9.3.2 Designing synchronization point dump files

This section describes the design considerations for synchronization point dump files.

Organization of this subsection
(1) Design considerations
(2) Design for improved reliability
(3) Defining the synchronization point dump files

(1) Design considerations

  1. Synchronization point dump files are required for each server, except for the system manager.
  2. You can create 2-60 groups of synchronization point dump files per server (if ONL is specified, 2-30 groups per server).
  3. You should create at least four synchronization point dump files per server.
  4. HiRDB uses synchronization point dump files in the order specified in the pdlogadfg -d spd operand.
  5. If a shortage of space occurs in a synchronization point dump file, HiRDB cannot be restarted. For this reason, the size of a synchronization point dump file should be set to be greater than the value specified for the maximum number of concurrent connections (pd_max_users) in the system common definition. For details about the calculation of synchronization point dump file size, see 17.2 Determining the size of synchronization point dump files.

(2) Design for improved reliability

(a) Example of file organization

As a safeguard against the possibility of hard disk failures, the synchronization point dump files should be created on separate hard disks. If this is not possible, adjacent generations of files should be created on separate hard disks, as shown in the example in the following figure.

Figure 9-5 Example of creating adjacent generations on separate hard disks (HiRDB/Parallel Server)

[Figure]

(b) Duplexing of the synchronization point dump file

When the synchronization point dump file is duplexed, HiRDB collects the same synchronization point dump on both system A and system B. This increases system reliability, because when a collected synchronization point dump is read and there is an abnormality in the file, the synchronization point dump can still be read from the other file. Duplexing also enables the number of guaranteed-valid generations to be set to one generation, yet reliability is not compromised and the number of synchronization point dump files in overwrite disabled status is reduced.

Specify the following operands in the server definition to enable duplexing of synchronization point dump files:

(c) Number of guaranteed-valid generations for synchronization point dump files

Each synchronization point dump acquired by HiRDB is stored in a separate synchronization point dump file. HiRDB uses the generation concept to manage synchronization point dump files. The HiRDB administrator specifies the number of generations of synchronization point dump files, and the corresponding system log files, that are to be placed in overwrite-disabled status. This concept is called the number of guaranteed-valid generations for synchronization point dump files, and it is illustrated in the following figure.

Figure 9-6 Number of guaranteed-valid generations for synchronization point dump file (HiRDB/Parallel Server)

[Figure]

Explanation
If there are two guaranteed-valid generations, the synchronization point dump files up to the second generation, and the system log files relevant to those synchronization point dump files, are in overwrite disabled status. The synchronization point dump files prior to the third generation, and the system log files relevant to those synchronization point dump files, are in overwrite enabled status.

The required number of synchronization point dump files is the number of guaranteed-valid generations + 1. Specify the number of guaranteed-valid generations for synchronization point dump files in the pd_spd_assurance_count operand in the server definition.

If synchronization point dump files are to be duplexed, it is recommended that only one guaranteed-valid generation be required. If duplexing is not to be used, two generations are recommended.

(d) Reduced mode operation for synchronization point dump files

If the number of synchronization point dump files available for use is reduced to the number of guaranteed-valid generations + 1 because of errors in synchronization point dump files, processing can continue with a minimum of two files. This is called the reduced mode operation for synchronization point dump files.

To perform reduced mode operation for synchronization point dump files, specify the pd_spd_reduced_mode operand in the server definition.

(e) Automatic opening of synchronization point dump files

When the number of synchronization point dump files available for use is reduced to the number of guaranteed-valid generations + 1 because of errors in synchronization point dump files, processing can continue by opening a reserved file and placing it in overwrite-enabled status (assuming that such a reserved file is available). This is called automatic opening of synchronization point dump files.

To perform automatic opening of synchronization point dump files, specify the pd_spd_reserved_file_auto_open=Y in the server definition.

(3) Defining the synchronization point dump files

The pdlogadfg and pdlogadpf operands are used to define the correspondence between file groups and the created synchronization point dump files.

If only the pdlogadfg operand is specified, synchronization point dump files can be added during HiRDB operation.