Known workarounds with RamDisc and ServiceGuard...

There are a few ServiceGuard  commands that will scan all known LVM volume groups
configured (even those not included in the cluster)  while performing their tasks.
The command may result with an error indicating a problem with /dev/vr0X (the 'X'
matching a RamDisc volume group) or /dev/dsk/crtXd0 (RamDisc device files).  This
error occurs when the command attempts to gather physical disk device LVM data on
a RamDisc device file. Here is an example of errors from the cmcheckconf command:

# cmcheckconf -C ramcluster.ascii
Error: Unable to determine a unique identifier for physical volume /dev/dsk/crt8d0
on node cluster1. Use pvcreate to give the disk an identifier.

This error will occur while using the cmapplyconf command as well. The workaround
is to use the '-k' option during cmcheckconf and cmapplyconf commands.  Here is a
cut-and-paste from the cmcheckconf man page: the -k option means that cmcheckconf
only checks disk connectivity to the  LVM volume group that are identified in the
ASCII file. Omitting the -k option (the default behavior)  means that cmcheckconf
tests the connectivity of all  LVM volume  groups on all the cluster nodes. Using
-k can result in significantly faster operation of the command.  This option will
ignore any RamDisc volume group configured on the system and allow the command to
succede simply by adding '-k' before the '-C' option in cmcheckconf & cmapplyconf.

The cmgetconf command does not support the '-k' option. There is a program in the
/usr/local/etc/ramdisc/sysadmin directory called ramcluster_devices that corrects
this error while the cmgetconf command is executing. Incuding a block disk device
file that points to the same RAID LUN as an option to the ramcluster_devices will
create a temporary redirect for all  RamDisc device files  allowing the cmgetconf
command to succede. The ramcluster_devices script must be executed simultaneously
on each cluster node just before executing the  cmgetconf command (or just before
the ServiceGuard GUI utility performs the command).
_________________________________________________________________________________

Known workarounds with RamDisc and Dynamic Root Disks...

Some options of the drd command  attempt to scan all LVM volume groups configured
group(even the non-Root disk volume groups) while performing their tasks. The drd
command may error indicating a problem while "Reading Current System Information".
This type of error occurs when the command attempts to gather physical drive info
using a RamDisc device file. A simple modification of an ASCII script file called
/opt/drd/lbin/drd_save_config will provide the solution to the workaround.

Backup the /opt/drd/lbin/drd_save_config file.

Change the following two red lines into the blue text within the file:

vgdisplay 2>/dev/null | grep "^VG Name" | awk ' { print $3 } ' |

vgdisplay 2>/dev/null | grep -v "/dev/vr0" | grep "^VG Name" | awk ' { print $3 } ' |

sort -u $whole_disk_temp_file |

sort -u $whole_disk_temp_file | grep -v "/dev/dsk/crt" |

Save the file and now all DRD commands will ignore RamDisc devices.

Please note that the RamDisc workarounds with ServiceGuard as documented above can only be
performed if your current RamDisc software revision is 4.00 or greater.

	Other pages within the AppMate website... [Home]  [Software]  [Contact Us]  [References]  [Geek stuff]
	AppMate Performance Software 207 Long Pond Rd Danville, NH 03819	Tel: (603) 617-2833 Fax: (603) 617-2891 Tel: (888) 601-3069