AFCV   

Explanation

A request made against a file opened in RLS mode was unable to acquire a record lock. It waited for the lock, but the wait time exceeded the maximum wait time applicable to that request.

System action

The task is abnormally terminated with a CICS transaction dump.

CICS prints message DFHFC0164 and message(s) DFHFC0165 or DFHFC0175 which identify the transaction(s) or Transactional VSAM unit(s) of recovery that were immediately in front of this transaction in the queue for the lock. Normally these transaction(s) or Transactional VSAM unit(s) of recovery are the owners of the lock, although this is not the case if a chain of requests for the record has built up.

User response

Retry the transaction.

If the problem recurs, see messages DFHFC0164 and DFHFC0165 or DFHFC0175 to determine the transaction or Transactional VSAM unit of recovery that is holding the lock. In most cases the problem lies with the lock owner rather than the transaction that has failed.

Examples of reasons why CICS transactions may cause a timeout:

  • The transaction that holds the lock has a design error. For example:

    • A conversational transaction updates a recoverable record and then issues a terminal control read. It does not issue syncpoint (and therefore does not release the lock) until the end user has responded to the terminal control read. It may therefore hold the lock for a considerable period.

    • A transaction updates very many records in recoverable files before issuing syncpoint. You are recommended to keep the number of updates made within a unit of work small and to issue frequent syncpoints to ensure that locks are released regularly.

  • The system in which the lock holder is running is experiencing severe performance degradation. Investigate the reason for the performance degradation.

  • There is a deadlock between RLS and another resource manager. For example one transaction may be holding an RLS lock and waiting for a lock on a transient data queue. The transaction that times out may hold the lock on the transient data queue and be waiting for the RLS lock. RLS can detect deadlocks only when all the locks involved in the deadlock are RLS locks. A deadlock such as this can appear to RLS to be a long wait for a lock and is reported as a time out. Examine the design of the transactions to determine whether resource manager deadlocks can occur.

  • It may be possible for RLS deadlocks to be reported as RLS timeouts if VSAM does not perform deadlock detection until after the time out value for the request occurred. For example, assume that DEADLOCK_DETECTION is specified as (15,4) in SYS1.PARMLIB, member IGDSMSxx. This means that VSAM does not attempt to detect cross-MVS deadlocks until 4 periods of 15 (that is, 60) seconds have elapsed. If DTIMOUT was not active for the transaction and the SIT specified FTIMEOUT=30, the RLS request times out after 30 seconds, before VSAM has attempted to detect cross-MVS deadlocks. Adjust FTIMEOUT, DTIMOUT, and DEADLOCK_DETECTION to avoid such effects.

DFHFC0175 messages identify Transactional VSAM units of recovery owning an RLS lock. If a Transactional VSAM application is the lock owner it should be investigated to determine why it is holding the lock. Some of the above considerations will be similar for Transactional VSAM applications.

Module

DFHEIFC, DFHDMPCA


AFCV.html | Timestamp icon Last updated: Sunday, 4 June 2017