Diagnosing persistent sessions problems
Consult this data when diagnosing problems with z/OS® Communications Server persistent sessions support.
Note: VTAM® is the previous
name for z/OS Communications
Server.
- Trace, TC level 1, 2, and exception in the range of AP FB10-FBFF.
- CEMT INQUIRE VTAM showing the PSTYPE and PSDINT values. The setting NOPS for PSTYPE means that persistent sessions support was not requested at startup of the CICS® region. A zero setting for PSDINT means that sessions are not retained.
- Console and CSNE logs:
- Persistent session messages (DFHZC0001 to DFHZC0162)
- Information produced by DFHZNAC
- Dumps taken by some of the messages listed earlier. If a NIBLIST was present at the time the dump was taken, you can examine it by printing the TCP section of the dump.
- Last flow information, that is, the CV29, FMH5, BIS, and BID information,
is useful if a session is in the wrong state after a persistent session
restart. This might have been diagnosed by an error message, or maybe
missed and message DFHZC0146 or DFHZC0156 issued.
TCTE_PRSS_CV29_PTR points to the CV29, FMH5, BIS, and BID information which was created by DFHZGRP and used by DFHZXPS or DFHZXRC. It is freed when DFHZNAC issues message DFHZC0146 or DFHZC0156. Otherwise, it is freed when the session is unbound.
The last flow information is traced by DFHZXPS as a TC level 1 trace. If you have a dump, but no trace level 1 available, it is dumped in the TCP section for each TCTTE for which it still exists.
- The contents of byte TCTE_PRSS are useful. Values other than X'00' and X'FF' indicate that something went wrong during the PRSS recovery. The possible values are listed in the Data areas. If a value remains in this byte, the meaning can indicate where the recovery went wrong. The values are described in Persistent sessions status byte (TCTE_PRSS).
- The contents of the state machines are useful.
- TCTECNTS, contention state machine
- TCTEBKTS, bracket state machine
- TCTECHSS, chain state machine
- TCTEUSRS, user state machine
- The contents of TCTE_BID_STATUS are useful. They are described in Bid status byte (TCTE_BID_STATUS).
Here are some possible problems:
- DFHZGRP can cause CICS to
stop during initialization for the following reasons:
- DFHZGRP has been called with invalid parameters.
- DFHZGRP cannot complete the receive any process.
- DFHZGRP has had a loop or abend.
- DFHZGRP cannot switch back to the QR TCB.
- DFHZGRP has failed before any NIBs have been obtained from z/OS Communications Server (with INQUIRE OPTCD=PERSESS).
- DFHZGRP or DFHZGUB has issued a z/OS Communications Server request that failed to respond within 5 minutes. Issued with message DFHZC0128 and a system dump.
- Sessions might be unbound by DFHZGRP for the following reasons:
- The open of the ACB is dynamic, for example, CEMT SET VTAM OPEN. However, if MNPS is in use, sessions are normally restored at this point.
- The TCTTE has not been found, probably because it has not been cataloged. Either the terminal was autoinstalled with AIRDELAY=0, or it was an APPC clone. No message is written because this state is considered to be normal.
- CICS does not support recovery for LU61 or pipeline sessions.
- The TCTTE does not match the NIB, possibly because of an operational failure. Has the correct global catalog data set been used?
- A terminal or session had RECOVOPT UNCONDREL|NONE specified.
- A connection had PSRECOVERY NONE specified.
- A matching mode group was not found. Have you got the correct global catalog data set?
- A suitable session was not found, perhaps because the CNOS values created many “up for grabs” sessions which were in use when CICS failed. This situation would occur if the session limit was high and the contention winners was low. The situation might also occur if CICS was in the process of CNOSing from a high session limit to a low session limit at the time CICS failed. Message DFHZC0111 is issued in this case.
- An URD was found for the session so the entire connection is unbound to allow the connection to recover correctly.
- APPC Sessions might be unbound by DFHZXPS for the following reasons.
Some of the reasons are known states for which the session cannot
be recovered. Others are unexpected errors. Known states for which the session cannot be recovered are as follows:
- The last flow was a positive response to a bid with data.
- Exchange log names (transaction CLS2) was running when the system failed.
- A bind or bind security had not completed when the system failed.
- Because of the last thing to flow, for example, SIGNAL, the state of the session at the time of system failure cannot be determined.
Unexpected errors are as follows:- A bad return code was received from a call to DFGZGDA.
- An attempt to reset the session from CS mode to CA mode or vice versa failed.
- The TCTE_PRSS byte contained an unexpected value on entry to DFHZXPS.
- The BIS, BID, or CV29 data pointed to by TCTV_PRSS_CV29_PTR contained an unknown value or was inconsistent.
- An error occurred during some other part of the recovery process.
- An internal logic error occurred in DFHZXPS.
- Sessions might be unbound by DFHZGDA for the following reasons:
- A SEND issued as part of Deallocate(Abend) processing has failed.
- A RECEIVE issued as part of Deallocate(Abend) processing has failed.
- A logic error is detected during Deallocate(Abend) processing.
- Sessions might be unbound by DFHZXRC for the following reasons:
- The user has specified RECOVOPT(RELEASESESS) and the session was in bracket at the time CICS failed.
- End-Bracket and Clear/SDT cannot be used to clean up the session.
- Cold Start has been requested for the session.
- Message DFHZC0124 can be issued with inconsistent counts if these
conditions occur::
- DFHZGRP loops or abends.
- The ACB is closed by z/OS Communications Server operator commands while DFHZGRP is in control.
- LU6.2 connections, which might be expected to persist, might be unbound if a persistent resource is associated with the connection when CICS fails (that is, there was an asynchronous processing request in progress at the time CICS failed).
- Following a persistent sessions restart, LU6.2 partners might experience a series of unexpected abends with sense code 08640001 from the persisting CICS. This condition can occur either because there was a conversation in progress at the time CICS failed, and CICS has ended the conversation with this code, or for synclevel 2 conversations, the partner has attempted to initiate a conversation before Exchange Lognames has run following a persistent sessions restart.
- Some APPC sessions might hang following a persistent sessions restart because CICS has determined that it was in RECEIVE state at the time of the CICS failure, and issued a RECEIVE for the expected data, but the partner has not sent the expected data; the RECEIVE will not time out in this situation, because RTIMOUT does not apply to sends issued by DFHZGDA.