If you configure a multisystem node in your RRSF network, at a
later time you might decide that you need to reconfigure the multisystem
node to make a different system the main system. This type of reconfiguration
is complex, and it is not a good solution when a system drops or a
connection is broken. Instead, when a system drops, re-IPL it. When
a connection breaks, let requests and returned output queue up in
the workspace data sets while you fix the connection. However, if
you need to shift the RRSF workload performed by the main system to
another system, you need to reconfigure the multisystem node. RACF® allows you to reconfigure
the multisystem node dynamically, without stopping the RACF subsystem on any system other than the
original local main system, or losing any RRSF requests or returned
output.
Restriction: You cannot change the main system
to or from a node for which a non-functional TCP/IP protocol instance
exists, or for which a protocol conversion is in progress. If you
try to do this, the TARGET command fails.
When you reconfigure a multisystem node, you cannot
safely just delete node definitions on live systems and redefine them.
You risk losing requests from automatic direction or password synchronization
while the nodes are not defined. Instead, to configure a different
main system in a multisystem node, follow these steps: - Drop TSO/E and JES on the original local main system.
Doing
this prevents any further RACF activity
on the system, and so prevents RACF databases
from becoming unsynchronized if you use automatic command direction.
- On the original local main system, issue the RACF STOP command to stop the RACF subsystem.
Doing this causes the system
to attempt to finish up all outstanding work and to close its workspace
data sets. New requests and returned output from remote nodes will
queue up in their OUTMSG workspace data sets.
- Make connections dormant:
- On the local system that is to be the new main, issue a TARGET
DORMANT command for its local connection. Also issue TARGET DORMANT
commands to make all connections with remote nodes dormant.
- On each remote node, issue TARGET DORMANT commands for the original
and new main systems. Do not perform step 7 until
the INMSG files for the original and new main systems on each remote
node have drained.
Tip: In
this step and others where you must issue the same command or set
of commands on every system (or every nonmain system, in this case)
on a multisystem node, you can make the task a little easier by creating
a RACF parameter library member,
IRROPTxx, containing the commands. Then
you can issue a SET INCLUDE(xx) command
on each system instead of manually entering the commands.
After
you issue the TARGET DORMANT commands, requests in the INMSG workspace
data sets are processed, and their output is queued up in the OUTMSG
workspace data sets. New requests and returned output also queue up
in the OUTMSG data sets. Issue TARGET LIST commands to verify that
the INMSG data sets on the local node have been drained before you
go on to the next step.
- If the workspace data sets for the original main system and the
new main system are not on shared DASD with a shared catalog, copy
the workspace data sets for the original main system to DASD accessible
to the new main system, using the same workspace data set names.
- On the new main system, issue a TARGET MAIN command
to make it the main system. For example, if the multisystem node name
is NODEABC, and MVSB is the new main system, issue:
TARGET NODE(NODEABC) SYSNAME(MVSB) MAIN
This
command causes RACF to transfer
requests and returned output from the original main system's OUTMSG
workspace data sets to the new main system's OUTMSG workspace data
sets.
If you have not specified the prefixes for the workspace
data sets and the LU names for the member systems consistently in
the TARGET commands that defined the local multisystem node, this
step will fail. See Defining RRSF nodes to RACF.
- Issue the same TARGET MAIN command that you issued in
step 5 on each nonmain system
on the local multisystem node. Issue this command on the original
main system only if it is to remain in the multisystem node.
- Issue TARGET LIST commands to verify that the INMSG
data sets on the remote nodes have been drained before you perform
this step.
On each remote system (that is, all remote systems
of all remote nodes), issue the same TARGET MAIN command that you
issued in step 5.
- On the new main system, issue TARGET OPERATIVE commands to make
the connection with itself and all connections with remote nodes operative.
- On each remote system (that is, all remote systems of all remote
nodes), issue TARGET OPERATIVE commands for the original main (if
it is to remain in the multisystem node) and new main systems.
- Update the TARGET commands in the RACF parameter libraries for all systems on
all nodes in the RRSF network to reflect the new main system.
If
you fail to update the RACF parameter
library for a system, the next time that system has its RACF subsystem restarted or is IPLed, the original
TARGET commands will be issued, and requests and returned output will
accumulate in the wrong OUTMSG workspace data set. However, RACF will issue appropriate error
messages and prevent communications.
You can update the parameter
libraries at the same time you issue the TARGET MAIN commands in steps 5, 6,
and 7. This approach helps to ensure
that no parameter library updates are missed.
- If the original main system is still part of the multisystem node,
(and assuming that you have updated its RACF parameter
library as discussed in step 10)
restart the RACF subsystem,
TSO/E and JES on the original main system.