Using your laptop or a customer ASCII terminal, log on to each OPS host.

For the first OPS host, enter the command nvram. The output should contain the line

scsihostid=

This output means that the SCSI host ID for this OPS host is 0.

For the second OPS host, repeat step 2 in the second window. The output should contain the line

scsihostid=2

This output means that the SCSI host ID for the second OPS host is 2.

If the customer has purchased a new workstation for use with OPS, install its software and configure it.

From the distribution CDs included in the IRISconsole kit, load the IRISconsole software as explained in Chapter 2 of the IRISconsole Multiplexer Installation Guide.

Use the IRISconsole GUI to set up the customer's site, which is the IRISconsole grouping of the two OPS hosts for administration purposes: as root; type /usr/sbin/ic. The IRISconsole main window appears.

In the Edit menu, select "Add New Site." The Change IRISconsole Config dialog box appears.

In the dialog box, type a name for the site, such as OPS. Click the Apply button; select "Save IRISconsole Config" in the File menu and close the dialog box.

To add the systems to the site, double-click on the site icon in the main menu. The site window appears.

In the site window, select "Add New System..." in the Edit menu. The Change Site Config dialog box appears, as shown in Figure 3-1.

Figure 3-1. Change Site Config Dialog Box

Enter information for the first OPS server. If necessary, check the multiplexer and the cables for the ports to enter. Ask the customer for the correct entry for the User: field. When the information for the first OPS server is correct, click the Apply button.

Click Reset to clear the fields and enter the information for the second OPS server. When the information for the second OPS server is correct, click Apply. Click Close to exit the dialog box.

Save the configuration by selecting "Save System Config" in the site window File menu. If the customer desires, save the configuration to a particular name (Save As). Exit the IRISconsole GUI.

Enter the fx command with appropriate parameters; for example:

fx -x dksc(6,6,2)

Output such as the following appears:

fx version 5.3, Jan 3, 1995
...opening dksc(6,6,2)
...controller test...OK
Scsi drive type == SGI RAID 5 0767
fx: Warning: bad sgilabel on disk

creating new sgilabel

----- please choose one (? for help, .. to quit this menu)-----
[exi]t             [d]ebug/           [l]abel/           [a]uto
[b]adblock/        [exe]rcise/        [r]epartition/     [f]ormat

Update parameters; at the fx> prompt, type the following:

fx> label/set/param

Output such as the following appears:

fx/label/set/parameters: Error correction = (enabled) 
fx/label/set/parameters: Data transfer on error = (enabled) 
fx/label/set/parameters: Report recovered errors = (enabled) 
fx/label/set/parameters: Delay for error recovery = (enabled) 
fx/label/set/parameters: Err retry count = (0) 
fx/label/set/parameters: Transfer of bad data blocks = (enabled) 
fx/label/set/parameters: Auto bad block reallocation (write) = (enabled) 
fx/label/set/parameters: Auto bad block reallocation (read) = (enabled) 
fx/label/set/parameters: Read ahead caching = (disabled) 

At the Enable CTQ prompt, type enable:

fx/label/set/parameters: Enable CTQ = (disabled) enable

At the CTQ depth prompt, type 12:

fx/label/set/parameters: CTQ depth = (2) 12 

Output such as the following appears.

fx/label/set/parameters: Read buffer ratio = (0/256) 
fx/label/set/parameters: Write buffer ratio = (0/256) 
 * * * * * W A R N I N G * * * * *
about to modify drive parameters on disk dksc(6,6,2)! ok?

At the prompt in the last line in step 4, type yes.

The following output appears:

----- please choose one (? for help, .. to quit this menu)-----
[exi]t             [d]ebug/           [l]abel/           [a]uto
[b]adblock/        [exe]rcise/        [r]epartition/     [f]ormat

To repartition the disk, type r.

Type exit to exit fx. The following message appears:

label info has changed for disk dksc(6,6,2). write out changes? (yes) 

Type y to write the changes to disk.

If there are more than ten volumes, you must edit master.d/xlv (see "Configuring the System for More Than Ten XLV Logical Volumes" in Chapter 7 of the IRIX Admin: Disks and Filesystems volume) and regenerate the kernel using autoconfig(1M).

Stop the RAID agent:

/etc/init.d/raid5 stop 

Use fx to create partitions for the log files on the volumes.

Create an xlv_make file for the volumes in all partitions.

Run xlv_make to create the volumes.

Edit /etc/init.d/xlv to add lines that change the ownership of the Oracle volumes to "oracle." Look for these commented lines:

# change ownership of xlv volumes here:
# chown $USER.$GROUP /dev/dsk/xlv/<volname>
# chown $USER.$GROUP /dev/rdsk/xlv/<volname>

When the volumes are created, change the owner to oracle.

---

Note: On every reboot, xlv recreates the volumes; they will be owned by root.sys, and the names will be different on each node.

---

Change the node name for all XLV volumes to be used by the OPS software so that the full XLV volume name is the same on both nodes.

When XLV creates volumes, it inserts a node name element in each volume name. For example, if volume finance1 is created on node ops1, XLV names the volume ops1.finance1. On node ops1, the device name for the volume appears as the matching volume /dev/dsk/xlv/finance1, but on node ops2, the full XLV volume name appears: /dev/dsk/xlv/ops1.finance1. The volume name has this form because XLV on ops2 sees that the volume finance1 was built on another server (ops1).

However, OPS requires that these shared volumes have exactly the same name. You must rename the volumes manually; do not use a node name. Give the shared volumes the same names on each node.Table 3-1 gives examples.

Restart the RAID agent:

/etc/init.d/raid5 start 

Log in on each server as the Oracle software owner.

Set environment variables:

Use systune -i to increase the number of semaphores in the kernel:

semaem = 16384 (0x4000)
semvmx = 32767 (0x7fff)
semume = 10 (0xa)
semopm = 200 (0xc8)
semmsl = 200 (0xc8)
semmnu = 100 (0x64)
semmns = 200 (0xc8)
semmni = 100 (0x64)

If necessary, adjust the shared memory segment size using systune.

Run the OPS install script as follows:

% cd $ORACLE_HOME/pso
% ./psoinst.sh

The psoinst.sh script does three things:

During installation, the script outputs messages informing you of the installation status. When the installation has finished, the following message appears:

Finished installing Oracle Parallel Server

If the message shown above does not appear, correct any errors and enter

% ./copyoh.sh 

The file osh is owned by root and is not copied automatically.

---

Note: If the message rcp: osh: Permission denied appears, ignore it.

---

To put the file osh on the second Challenge server, log in as root and copy this file, as follows:

# rcp &ORACLE_HOME/bin/osh \
   $REMOTE_NODE:$REMOTE_ORACLE_HOME/bin
# rsh &REMOTE_NODE chmod 4711 $REMOTE_ORACLE_HOME/bin/osh

If you encounter errors, make sure that the user equivalence is set up for root, or execute the following steps as root:

# cd $ORACLE_HOME/bin
# ftp $REMOTE_NODE

Enter the username root and the root user password.

ftp> cd $REMOTE_ORACLE_HOME/bin
ftp> bin
ftp> put osh
ftp> quit
# rlogin $REMOTE_NODE -l root
Password: password
# chmod 4711 $REMOTE_ORACLE_HOME/bin/osh

Install any optional software that the customer has purchased.

As root on the workstation, start the OPS node controller software by entering

/usr/opsnc/bin/opsnc

Start IRISconsole by entering

/usr/sbin/ic

In the IRISconsole site window, click the GetConsole button for each OPS host.

Check for the presence of the CM lock file in /tmp. This filename has the format .nn, where each digit stands for DLM domain and instance, for example, .00. If this file exists, delete it.

Check if the CM is already running by entering

ps -ef | grep opscm

If it is running, type the following to kill it:

killall -TERM opscm

Check if the DLM is already running by entering

ps -ef | grep dlm

If the DLM is running, enter the following to kill its processes:

killall -TERM dlmmon
killall -TERM dlmd

Run ipcs to determine the shared memory and semaphores used on the host. The following is an example output:

IPC status from /dev/kmem as of Thu May 18 14:31:22 1995
T     ID     KEY        MODE       OWNER    GROUP
Message Queues:
Shared Memory:
m      0 0x53637444 --rw-r--r--     root      sys
m    301 0x000022bb --rw-rw----   oracle      dba
m   2202 0x0c33b7c9 --rw-r-----   oracle      dba|
Semaphores:
s   2200 0x00000000 --ra-r-----   oracle      dba

If Oracle or DLM is using any shared memory segments or semaphores, save them to another location if you need them for debugging a DLM or Oracle crash; otherwise, delete them with ipcrm. In the example in step 7, you would use

ipcrm -m 301 -m 2202 -s 2200

For each host, set environment variables in the /etc/init.d/opscmgr script:

ORACLE_HOME=/usr/people/oracle
ORACLE_SID=sidname
LKDOM=0
LKINST=0
USER=oracle
GROUP=dba

export ORACLE_HOME ORACLE_SID LKDOM LKINST

/usr/opscm/bin/opscm 
$ORACLE_HOME/bin/lkmgrd -u $USER -g $GROUP

In each script, make sure that the values for LKDOM= and LKINST= are accurate for the domain and instance on that host. Theses values must match those in /usr/opscm/conf/sidconf, as explained in "Editing sidconf," earlier in this chapter.

Run the startup script on each host to bring up OPS.

In each host console window, start the Oracle database.

To check DLM operation, run dlmstat:

setenv ORACLE_SID sidname
$ORACLE_HOME/bin/dlmstat -i -q 

No error messages should appear.