Routing BTS activities

This section describes how to use a distributed routing program to dynamically route CICS® business transaction services (BTS) processes and activities. It assumes that you have read the introduction to the distributed routing of BTS processes and activities in the CICS Business Transaction Services manual.

Which BTS activities can be dynamically routed?

Not all activations of BTS processes and activities can be routed.

Processes and activities that are activated asynchronously with the requestor--by means of a RUN ASYNCHRONOUS command--can be routed either dynamically or statically.

Processes and activities that are activated synchronously with the requestor--by means of a RUN SYNCHRONOUS or LINK command--are always run locally. They cannot be routed, neither dynamically nor statically. A RUN SYNCHRONOUS or LINK command issued against an activity whose associated transaction is defined as DYNAMIC(YES), or as residing on a remote region, results in the activity being run locally.

Thus, to be eligible for dynamic routing:

  1. A BTS process or activity must be run asynchronously with the requestor, by means of a RUN ASYNCHRONOUS command.
  2. The TRANSACTION definition for the transaction associated with the process or activity must specify DYNAMIC(YES).

"Daisy-chaining" is not supported. That is, once a BTS activity has been routed to a target region it cannot be re-routed from the target to a third region, even though its associated transaction is defined as DYNAMIC(YES).

When the distributed routing program is invoked

For BTS processes and activities started by RUN ASYNCHRONOUS commands, CICS invokes the distributed routing program at the following points:

On the routing region:10
  1. Either of the following:
    • For routing the activity. This occurs when the transaction associated with the activity is defined as DYNAMIC(YES).
    • For notification of a statically-routed request. This occurs when the transaction associated with the activity is defined as DYNAMIC(NO). The routing program is not able to route the activity. It could, however, do other things.
  2. If an error occurs in route selection--for example, if the target region returned by the routing program on the route selection call is unavailable. This gives the routing program the opportunity to specify an alternate target. This process iterates until the routing program selects a target that is available or sets a non-zero return code.
  3. After CICS has tried (successfully or unsuccessfully) to route the activity to the target region.

    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.

On the target region:
These invocations occur only if the target region is CICS TS OS/390®, Version 1 Release 3 or later and the routing program on the routing region has specified that it should be reinvoked on the target region:
  1. When the activation starts on the target region (that is, when the transaction that implements the activity starts).
  2. If the routed activation (transaction) ends successfully.
  3. If the routed activation (transaction) abends.

Figure 48 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 run the activity locally.

Figure 48. When and where the distributed routing program is invoked
 The picture shows two rectangles, one representing the requesting region and the other the target region. Within the requesting region, four labels indicate the points at which the distributed routing program is invoked:   Route selection Notification Route selection error Routing attempt complete Within the target region, three labels indicate the points at which the distributed routing program is invoked:   Transaction initiation Transaction termination Transaction abend

Changing the target CICS region

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 process or activity 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.

Telling CICS whether to route the activity

When the routing program is invoked for routing, if you want the process or activity 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 treat the request as unserviceable, return a non-zero value. For information about unserviceable requests, see the CICS Business Transaction Services manual.

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

If an error occurs in route selection--for example, if the sysid returned by the distributed routing program is unavailable or unknown--the distributed routing program is invoked again. When this happens, you have a choice of actions:

  1. You can try to route the request to a different target region, by changing the sysid, and issuing a return code of ‘0’ in DYRRETC.

    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.

  2. You can tell CICS to treat the request as "unserviceable", by issuing a non-zero return code in DYRRETC.

    Sometimes, perhaps because of a transaction affinity, it is essential that an activity should execute on a particular target region, and on no other. If this is the case, and the target region is unavailable, classify the request as unserviceable. Instead of reinvoking the routing program for a route selection error, CICS:

    1. Tries repeatedly to route the request to the specified target region, at 1-minute intervals.

      If one of these attempts is successful, CICS issues message DFHSH0108. The routing program is invoked on the routing region for "routing attempt complete", and, if specified, on the target region for "transaction initiation".

    2. Every hour, if the target region is still unavailable, issues message DFHSH0106.
    3. If the target region is still unavailable 24 hours after the request was issued, issues message DFHSH0107, and stops trying to route the request, which is discarded. The routing program is invoked on the routing region for "routing attempt complete".

Invoking the distributed routing program on the target region

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:

Each time it is invoked on the target region, the routing program could update a count of BTS activities 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.

Related concepts
Differences between the distributed and dynamic routing interfaces
Related tasks
Routing method requests for enterprise beans and CORBA stateless objects
Routing non-terminal-related START requests
Routing by user ID
Dealing with an abend on the target region
Some processing considerations
Naming your distributed routing program
Rewriting user-replaceable programs
Assembling and link-editing user-replaceable programs
Related reference
Parameters passed to the distributed routing program
Distributed transaction routing sample programs

10.
For BTS processes and activities, the routing region--the region which routes the request--is always the requesting region, the region on which the RUN ASYNCHRONOUS command is issued.

[[ Contents Previous Page | Next Page Index ]]