Introduction to the Backup Process

Backing up, or dumping, data is the most basic operation performed with the Backup System. Data must be dumped to tape before it can be tracked in the Backup Database and before it can be restored from tape to the file system. This section provides an overview of the backup process.

Dumping a fileset makes it inaccessible to other file system users for the duration of the dump process. To reduce inconvenience, create a backup version of a fileset (a version with a .backup extension) and dump the backup version rather than the read/write version; this does not interrupt a user's work. Creating a backup version of a fileset, using the fts clone or fts clonesys command described in detail in Making Filesets and Aggregates Available, does make its read/write source fileset unavailable for a short period of time; therefore, you may wish to create backup versions during periods of low system usage, using bos commands to create a cron process to automate the procedure. (See Monitoring and Controlling Server Processes for a description of the bos commands.)

Occasionally, the Backup System cannot access a fileset, perhaps because of a File Server machine or Fileset Server outage. When this happens, it attempts to access the fileset three times over the course of the operation. If it still cannot access the fileset after the third attempt, it omits the fileset from the dump rather than aborting the dump or waiting for the fileset to become accessible. If the access failure occurs during a full dump, the next incremental dump of the fileset includes the entire fileset; if the failure occurs during an incremental dump, the next incremental dump of the fileset includes all files modified since the last successful dump of the fileset. You can set the Tape Coordinator performing the dump to notify you of the omission in its monitoring window (by specifying a value of 1 with the \-debuglevel option of the butc command used to start the Tape Coordinator). The Tape Coordinator's error file also records the fileset's omission.

Following is a summary of the process the system uses to perform a typical backup. The example assumes that a backup is being performed on a Wednesday; the fileset family usersys is to be dumped at the dump level whose name in the dump hierarchy is /sunday/wednesday in this example. Note that the Backup System makes no implied connection between the name of a dump level and the date and time at which a dump at that level is to occur; descriptive dump level names serve merely as reminders to system administrators of when dumps are to be performed.

· The Backup System reads the dump hierarchy in the Backup Database to see if /wednesday is an incremental dump and, if so, to determine which preceding level is its parent. In this example, the /wednesday level is incremental, and /sunday is its parent.

/sunday

/monday

/tuesday

/wednesday

/thursday

/friday

If /sunday were specified, this would be a full dump; the system would copy the complete contents of each fileset in usersys. Because /sunday/wednesday is an incremental dump level, the dump set includes only those files that changed since usersys was dumped at the /sunday level.

· Because /sunday is the parent for /wednesday, the Backup System checks the Backup Database for the date and time of the last dump of usersys at the sunday level.

· The Backup System reads the fileset family usersys in the Backup Database to learn which fileset family entries it contains. The fileset family and its entries were created beforehand with the bak addftfamily and bak addftentry commands. In this example, the entries are

.* .* user.*

.* .* sys.*

· The Backup System scans the FLDB to match the wildcards from each fileset entry and generates a complete list of the filesets to be included in the dump. If duplicates are found, they are not dumped; only one occurrence of any fileset is included.

· The Backup System reads the label on the tape in the drive to verify that the tape name is acceptable and that the tape does not contain an unexpired expiration date.

· The system transfers the list of filesets to be backed up to the appropriate Fileset Server processes, which determine which data in the filesets was modified after the date and time of the last dump at the /sunday level.

· The designated Tape Coordinator puts the gathered data onto tape; the expiration date and other information associated with the dump are stored in the tape's label, and a unique dump ID number is assigned to the dump. If one tape is not large enough to hold the entire dump set, the Backup System prompts the operator to place additional tapes in the drive, as needed.