Specifications
Chapter 6 Troubleshooting Common Problems 6-15
NOTES:
■ You can run the add16lun or add32lun script that comes with RAID Manager
6.22. It will do all the steps needed for 16 or 32 LUNs support (rdriver.conf
gets modified).
■ Another new command, rdac_disks, cleans up the device tree so there is no
confusion between VERITAS Volume Manager device tree and the /dev/osa
device tree. If this step is omitted you will likely find phantom controllers, and
that lad and format using different path’s etc.
■ If you are using any version prior to VxVM 3.0.2 you must disable DMP as
described in the Sun StorEdge RAID Manager 6.22 Release Notes or FIN I0511-2.
■ It is very popular out in the field for people to run the following commands:
drvconfig, devlinks, disks,oradd_drv. This will result in multiple RDAC
links to a single device and the names will not agree with the RAID Manager path
names. This situation can be corrected by the execution of
/etc/raid/bin/rdac_disks. You can also look at the man pages for
rdac_disks for more information about this.
6.9 BootDelay (WhyBootingTakes SoLong)
Several things can be done to reduce delays in booting, especially reconfiguration
reboots and drvconfig calls which are done by rdac_disks (1M) or add_disk
(1M). drvconfig (1M) is also called when controllers are brought back online.
If you using A3500FC disk arrays only, edit the rmparams file and remove the sd:
from the end of the variable Rdac_NativeScsiDrivers:sd:ssd:.
Also, clean up the sd.conf file so the SCSI device discovery process doesn’t have to
explicitly access each potential device. This is explained in FIN I0551.
Sometimes the number of device instances is so large that the drivers are spending
time looking for non-existent devices. One way to clean up these instances is to use
disks -C or under Solaris 8, devfsadm -C. This should not be done if the host is
connected to controllers that are failed over, temporarily removed, or if the LUNs are
not properly balanced because the extra device links will be removed.
Another approach that has been taken for a couple of escalations is to remove the
line rdnexus scsi from /etc/driver_classes and replace class=scsi by
parent=rdnexus in rdriver.conf as described in RFE no. 4374861. However,
this approach has not been thoroughly tested.
Under Solaris 9, a delay of five to ten minutes has been reported. See bug 4630273.