DFHFCRC (file control recovery control program)

DFHFCRC provides recovery control for file control. All calls from the Recovery Manager domain to file control come through DFHFCRC.

DFHFCRC is called by the Recovery Manager domain to participate in syncpoint and in warm and emergency restart.

Early on during startup File Control registers as a client of the CICS Recovery Manager. During File Control initialization, File Control will add its recovery gate to the kernel, specifying DFHFCRC as the entry point, and then declares the recovery gate to the CICS Recovery Manager via an RMCD SET_GATE call.

At syncpoint, a resource owner such as File Control may be called either
  1. to prepare, optionally followed by shunt-unshunt pairs, followed either by calls to backout (as in 2) or a call to commit.
  2. to backout, which involves start_backout, optional delivery of backout data, and end_backout, followed by prepare and commit, optionally followed by backout retries (which consist of shunt-unshunt pairs followed by the start_backout - delivery of backout data - end_backout - prepare - commit sequence).

At warm or emergency restart, a resource owner such as File Control will be called with start_delivery, optional deliver_recovery and deliver_forget calls, followed by end_deliver.

The Recovery Manager functions processed by DFHFCRC are:
  • RMRO PERFORM_PREPARE
  • RMRO PERFORM_COMMIT
  • RMRO START_BACKOUT
  • RMRO DELIVER_BACKOUT_DATA
  • RMRO END_BACKOUT
  • RMRO PERFORM_SHUNT
  • RMRO PERFORM_UNSHUNT
  • RMKP TAKE_KEYPOINT
  • RMDE START_DELIVERY
  • RMDE DELIVER_RECOVERY
  • RMDE DELIVER_FORGET
  • RMDE END_DELIVERY

DFHFCRC performs different processing depending on the function with which it has been called:

PERFORM_PREPARE

Any active VSAM requests are terminated, and a vote of READ_ONLY is returned if the unit of work did not make any recoverable file control updates, a vote of YES if the prepare was successful, or a vote of NO otherwise.

PERFORM_COMMIT

For a forwards syncpoint, any changes made by the unit of work to recoverable user-maintained data tables are committed. For a backwards syncpoint, locks for any backout-failed data sets are retained. All other locks are released.

On transaction termination, the FLABs and FRAB are freed unless there are FLABs marked for retention. On an intermediate syncpoint, various flags in the FLABs and FRAB are reset to indicate that a commit has been performed.

START_BACKOUT

Any active VSAM requests are terminated, and any changes made by the unit of work to recoverable user-maintained data tables are backed out.

DELIVER_BACKOUT_DATA

The recoverable file control change represented by the log record delivered to DFHFCRC is backed out via calls to DFHFCFR which reverse the update. The change is not backed out if the unit of work has already suffered a backout failure for the data set, or if the data set is in a ’non-RLS update permitted’ state, or if this call is being made as part of a CEMT or EXEC CICS SET DSNAME RESETLOCKS request.

If a failure occurs during the backout, then backout failure processing is carried out.

END_BACKOUT

Under normal conditions there should be no processing required at END_BACKOUT, but it is conceivable that there might be outstanding active VSAM requests to be terminated.

PERFORM_SHUNT

The failed parts of the unit of work's file control structures are put into a condition to survive without an executable transaction environment. This involves retaining any FLABs that are marked for retention, which will allow files to be closed, but not to be reallocated to a different data set.

If this is an intermediate syncpoint, and the shunt is due to a failure in phase 2 of syncpoint, the transactional parts of the unit of work are copied into a new control structure to be passed to the follow-on unit of work. A new FRAB is acquired to anchor this control structure. If this is transaction termination, or the shunt is due to a failure in phase 1 of syncpoint, the transactional parts are cleaned up.

PERFORM_UNSHUNT

The file control structures are converted back into a condition suitable for a unit of work that is in an executable state. Retained FLABs for the unit of work are restored.

TAKE_KEYPOINT

DFHFCRC is called when CICS takes a keypoint, to perform processing required by BWO backup on non-RLS data sets. This involves the writing of a set of ’tie up records’ and the calculation of a new BWO recovery time.

START_DELIVERY

DFHFCIR is called to process the call.

DELIVER_RECOVERY

DFHFCIR is called to process the call.

DELIVER_FORGET

DFHFCIR is called to process the call.

END_DELIVERY

DFHFCIR is called to process the call.



dfhs3fu.html | Timestamp icon Last updated: Monday, 5 June 2017