Connections that do not fully support shunting

The information in previous sections assumes that you are using MRO or APPC parallel-session connections to other CICS® Transaction Server for z/OS® systems--that is, that your network consists solely of current systems that fully support shunting. Much of the preceding information applies equally to other types of connection. This section describes exceptions that apply, for example, to connections to back-level systems.

LU6.1 connections

This section describes the ways in which LU6.1 connections differ from APPC parallel-session connections and MRO connections to CICS TS z/OS systems.

Recovery functions and interfaces

Some recovery functions are not available to LU6.1 connections:

Restriction on shunting support

There is no LU6.1 protocol by which one system can notify another system that a unit of work has been shunted. The only time when a UOW that includes an LU6.1 session can be shunted is when all the following are true:

Under these conditions, the UOW can be shunted, because there is no need for the LU6.1 partner to be notified of the shunt.

Under other conditions, a UOW that fails in the in-doubt period, and that involves an LU6.1 session, takes a unilateral decision. If WAIT(YES) is specified on the transaction definition, it has no effect--WAIT(NO) is forced.

Unsupported commands

The following commands are not supported on LU6.1 connections:

Lack of SYNCPOINT ROLLBACK support

There is no LU6.1 protocol by which one system can notify another that a UOW has been backed out, without terminating the conversation. An attempt to issue an EXEC CICS SYNCPOINT ROLLBACK command in a UOW that includes an LU6.1 session results in an ASP8 abend. This abend cannot be handled by the application program.

Any resources in the UOW are backed out, but the transaction is not able to continue.

Session-by-session resynchronization

Unlike APPC parallel-session connections and CICS TS z/OS-CICS TS z/OS MRO connections, LU6.1 sessions are resynchronized one by one, as they are bound. Therefore, any UOW that requires resynchronization is not resynchronized until the session that failed is reconnected.

Initial and cold starts

The LU6.1 connection definition contains sequence numbers used for recovery. If you perform an initial or cold start of CICS when there are LU6.1 connections on which recovery is outstanding, the sequence numbers are lost, and it becomes impossible for the partner systems to resynchronize their outstanding units of work.

Lognames are not used. Therefore, the XLNACTION option of the CEDA DEFINE CONNECTION command is meaningless for LU6.1 connections.

Managing connection definitions

Recovery information for a remote system is not stored independently from the connection definition for the system--the LU6.1 connection definition contains sequence numbers used for recovery. Therefore you should not modify or discard connections for which recovery information may be outstanding.

APPC connections to non-CICS TS z/OS systems

Some non-CICS Transaction Server for z/OS systems that can be connected to by APPC links do not support shunting, and always take unilateral action if a session failure occurs during the in-doubt period. Inevitably, communication with a system that does not support shunting involves a risk of damage to data integrity through the taking of unilateral decisions. It is not possible for CICS to distinguish systems that do not support shunting from others that do support shunting. Therefore, it cannot preferentially select such a system to be the coordinator of a unit of work.

Note the following:

APPC single-session connections

Normal syncpoint protocols cannot be used across a connection that is defined as SINGLESESS(YES). If function shipping is used (inbound or outbound), CICS communicates the outcome of a unit of work as described in the CICS Family: Communicating from CICS on System/390® manual. However, resynchronization cannot be performed in the case of session failure.

CICS issues a message to inform you of the shunting--but not the unshunting-- of a unit of work.

If the connection to which a function-shipped request is made is defined as remote (that is, it is owned by a remote region), the connection to the remote region must be defined as a parallel-session link, if recovery protocols with the resource-owning system are to be enabled.

Related concepts
Syncpoint exchanges
Recovery functions and interfaces
Initial and cold starts
APPC connection quiesce processing
Related tasks
Managing connection definitions
Problem determination
Related reference
Terminology
[[ Contents Previous Page | Next Page Index ]]