Spooler Programmer's Guide

Using the Spooler Interface Procedures

Spooler Programmer’s Guide—522287-002

2-17

Spooling From a NonStop Process Pair

An application process sending data to a collector can run as a NonStop process pair.

Programmers writing such applications should be aware of the checkpointing

considerations described in this subsection.

This discussion assumes that you already have knowledge of fault-tolerant

programming (coding NonStop process pairs).

For an application program to run as a NonStop process pair, the system spooler

should also be running as a NonStop process pair. Otherwise, an attempt at NonStop

spooling from the application will not be fully effective.

If the collector is running as a NonStop process pair, you can spool with or without the

spooler interface procedures. However, the considerations are different in each case.

You must decide whether or not the application can tolerate duplication of a line of data

in the event of a failure of the application’s primary process. If the application cannot

tolerate duplication of any lines, then sync depth must be specified when the collector

is opened.

Use of Sync Depth

When a NonStop process pair opens the collector, it can set the sync-depth parameter

of the file-system OPEN or FILE_OPEN_ procedure to a value of 1 or more, up to 15.

Each write to the collector is then tagged by the file system with a sync ID. This

ensures that, in the event of a failure of the primary processor, the collector will

recognize any line rewritten by the backup.

If you have opened the collector with a nonzero sync depth, the information

checkpointed should include the synchronization block of the file to the collector. A

sync depth greater than 1 allows the application to perform less-frequent checkpoints.

If you open the file to the collector with a sync depth of n, the application need only

checkpoint before every nth write operation.

Spooling—Levels 1 and 2

The considerations for spooling on levels 1 and 2 from a NonStop process pair are

very similar. Your main concern while performing such spooling from an application

program is whether you can afford duplication in your spooled job.

Spooling With a Zero Sync Depth

If duplication of data lines can be tolerated, then a reasonable checkpointing strategy is

to checkpoint the primary processor stack, the line of data to be spooled, and the

Note. The examples in this subsection deal with checkpointing only to ensure the integrity of

the application-collector interface. Checkpointing the remainder of the program is left to the

programmer.