ReliantHA1.1.3a

Dear SCO Customer,

This ReliantHA® Software Supplement, ReliantHA1.1.3a, is a complete release of ReliantHA. It can be installed on UnixWare 7.1.3, 7.1.2 or 7.1.1 as well as OpenUnix 8. It can be used for a fresh install of ReliantHA or it can be used to upgrade a previous installation of ReliantHA 1.1.X.

New Features added in version 1.1.3a

The maximum length of the name of a mount point has been increased from 19 characters to 31 characters.

New Features added in version 1.1.3

Added a configuration utility to simplify the configuration process. The new script is /usr/opt/reliant/bin/rcu. Documentation for it will be available in the final product
Added a minimal sanity diagnostic tool for ReliantHA configurations. The new script is /usr/opt/reliant/bin/rdu. Documentation for it will be available in the final product.
Added a sample script utility that copies configuration files from one machine to all other nodes in a Reliant cluster. The new script is /usr/opt/reliant/bin/samples/hvcpconf. Documentation for it will be available in the final product.
When serial or other private links fail, a warning is now logged in the switchlog file.
The gab driver has been changed to allow the system administrator to run the gabconfig command to make the system panic when a gab_halt occurs, allowing a dump to be taken. The command and its options are:

/sbin/gabconfig -P [0|1]

A '1' enables the panic scenario, and a '0' disables it. It is disabled by default.

Oracle failover scripts capable of handling Oracle9i have been provided as sample scripts in the /usr/opt/reliant/samples directory. They have not yet been fully certified, so use them at your own risk. When certified, they will replace the Oracle scripts with the same names in /usr/opt/reliant, which are currently not Oracle9i-capable.
The Ip and Mount detector scripts have been improved, and now allow you to set the debug mode for each, as follows:

Desired Logging Behavior	-d flag in the Ip or Mount entry in /usr/opt/reliant/etc/rkind/names	.Ipdebug_min or .Mtdebug_min file in rkind directory
log failure messages and state changes only	set	exists
log all messages	set	removed
log minimum detector messages and don't log bm messages	unset	exists
log no messages	unset	removed

An experimental, improved Nw detector has been added as a sample; it should be used with care on production systems, or it may pose a security risk

The vulnerability is that rsh is invoked within the binary, and root permission in the .rhosts file on each node must be enabled for the experimental version to work properly. By doing this, you are allowing the root login on the system running the experimental version to log in to any node with root permissions.

To use the experimental Nw detector:

cp /usr/opt/reliant/bin/samples/Nw /usr/opt/reliant/bin/Nw

On each node, create a /.rhosts file that contains the system name (uname -s)

The experimental Nw detector, upon detection of a Network failure, will remotely check the ability of other nodes to access the network. If they are unable to access the network either, then the applications will stay in the current node. If the node to which the application is to move can successfully access the net, then the switchover will proceed as configured.

When none of the nodes are able to access the public network, execution of hvdisplay will show all applications (as well as the Nw Resource) to be "online". The Nwlog will log messages indicating that the network is inaccessible. The administrator must inspect the log, when the network access is questionable, before executing an hvswitch. To execute hvswitch under this condition will not restore the network and may result in an unnecessary switch.

Setting the debug level is the same as for Ip and Mount, except the resource name is Nw and the file is .Nwdebug_min.

Problems Fixed

Fixed from version 1.1.3:

A bug in the installation process caused the files hvipalias, hvenv and hvhosts to get overwritten with fresh versions. These files are now merged with the pre-existing versions correctly.

Fixed from version 1.1.2:

The example line in the hvipalias files has been updated.

A bug in the Process_Offline script has been corrected, so that the appropriate script processes get killed, and do not accidentally kill non-ReliantHA processes and shells. It also appropriately kills any children spawned by the ReliantHA process.

Fixed from version 1.1.1:

In some conditions a node's resources and applications may not have failed over when a complete system failure occurred.

There was no provision for handling LKP processes.

Oracle 8 failover did not proceed correctly in some circumstances. The Oracle scripts have been updated accordingly.

Sample configurations for the Messaging Server have been included in /usr/opt/reliant/build/samples/.

Under heavy load a system may have begun to run out of STREAMS resources and subsequently would miss a GAB halt sent from another node. In this case, two nodes might try to use the same shared disk which could result in data corruption. A fix has been made to cause a system to failover when it detects it is running out of STREAMS resources instead of trying to wait for a GAB halt from another node.

In some cases ReliantHA would hang during shutdown, thus preventing the node to shutdown completely.

The number of levels resources can be nested in a configuration file was limited to 16. This has been increased to 32.

Fixed from version 1.1.0:

Failover did not work across an RFC1821-compliant router. Gratuitous ARP capabilities have been added to make this possible.

If ReliantHA was not completely configured correctly, the system may hang on shutdown.

In some instances hvgen did not allow two distinct mount points to coexist in the same configuration file.

A race condition sometimes occurred while waiting for an IOCTL response when a device was closed.

Under heavy load a GAB halt may have occurred when it shouldn't. Timing in the llt driver has been modified to accommodate this.

The license manager may have reported strange licenses for ReliantHA.

The mounting of file systems may have failed if a process is already trying to use the file system's mount point. Mount_Online has been modified to check for this condition.

Extra unnecessary debug information was logged even if debugging was turned off.

During the installation of ReliantHA "mktitles" generated errors.


	/etc/conf/pack.d/gab/Driver.o
	/etc/conf/pack.d/llt/Driver.o
	/etc/conf/pack.d/msw/Driver.o
	/etc/conf/pack.d/sdl/Driver.o
	/etc/conf/pack.d/vmac/Driver_atup.o
	/etc/conf/pack.d/vmac/Driver_mp.o
	/etc/rc0.d/K14reliant
	/etc/rc0.d/K67msw
	/etc/rc1.d/K14reliant
	/etc/rc2.d/S92gab
	/sbin/gabconfig
	/sbin/lltconfig
	/sbin/lltstat
	/sbin/mswconfig
	/sbin/sdlconfig
	/sbin/sdlistty
	/usr/bin/vmacctl
	/usr/lib/Nlibgab.a
	/usr/lib/Nlibgab.so
	/usr/lib/scohelp/en_US.ISO8859-1/man/html.1Mha/CONTENTS.html
	/usr/lib/scohelp/en_US.ISO8859-1/man/html.4ha/CONTENTS.html
	/usr/opt/reliant/bin/DiskGroup
	/usr/opt/reliant/bin/File
	/usr/opt/reliant/bin/Ip
	/usr/opt/reliant/bin/LinuxInit
	/usr/opt/reliant/bin/LinuxInit_Offline
	/usr/opt/reliant/bin/LinuxInit_Online
	/usr/opt/reliant/bin/Mount
	/usr/opt/reliant/bin/Mount_Online
	/usr/opt/reliant/bin/Nw
	/usr/opt/reliant/bin/Oracle
	/usr/opt/reliant/bin/Oracle_Online
	/usr/opt/reliant/bin/Oracle_aux
	/usr/opt/reliant/bin/PFile
	/usr/opt/reliant/bin/Pdisk
	/usr/opt/reliant/bin/Process
	/usr/opt/reliant/bin/Process_Offline
	/usr/opt/reliant/bin/Share
	/usr/opt/reliant/bin/Volume
	/usr/opt/reliant/bin/bm
	/usr/opt/reliant/bin/chklic
	/usr/opt/reliant/bin/commdc
	/usr/opt/reliant/bin/commds
	/usr/opt/reliant/bin/generic_aux
	/usr/opt/reliant/bin/guiD
	/usr/opt/reliant/bin/hvassert
	/usr/opt/reliant/bin/hvdisp
	/usr/opt/reliant/bin/hvenv
	/usr/opt/reliant/bin/hvgen0
	/usr/opt/reliant/bin/hvifc
	/usr/opt/reliant/bin/hvrkind
	/usr/opt/reliant/bin/hvsend
	/usr/opt/reliant/bin/hvshut
	/usr/opt/reliant/bin/hvstart
	/usr/opt/reliant/bin/hvswitch
	/usr/opt/reliant/bin/hvutil
	/usr/opt/reliant/bin/ipalias_off
	/usr/opt/reliant/bin/ipalias_on
	/usr/opt/reliant/bin/keyfail
	/usr/opt/reliant/bin/mskx
	/usr/opt/reliant/bin/nodes/ckeys
	/usr/opt/reliant/bin/rcvm
	/usr/opt/reliant/bin/t_bm
	/usr/opt/reliant/bin/updfile
	/usr/opt/reliant/build/Ip.c
	/usr/opt/reliant/build/LinuxInit.c
	/usr/opt/reliant/build/Nw.c
	/usr/opt/reliant/build/Oracle.c
	/usr/opt/reliant/build/make_LinuxInit
	/usr/opt/reliant/build/samples/INSTALL.messaging
	/usr/opt/reliant/build/samples/grat_nfs.hl
	/usr/opt/reliant/build/samples/messaging-reliant-setup
	/usr/opt/reliant/build/samples/messaging.hl
	/usr/opt/reliant/conf/config
	/usr/opt/reliant/conf/reliant
	/usr/opt/reliant/etc/HelpTool
	/usr/opt/reliant/etc/templates/Resource.us
	/usr/opt/reliant/include/poststat.h
	/usr/opt/reliant/include/rkind.h
	/usr/opt/reliant/lib/detlib.a
	/usr/opt/reliant/lib/dettlib.a
	/usr/opt/reliant/lib/libmisc.a
	/var/opt/llt/dlpiping
	/var/opt/llt/llttest
	/var/opt/msw/floodin
	/var/opt/msw/floodout
	/var/opt/msw/setmac

Software Notes and Recommendations

ReliantHA1.1.3a should only be installed on:

UnixWare 7.1.3, 7.1.2 or 7.1.1
OpenUNIX 8.0.0

Installation Instructions

Download the ReliantHA1.1.3a file to the /tmp directory on your machine
As root, uncompress the file and add the package to your system using these commands:

	$ su
	Password: 
	# pkgadd -d /tmp/ReliantHA1.1.3a
	# rm /tmp/ReliantHA1.1.3a

Alternatively, this SLS package may be installed in quiet mode, that is, without displaying the release notes and asking for confirmation. To do this, use these commands:

	$ su
	Password
	# pkgadd -qd /tmp/ReliantHA1.1.3a all
	# rm /tmp/ReliantHA1.1.3a

Reboot the system after installing this package.

Removal Instructions

As root, remove the package using these commands:

	$ su
	Password: 
	# pkgrm ReliantHA

Reboot the system after removing this package.

Please NOTE:

Two files are modified as opposed to being replaced during upgrade to guard against the overwriting of any customization which might have been made by a system administrator. These are:

/usr/opt/reliant/etc/hvenv /usr/opt/reliant/etc/rkind/names

Because they are edited and not replaced, the changes made to these files will not be removed if the ReliantHA package is removed. However, the changes should not adversly affect the system if they remain after removing the ReliantHA package.

If you have questions regarding this supplement, or the product on which it is installed, please contact your software supplier.