This document discusses CICS node error program considerations.
The following topics are covered:
See also:
For information on installing a CICS node error program, refer to the corresponding section in Installing Natural CICS Interface for z/OS in the Natural Installation documentation.
Whenever a Natural session is active, CICS resources such as thread storage, roll facility entries (that is, records in a VSAM RRDS file or in a CICS temporary storage queue), swap pool slots etc. are used.
If these resources are under the control of the Natural CICS Interface, they are correctly released whenever a session terminates normally or abnormally.
The following cases cannot be controlled by the Natural CICS Interface:
A non-Natural program called by Natural issues an EXEC
CICS ABEND CANCEL command or the equivalent CICS macro request:
the CICS task is canceled without the Natural CICS Interface receiving control
to properly release all session resources.
Some CICS monitor products offer tools to purge CICS tasks, bypassing any abnormal termination exit set by the application. If a Natural task is canceled this way, the Natural CICS Interface has no chance to release the resources still owned by the session.
A user disconnects a terminal from the CICS region (by switching the power off or using an adequate session manager function) while a Natural session is currently not active in CICS (pseudo-conversational screen I/O).
The Natural CICS Interface provides some recovery mechanisms to recover from such situations; for example:
Whenever a new Natural session is to be started, a table is scanned for another Natural session still active with the same terminal ID. If such a session exists, it is logically terminated, and all its resources are released prior to starting the new one.
However, it may take quite a long time between logically terminating the session and releasing its resources, and there may also be a security concern:
When the NTCICSP macro parameter
COMARET is
set to OFF, the information to resume a Natural session is kept in
a CICS temporary storage record with the terminal ID being part of the
temporary storage queue name. If another CICS user tries to start Natural with
this terminal ID, he/she will resume the old Natural session rather than
starting a new one.
The third case in the above list is the most crucial one. CICS provides
a node error program (NEP) exit interface, which can be used in these cases to
trigger the Natural CICS Interface to terminate the lost session. An
appropriate module called NCIZNEP is provided in the Natural CICS
Interface (see Natural CICS Sample
Programs); it must be called by a DFHZNEP
node error program.
Based on session restart information
(NEXTTRANSID and restart data in COMMAREA or CICS
temporary storage), the Natural CICS Interface tries to resume the session and
terminate it logically.
Based on restart information (NEXTTRANSID and
restart data in COMMAREA or CICS temporary storage) of a terminal bound
session, which is currently pending in a pseudo-conversational screen I/O,
NCIZNEP tries to resume the session asynchronously and terminate
it logically.
Optionally, NCIZNEP can also try to purge an active Natural
session. This functionality is available for CICS Transaction Server systems
only. To enable this functionality, specify
PURGE=YES
for the NCIZNEP module link. See the installation procedure for
the Natural CICS Interface and the section CICS Startup
Parameters in Installing Natural CICS Interface
for z/OS
in the Natural Installation documentation.
MRO/CICSPlex environments:
The module NCIZNEP must be defined also in the AOR.
A transaction ID has to be defined for the module NCIZNEP
in the AOR.
This transaction ID must also be set via the
NEPTRAN parameter in the CICS startup parameters in the
NCIZNEP module as described in Installing Natural CICS
Interface for
z/OS
in the Natural Installation documentation.
Upon completion, NCIZNEP indicates to the caller (normally
DFHZNEP), by clearing the parameter input, whether it has
successfully completed its work; that is, launching the Natural session
clean-up task.
This is in particular of interest, if more than one Natural CICS
versions are active in a CICS system: clean-up function of a specific
NCIZNEP may fail, because the Natural session to be tidied up is a
different Natural CICS Interface version. DFHZNEP now can test
whether the called NCIZNEP (for example,
NCIvrNEP, where
vr represents the current Natural
version) was
successful, and if not, call another NCIZNEP (for example,
NCInnNEP, where
nn may be any preceding Natural
version).
There are still some items to be considered:
The Natural CICS source library contains the sample node error program
XNCINEP2 for CICS. This sample programs does not perform anything
special for the Natural CICS environment, it merely calls (via LINK) the
NCIZNEP module, which then deals with Natural under CICS.
DFHZNEP may already be customized for a specific
installation; as only one node error program is possible, the logic of the
relevant XNCINEPx program should be
adapted to the existing DFHZNEP logic.
In MRO environments, DFHZNEP and NCIZNEP
must be installed in the TOR.
The NCIZNEP module must be defined with
EXECKEY(CICS).
In the case described under item 3. above, DFHZNEP may
receive control more than once for various internal error codes, since each
internal error code is related to a specific CICS error message, but there may
be more than one error message resulting from a given action.
Upon completion NCIZNEP indicates to the caller (normally
DFHZNEP) — by clearing the parameter input — if it has
successfully completed its work, that is, launching the Natural session
clean-up task.
The CICS control block constellation may have changed each time a node
error program has been invoked, for example, the COMMAREA and
NEXTTRANSID information in the TCTTE may have been
lost after a certain node error event. In this case, the NTCICSP
parameter COMARET must be set
appropriately, which means that you cannot choose a node error event for your
node error program to be invoked when passing the Natural pseudo-conversational
session restart data in a CICS COMMAREA that has already been cleaned up by
CICS.
If you want to know how many times and with what error codes
DFHZNEP is invoked on certain actions and how the TCTTE should
look, write a dummy node error program, which only issues CICS trace requests
showing the requested information.
The following sample enables a DFHZNEP error processor to
receive control for all possible error codes passed to
DFHZNEP:
.
DFHSNEP TYPE=INITIAL
ORG NEPTT
DC 256X'03' invoke error processor '3' for ALL error codes
ORG ,
DFHSNEP TYPE=ERRPROC,GROUP=3,CODE=49
.
set up requested information and issue trace request(s)