FORTRAN Reference Manual

Fault-Tolerant Programming
FORTRAN Reference Manual—528615-001
16-5
Overview of Fault- Tolerant Programs
•
The START BACKUP statement does not establish a takeover point. You must
execute a CHECKPOINT statement that does not specify STACK = 'NO' to
establish a takeover point.
•
If the primary process, or the processor in which it runs, fails, the backup process
begins executing instructions at the most recent takeover point. (Takeover points
are described earlier within this subsection.) The backup process becomes the
new primary process.
•
In some cases, FORTRAN starts a new backup process in the former primary’s
processor. Under some circumstances, however, FORTRAN either does not or
cannot start a new backup process. For example, if the processor in which the
primary process was running is down, FORTRAN cannot start a new backup
process in that processor and returns backup status 102. If the primary process
has failed more than ten times, FORTRAN does not start a new backup process.
Table 7-10 in Section 7, Statements lists the backup status codes.
If FORTRAN returns a backup status code that indicates that it has not started a
new backup process, your program must execute a START BACKUP statement if
you want to continue running as a NonStop process.
Figure 16-1 on page 16-6 illustrates a typical fault-tolerant application. The backup
process remains in the monitor state while the primary is operational. If the primary
fails, the backup leaves the monitor loop and begins executing instructions at the most
recently established takeover point.