NAME

uncluster − uncluster a rootserver

SYNOPSIS

uncluster [ -i ] -p | -u [ -n ]

DESCRIPTION

This program will uncluster a rootserver’s file system by restoring all CDFs back to their standalone location, and by removing /etc/clusterconf and /etc/rbootd.log.  After running uncluster and rebooting, a system that was a cluster server will boot up and run as a standalone system.  Uncluster works on HP9000 Series 300, 400, 700 and 800 systems.  Uncluster uses the find(1) command to search for all CDFs on the system.  Each CDF is examined, using /bin/getcontext, in order to find the subfile that belongs to the rootserver.  For each CDF, the rootserver’s subfile is saved in a temporary area, the CDF is removed, then the saved subfile is placed into its standalone location with its standalone name. 
 
All actions taken by uncluster are recorded in the log file, /tmp/uncluster.log. 

Options

Uncluster is executable by the Super User with the following options:

-i This option will run uncluster in interactive mode.  The user will be prompted with any error information and will be asked whether or not to "Perform Task?".  The -i option must be accompanied with either the -p or -u option. 
 
Note that if the -i option is not specified, uncluster will not print out any error information and will not prompt the user to "Perform Task?". 

-p This option will put uncluster into preview mode.  In preview mode, uncluster displays actions that it would take on the file system.  These steps are recorded in the log file; however, no file system changes are made. 

-u This option will put uncluster into uncluster mode.  In this mode, uncluster will display actions that it is taking, and record these actions in the log file.  When uncluster has finished, the system is in an inconsistent state and will automatically be rebooted unless the -n option was specified. 

-n The "no reboot" option turns off uncluster’s automatic reboot capability.  The -n option can only be used with the -u option.  All kernel and file system changes will be in place after uncluster finishes.  When ready, the user will need to execute the reboot(1) command to get the system back into a consistent state. 

EXAMPLES

Uncluster preview (for rootserver only):  (No file system changes)

uncluster -i -p

Uncluster a rootserver:  (File system is modified)

uncluster -i -u

Uncluster a rootserver without rebooting:  (File system is modified)

uncluster -i -u -n

DEPENDENCIES

Parse_cdfs: Uncluster uses this program to parse the pathname of a CDF and return a standalone pathname.  Parse_cdfs must be present, must run on the current architecture, and must have a revision that matches uncluster’s revision.  To uncluster you must be:

- super user
- on the rootserver
- in single user mode (recommended, not enforced)
- in the root (/) directory

And you must have:

- /etc/parse_cdfs (proper revision)
- /etc/conf/gen (S800) or /etc/conf/dfile (S300 S400 or S700)

WARNINGS

Uncluster is NOT interruptible while executing in uncluster mode.  This is necessary to preserve the integrity of the file system.  Only preview mode is interruptible.  The log file may look slightly different when uncluster executes in uncluster mode as opposed to preview mode.  This is because nested CDFs require special processing as the file system is modified: Uncluster implements pathname adjusting and records this activity in the log file.  Uncluster will execute "mount -a -t hfs" to gain access to all of the file system. 

AUTHOR

Uncluster and parse_cdfs were developed by HP. 

FILES

/etc/uncluster

/etc/parse_cdfs

/etc/parse_cdfs.c

/tmp/uncluster.log

Hewlett-Packard Company  —  HP-UX Release 9.0: July 1992