This section describes how to use a distributed routing program to dynamically route non-terminal-related EXEC CICS START requests.
For a non-terminal-related START request to be eligible for dynamic routing, all of the following conditions must be met:
If the request is fully eligible for dynamic routing, the distributed routing program is invoked for routing. The START request is function-shipped to the target region returned by the routing program.
"Daisy-chaining" is not supported. That is, once a non-terminal-related START request has been dynamically routed to a target region it cannot be dynamically routed from the target to a third region, even though the transaction is defined as ROUTABLE(YES) and DYNAMIC(YES). The transaction may, however, be statically routed from the target region to a third region.
For definitive information about which non-terminal-related START requests are eligible for dynamic routing, see the CICS Intercommunication Guide.
For non-terminal-related START requests that are eligible for enhanced routing, CICS invokes the distributed routing program at the following points:
The routing program is not able to route the request. It could, however, do other things.
This invocation signals that (unless the routing region and the target region are one and the same) the routing region’s responsibility for this transaction has been discharged. The routing program might, for example, use this invocation to release any resources that it has acquired on behalf of the transaction.
Figure 51 shows the points at which the distributed routing program is invoked, and the region on which each invocation occurs. Note that the "target region" is not necessarily remote--it could be the local (routing) region, if the routing program chooses to execute the START request locally.
The DYRSYSID field of the communications area passed to the distributed routing program initially contains the system identifier (sysid) of the default target region to which the request is to be routed. This is derived from the value of the REMOTESYSTEM option of the installed transaction definition on the routing region. If REMOTESYSTEM is not specified, the sysid passed is that of the local CICS region.
When it is invoked for route selection, the distributed routing program can change the target region by changing the value in DYRSYSID.
If the specified sysid is invalid, or cannot be found, SYSIDERR is returned to the distributed routing program--which may deal with the error by returning a different sysid--see If an error occurs in route selection.
If the routing program changes the sysid when it is invoked for notification, routing complete, transaction initiation, transaction termination, or abend, the change has no effect.
When the routing program is invoked for routing, if you want the request to be routed (whether you have changed any values or not) return a zero value to CICS in field DYRRETC of the communications area. When you return control to CICS with return code zero, CICS first compares the returned sysid with its own local sysid:
If you want CICS to reject the START request, return a non-zero value. The EXEC CICS START command receives a SYSIDERR condition, with a RESP2 value indicating that the START request has been rejected by the routing program.
Returning a value in DYRRETC has no effect when the routing program is invoked for notification, routing complete, transaction initiation, transaction termination, or abend.
If an error occurs in route selection--for example, if the sysid returned by the distributed routing program is unavailable or unknown--the routing program is invoked again. When this happens, you have a choice of actions:
If this region too is unavailable, the routing program is again invoked for a route selection error. A count of the times the routing program has been invoked for routing purposes for this request is passed in field DYRCOUNT. Use this count to help you decide when to stop trying to route the request.
You can use an XICERES global user exit program to check that all resources required by the started transaction are available on the target region.
To use the XICERES exit, both the routing region and the target region must support the "resource unavailable" condition (RESUNAVAIL). All the following support the "resource unavailable" condition:
The XICERES exit is invoked, if enabled, on the target region before CICS processes a dynamically-routed START request.
If, for example, the transaction to be started is disabled on the target region, or a required file is missing, your exit program can give the distributed routing program the opportunity to route the request to a different region. To do this, it should set a return code of UERCRESU. This causes CICS to:
For information about writing an XICERES global user exit program, see The XICERES global user exit.
If a required resource is unavailable on the target region, but the XICERES exit is unavailable or disabled (or is enabled but does not set the UERCRESU return code), the client program receives an error response.
The route selection, notification, route selection error, and routing complete invocations of the distributed routing program all occur on the routing region. If the routing program wants to be re-invoked on the target region, it must set the DYROPTER field in the communications area to 'Y'. It must do this on its initial (route selection or notification) invocation--and again, if it is reinvoked for a route selection error.
If the routing program sets DYROPTER to 'Y', it is re-invoked on the target region:
This is effective only if the target region is CICS TS OS/390, Version 1 Release 3 or later.
Each time it is invoked on the target region, the routing program could update a count of transactions that are currently running on that region. When it is invoked for routing, the routing program could use the counts maintained by all the regions in the routing set (including itself) as input to its routing decision. This requires that each region in the routing set has access to a common data set on which the counts are recorded.