/system file: cp /etc/system /etc/system.old.
2. To assure you use the default TOV, make sure that no sd_io_time values are set in the
/etc/system file or modify the /etc/system file to show the following values:
set sd:sd_io_time=0x3c
set ssd:ssd_io_time=0x3c for Oracle generic FCA
3. To set the queue depth, add the following to the /etc/system file:
set sd:sd_max_throttle = x
set ssd:ssd_max_throttle = x for Oracle generic FCA
(for x value, see
Example:
set sd:sd_max_throttle = 16
set ssd:ssd_max_throttle = 16

Configuring FCAs with the Oracle SAN driver stack

Oracle branded FCAs are only supported with the Oracle SAN driver stack. The Oracle SAN
driver stack also supports current Emulex and QLogic FCAs.
92 Solaris
Table 20 (page 92))
<— Add this line to /etc/system
<— Add this line to /etc/system (for Oracle generic FCA)
Requirements
XP10000, XP12000, SVS 200: Queue_depth
default.
XP20000, XP24000: Queue_depth
CAUTION: The number of issued commands must be
completely controlled. Because queuing capacity of the
disk array is either 1024 or 2048 per port (depending on
the disk array), you must adjust the number of issued
commands from Solaris system to less than 1024 or 2048.
Otherwise, memory allocate failed messages may occur
on the Solaris system, and all read/write I/O may stop,
causing the system to hang.
XP10000, XP12000, SVS 200: Number of LUs x
queue_depth 1024 and queue_depth
XP20000, XP24000: Number of LUs x queue_depth
2048 and queue_depth
1024
2048 default.
32
32