Configuring SavWareHA

Before configuring SavWareHA, the following items must be installed and functional.

 
SCO TCP/IP

SCO TCP/IP must be installed on both machines. IP addresses and machine name s must be uniquely determined for each card on each machine (a good discussion of Internet addresses can be found in the SCO “TCP/IP Runtime System” manual ). Avnet Technology Solutions recommends using a 100BaseT or 10BaseT Ethernet point to point connection for the data link because of it’s ease of support.

Administration Recommendation: SCO assigns a unique system name to each network card in each system. It is recommended that the network card for the data link on the primary system be given the name “neptune” and the network card for the data link on the standby system be given the name “pluto”. This is not a requirement for SavWareHA to operate correctly but can help simplify the system administration process as well as help prevent the confusion in the role reversal situation that naming the systems “primary” and “standby” would have.

 
Host Equivalence

Each machine must allow the root account of the other machine to have trusted access . This is accomplished by adding entries for both machines into the files ‘/etc/hosts ’ and ‘/.rhosts ’. (Refer to the SCO “TCP/IP Configuration Guide” for more detail). Please note that ‘/.rhosts’ must be owned by root, have read-only set for user, group and other (use ‘chown root /.rhosts’ and ‘chmod 400 /.rhosts’ to accomplish this). Also, a password must be assigned to ‘root’ on each machine.

 
Primary / Standby Designation

One machine must be designated the Primary System (the other must then be designated the Standby System ). If both systems are of equivalent configuration and power, no special requirements are placed on the selection of the Primary System. However, if the two hosts are not equivalent in power or configuration, the more powerful (or more fully configured) host should be chosen as the primary.

 
Configuration Process

SavWareHA must be configured using the following procedure after installation is complete on both systems. It may be useful to refer to the SavWareHA configuration worksheet (see ‘Application Notes’ for this form) for an overview of the information that will be required during the configuration process.

1. Boot both machines into multi-user mode.

2. Mirroring a Unix Filesystem: Unmount the filesystems that will be mirrored on each machine. Use the ‘umount ’ command or the ‘sysadmsh ’ program (choose ‘filesystems’, then ‘unmount’ ). NOTE: All the data that might have existed on the Standby System ’s half of the mirror will be overwritten with the Primary System ’s data during the regeneration phase!

3. Mirroring a Unix Filesystem: Remove filesystem configuration entries.

SCO OpenServer5 and SCO Unix: Use the ‘mkdev fs’ command or the ‘sysadmsh ’ program (choose ‘Filesystems, then ‘Remove’; this allows you to remove existing filesystem definitions). The filesystem configuration entries are held in the files ‘/etc/default/filesys’ and ‘/etc/checklist’; these files may be edited by hand.

SCO UnixWare7: Use the Filesystem Manager in the Graphical Interface to Remove Mount Configuration for the disk that is to be mirrored. (The information is stored in /etc/vfstab.)

Mirroring a Raw Database Partition: The application using the raw database partition must be stopped. The database must also be shut down. Refer to the shutdown procedure in the database documentation. (After the configuration process is complete, the database application will need to be re-configured to access the device /dev/rmirror00 instead of the original SCO disk device).

4. Log in as root on the Primary System . Attempt to log into the Standby System with the ‘rlogin ’ command; if the ‘password:’ prompt is seen, then ‘/.rhosts’ has not been properly configured on the Standby System. Perform the same ‘rlogin’ command, but this time (since you are on the Standby System), attempt to log into the Primary System. The ‘password:’ prompt should not be seen here either. To sign off both systems, enter ‘exit’ twice or [CONTROL+D] twice; this will return your focus to the Primary System.

5. Give the ‘sentinel’ command. The SavWareHA program uses the arrow keys or space bar to move between selections and the [RETURN] key to make a selection. Use the [END ], [ESC] or [CONTROL-E] key to exit or cancel an operation.

6. When prompted, enter the system name for the network card of the data link on the primary system.

Administration Recommendation: It is recommended that the network card for the data link on the primary system be given the name “neptune”. This must be done with netconfig before SavWareHA is configured.

(Press [F3 ] for a listing of defined system names). This should be the same name as that used to ‘rlogin ’ from the Standby System to the Primary System.

NOTE: [F3 ] here passes through the ‘/etc/hosts’ file looking for valid addresses that can be reached via ‘ping’. This should serve as a further verification that the hosts are properly configured.

7. When prompted, enter the system name of the network card that is used for the data link on the standby system.

Administration Recommendation: It is recommended that the network card for the data link on the standby system be given the name “pluto”. This must be done with netconfig before SavWareHA is configured

(Press [F3 ] for a listing of defined system names). This should be the same name as that used to ‘rlogin ’ from the Primary System to the Standby System.

8. Choose the "Utilities " option and press [RETURN].

9. Choose the "Build a mirror " option and press [RETURN]. This will be non-destructive to the data on the Primary System ; but will overwrite all data on the Standby System ’s filesystem .

10. Enter the disk partition device name on the Primary System of the filesystem that will become the SavWareHA master mirror side (Press [F3 ] for a listing of all disk partitions on the Primary System.

NOTE: [F3 ] here passes through ‘/etc/conf/cf.d/mscsi’, and attempts to access devices that appear to be valid disk (Sdsk) entries. If the disks do not actually exist, or are not completely configured (via the ‘mkdev hd’ command), errors will be reported to the system console. Depending on the configuration of the system, these errors may be ignored.

11. Enter the disk partition device name on the Standby System that will contain the slave mirror side. (Press [F3 ] for a listing of all disk partitions on the Standby System). Any existing data in this disk partition will be overwritten with the Primary System ’s data !! This disk partition must be at least as large as the disk partition on the Primary System.

12. Mirroring Unix Filesystem: Enter the mount point for the filesystem (such as "/u"). This same mount point name will apply to both systems.

Mirroring Raw Database Partition: Enter “N” at the mount point prompt.

13. When prompted with "OK to continue?", double check the entries and press "y" to complete the mirror configuration. At this time, the regeneration process will begin. Both systems should show massive disk access activity, as will the data link .

14. (Versions of SavWareHA before 2.02c): If prompted "mkfs: Make a Unix Filesystem (y/n)": Answer "n" (no) if there is existing data in the disk partition on the Primary System . An answer of "y" to this question will DESTROY all data on the mirrored disk partitions on BOTH systems!!! Answer "y" (yes) if there was no filesystem in the disk partition on the Primary System. If in any doubt at all, entering ‘n’ is always the best choice.

15. Press [RETURN] to continue. The mirror generation will continue and mirrored access is now available to the users. The mirror is mounted automatically at this time on the specified mount point .

Mirroring Raw Database Partition: The database application must now be configured to access /dev/rmirror00 instead of the original SCO device (i.e. /dev/ru).

16. Exit the SavWareHA program by pressing the “q”, [END] , [ESC] or [CONTROL-E] key.

17. Test your application’s access to the mounted filesystem (s) on the Primary System . User applications will not be able to access the filesystem(s) on the Standby System unless SavWareHA is switched to fallback mode .

18. Configure the SavWareHA monitor program links as necessary. By default, the initial link is set to ‘net’. (See the section “Configuring the SavWareHA Monitor ” in this manual for more information).

19. Configure the terminals (‘stty’ settings, terminal types) as required for the application. Remember that this will have to be performed on both the Primary and Standby System s. See application note for S75inittab and mracdb for details on keeping the standby root information up to date.

20. Add installation-specific ‘Start’ and ‘Stop’ scripts to the directory ‘/etc/sentinel.d’. The scripts should follow the SCO default of ‘S##xxxx’ and ‘K##xxxx’. ‘S##xxxx’ is the startup script (‘S’), sequence number (‘##’), and descriptive name (‘xxxx’). ‘K##xxxx’ is the stop or kill script (‘K’), sequence number (‘##’), and descriptive name. Usually, one script is written to handle both start and stop, named ‘S##xxxx’, and is then linked with the ‘ln’ command to the ‘K##xxxx’ name.

21. Test the SavWareHA system as described in the “SavWareHA Checkout” section of this manual .