List files for optimized tape recalls

You can preview a list of files for optimized tape recalls without starting the recall process.

The preview option with the dsmrecall command generates the list files without starting the recall process. Each run with the preview option generates a set of files in a different directory. You can run simultaneous dsmrecall processes on the same node or on different nodes with different input list files. You can run simultaneous dsmrecall processes for different file systems or for the same file system.

The dsmrecall command generates the list files every time that it runs, even if you do not specify the preview option. The generated file lists are not deleted automatically. You must delete the list files when you no longer require them.

dsmrecall generates ordered list files — Figure 1. Ordered list file generation

Input file lists

In addition to the rules specified for list files, the files listed in the input list file must follow this rule:

Each entry in the list file must be from the same file system, and you must specify the file system when you start the dsmrecall command.

Any files from other file systems are skipped and written to a list of unprocessed files. This list of unprocessed files is stored in .SpaceMan/tapeOptimizedRecall/FileList.unprocessed.file_system_name. The list file of unprocessed files can be used as a new input list file for the next dsmrecall command.

Generated list files

There are two types of generated list files. All generated files are stored in .SpaceMan/tapeOptimizedRecall in the space-managed file system. Each time dsmrecall creates ordered list files, they are created in a new directory. You can run several processes simultaneously without overwriting the generated files. The directory includes the node ID (node_ID) and process ID (PID): .SpaceMan/tapeOptimizedRecall/node_ID/PID.

Tape and disk list files

One ordered list file is generated for each tape. The tape list file has the following properties:

The tape list file has a name with this format: filelist.ordered.tape.tapeID.

Each entry in the tape file has this format:

order_ID extObjID file_system_ID /filesystem/path/filename

Files in these lists are ordered by their position on the tape.

There is one file list for all files stored on a disk pool on the server. The disk list file has the following properties:

The disk list file has a name with this format: filelist.nonTape.00000000.

Each entry in the disk file has a format with this format:

order_ID extObjID file_system_ID /filesystem/path/filename

Files in the disk list are not ordered since they are on disk.
You cannot specify a tape list file or a disk list file on the filelist option of the dsmrecall command.

Collection files

There is one collection file that includes the names of the tape list files and the disk list file. The collection file has these properties:

The collection list file has a name with this format:
```
filelist.ordered.collection
```
Entries in this list are the names of the tape and disk list files.
Entries in the collection file have this format:
```
filecount byteCount filelist.nonTape.00000000
filecount byteCount filelist.ordered.tape tapeID1
filecount byteCount filelist.ordered.tape tapeID2
```
filecount is the number of files in the list, and byteCount is the space that is occupied in the system when all files from the list are recalled.
By default the collection file is ordered by the tape ID in ascending order. The disk list file is in the first position. The recall is done in order of the entries, thus the first entry has the highest recall priority, and last entry has the lowest recall priority. You can manually rearrange the order to suit your environment.
You can specify a collection file on the filelist option of the dsmrecall command.

dsmrecall generates three kinds of ordered list files: tape files, a disk file, and a collection file — Figure 2. Generation of list files