UnixWare® 7 NonStop™ Clusters Release 7.1.1+IP is the second release of NonStop Clusters based on SCO's UnixWare 7 Release 7.1.1 operating system. The emphasis of this release is on the addition of the Open Cluster Interconnect with IP.
This README file supplies you with online images of the following documents for UnixWare 7 NonStop Clusters Release 7.1.1+IP:
For details of SCO products, updates and services, please refer to the Cover Letters accompanying your UnixWare 7 NonStop Clusters Release 7.1.1+IP media.
The Open Cluster Interconnect with IP feature supports the establishment of a two-node cluster using any interconnect (with restrictions, listed below) supporting the IP protocol. The choice between ServerNet and IP interconnection is made during installation. IP-based cluster operations are supported only in a two-node, co-located configuration.
IP protocol interconnect restrictions are:
The Fibre Channel Direct Disk Access by Pairs of Nodes feature allows direct access to FCAL disk enclosures from each node of a cluster. This allows the user to configure either node of a failover pair to directly access disks within a disk enclosure rather than forcing all disks to be directly accessible from only one node of the failover node pair. For FCAL disk subsystems, users can designate a preferred node for each disk enclosure. All filesystems within the disk enclosure are mounted on the preferred node whenever the preferred node is up. The filesystems failover to the specified backup node when the preferred node goes down due either to failure or to orderly shutdown. All filesystems within a disk enclosure are automatically remounted on the preferred node when the preferred node rejoins the cluster.
The GUI Configuration for File System Manager adds the following features to the SCOadmin File System Manager:
The System Availability Monitor (SAM) records and reports on the failures of a cluster and its individual nodes. Up and down events are recorded in a log file, which is subsequently analyzed by a report generator. Both graphical and command line interfaces are provided for user access to the report generator. The installed online documentation set provides a manual page, intro_sam(1M), that gives an overview of the operation and use of SAM.
The Keepalive Configuration Manager (KCM) provides an interface for creating sets of configuration files for the processes you want the keepalive subsystem to monitor. Dialogs provide fields for specifying the process node, favored and backup nodes, node selection policy for restarting the process, number of instances of the process to register, a process restart script, and the registration command.
NonStop Clusters licensing is now enforced. The kernels on all nodes now know how many licenses have been entered. Installation of more nodes than the current number of licenses will succeed, but having a node boot into the cluster without a license will result in error messages and the halting of the offending node. No enforcement will be in effect for the single root node, so that a single node cluster does not need a license.
IP aliasing is a feature that allows network interfaces to support more than one IP address. The IP Aliasing on CVIP feature provides support for adding aliases to the CVIP interface, thereby getting failover of those interfaces.
ftp://ftp.sco.com/SLSand check for any relevant NonStop Clusters Support Level Supplements (SLS). These have the prefix "nsc": the compressed binary is the file with the .Z suffix, and is accompanied by a .txt readme text file.
LANG=C export LANGbefore executing the following:
pkgadd -d cdrom1 -q NSC
UX:cp: ERROR: Cannot access /mnt/nscdoc/install/inst/locale/C/nscdoc.pkg: No such file or directory UX:cp: ERROR: Cannot access /mnt/nschelp/install/inst/locale/C/nschelp.pkg: No such file or directory UX:cp: ERROR: Cannot access /mnt/nscman/install/inst/locale/C/nscman.pkg: No such file or directory
The workaround is to update the OS only when all nodes are up and joined to the cluster, and node 1 is the root node. Otherwise, reinstall the dependent node that was down during the OS update.
The workaround is to run swap -a manually after dependent node installation, for swap areas greater than 2GB.
The workaround is to remirror the disks using the following command:
/sbin/vxconfigd -k /etc/vx/bin/vxreattach -r -b
The workaround is to reboot the cluster with node 1 and the root node.
The workaround is to fix any incorrect entries in /etc/cluster/fdevtab. Entries must be of the form:
fs(a) N:/dev/dsk/nNcxby*; M:/dev/dsk/nMcxby*
The workaround is that users with additional swap devices should edit the /etc/swaptab file, and change /dev/dsk/* devices to /dev/dsk/n?* devices.
When performing administrative functions on multi-ported disks, the system does not detect that the same physical device is being accessed simultaneously from multiple paths. This may result in filesystem corruption if you inadvertently cause the same disk to be accessed for write at the same time. The workaround is to avoid multiple simultaneous actions that cause writes to the same disk partition.
This problem occurs because all the interfaces for the default route's subnet are marked down. This is generally due to a cable being pulled or some other hardware failure outside the cluster. Before the routing problem can be corrected, one of the interfaces for the subnet must be in the up state. If the hardware issue is fixed, and a valid link is detected, the interface should be automatically marked as being back up in approximately 15 seconds.
Once this has taken place, look for the route add default command in /etc/inet/config, and run it by hand.
There is no known workaround.
The workaround is to kill the dlpid daemon, wait for keepalive to restart the daemon, and then reboot the backup root node.
The workaround is to set the "AUTOBOOT" entry in the /stand/<node>/boot file to greater than 120. 300 is recommended.
Please ensure that you are familiar with the procedure laid out in this document before attempting to install the UnixWare 7 NonStop Clusters Release 7.1.1+IP system.
This installation document will also be available as part of your SCOhelp system once the supplied UnixWare 7 NonStop Clusters software and documentation packages have been installed. Other documentation available through the SCOhelp interface includes:
If this release of sendmail is required, do the following after installing the NonStop Clusters packages and mounting the CD-ROM in /mnt:
pkgadd -d /mnt/NSC_Sendmail
Then install the package found there.
The Release Notes accompanying this image of sendmail are available in the PTF7130a: sendmail release 8.10.1 document.