CICS Recovery

CICS recovery is the set of facilities that protects your data when a task fails or the CICS region stops. It guarantees that changes to recoverable resources are either fully completed or fully backed out - never left half-done.

What is CICS recovery

• Recovery makes every unit of work atomic: all of its changes are committed together, or all are backed out together.
• Before CICS applies a change to a recoverable resource, it writes a journal (log) record describing the change.
• If the task or the region fails, the CICS recovery manager uses the journal to undo incomplete work and redo committed work.
• Recovery works across resource types - VSAM files, temporary storage, transient data and DB2 tables can all be recovered as one unit.

Recoverable resources

• A VSAM file is recoverable only when its resource definition specifies recovery (RECOVERY attribute); otherwise CICS cannot back out its updates.
• Temporary storage queues and transient data queues can be defined as recoverable so their updates are journalled.
• DB2 tables accessed through CICS are recoverable and are coordinated with CICS by the two-phase commit protocol.
• Non-recoverable resources are faster but risky: a failure in the middle of updates leaves the data inconsistent and must be fixed manually.

How CICS performs recovery

• During normal running, every update to a recoverable resource is journalled with a before-image and an after-image of the record.
• When a task abends, dynamic transaction backout automatically undoes that task's changes while CICS keeps running.
• When the whole region fails, emergency restart reads the system log, backs out in-flight units of work and completes the ones that were committed.
• Example - a program that updates two accounts and commits them as one unit of work:-
  3000-UPDATE-ACCOUNTS.
  EXEC CICS
  READ FILE('ACCTMAS') INTO(ACCT-RECORD)
  RIDFLD(ACCT-NO-1) UPDATE
  END-EXEC.
  SUBTRACT TRANSFER-AMT FROM ACCT-BALANCE.
  EXEC CICS
  REWRITE FILE('ACCTMAS') FROM(ACCT-RECORD)
  END-EXEC.
  *-- read and update the second account the same way
  EXEC CICS
  SYNCPOINT
  END-EXEC.
• If the task abends before the SYNCPOINT, both account updates are backed out and the balances stay consistent.