Administration Guide

Recovery Method: Version Recovery

Version recovery using the BACKUP command in conjunction with the RESTORE command puts the database in a state that has been previously saved. You use this recovery method with non-recoverable databases (that is, databases for which you do not have archived logs). You can also use this method with recoverable databases by using the WITHOUT ROLLING FORWARD option.

In this section, planning considerations and how to invoke the specific utilities or commands to carry out the method are reviewed. Then, any concepts or related issues that allow effective use of this method are presented.

The following topics provide additional information:

• "Backing Up a Database"
• "Restoring a Database"
• "Recovery History File Information".

Backing Up a Database

To make a backup copy of the database, you use the BACKUP command or the Control Center. Within the Control Center, you select the database to be backed up and then select the backup action.

Figure 53. Creating a Database Image

In a partitioned database system, you back up database partitions individually using the BACKUP DATABASE command. The operation is local to the database partition server where you issue the command. You can, however, issue db2_all from one of the database partition servers in the instance to submit the BACKUP command on a list of servers, which you identify by their node number. If you do this, you must back up the catalog node first, then back up the other database partitions. You can also use the Control Center to backup database partitions.

In a partitioned database system, you can use the LIST NODES command to determine the list of nodes (database partition servers) that have user tables on them. Because this recovery method does not support roll-forward recovery, regularly back up the database on this list of nodes.

In a distributed request system, the BACKUP and RESTORE commands apply to the distributed request database and the metadata stored within that database catalog (wrappers, servers, nicknames, etc.). Data source objects (tables and views) are not backed-up or restored unless those objects are stored in the distributed request database.

You must keep in mind the recovery method to be used. The following sections provide requirements and other considerations that apply to this task:

• "Planning to Use the BACKUP Command"
• "Invoking the BACKUP Command"
• "Backup Images Created by BACKUP".

Planning to Use the BACKUP Command

Your planning considerations should include:

• You must have SYSADM, SYSCTRL, or SYSMAINT authority to use the BACKUP command.
• The database may be local or remote. The backup remains on the database server unless a storage management product such as Tivoli* Storage Manager (TSM) is used.
• You can back up a database to a fixed disk, a tape, or a location managed by TSM or another vendor storage management product. See Tivoli Storage Manager for information on TSM.

  Under OS/2, you can also back up to diskette or to a user exit.

  Note:

  In OS/2, when backing up a database online to a user exit, note that the database will be quiesced before the backup starts. As such, the backup will wait for all transactions to either commit or rollback before it starts. While the backup is running, all new transactions will wait until the backup is complete, and, once the backup is completed, all transactions will continue processing as usual.

• When creating a backup image (or restoring a backup image) the buffer size is 1024 pages (of 4 KB size). This is important if you are using tape as your backup device. If you are using variable block sizes you must lower your buffer size to a range which your tape drive uses.
• Under supported Windows operating systems, you can back up to diskette.
• User most versions of Linux, using DB2's default buffer sizes for backup and restore to a SCSI tape device results in the error message: SQL2025N, reason code "75". To prevent the overflow of Linux internal SCSI buffers, use this formula:

     bufferpages <= ST_MAX_BUFFERS * ST_BUFFER_BLOCKS / 4
  

  where bufferpages is the value of either backbufsz or restbufsz. ST_MAX_BUFFERS and ST_BUFFER_BLOCKS are defined in the Linux kernel under drivers/scsi directory.
• Under OS/2, a user exit is used when backing up to tape because the operating system has no native tape support.

  Under UNIX-based operating systems and Windows NT, native tape support is available.

  Note:

  If you use a variable block size with your tape devices, ensure that the DB2 buffer size is either less than or equal to the maximum variable block size that the device is configured for. Otherwise, the backup will succeed but the resulting image is not guaranteed to be recoverable.

• Multiple files may be created to contain the backed up data from the database.
• At the completion of an online backup, the active log is closed and written to disk.
• In a partitioned database environment, an offline backup uses an exclusive connection to the database at that database partition server (that is, the operation requires an exclusive connection to the database partition), so no other application can be connected to the database partition. When you do an offline backup of the catalog node, there can be no activity on the entire database, including backups of the database on non-catalog database partition servers. You can use db2_all to back up the database, but you must ensure that the catalog node is backed up first. After the catalog node is backed up, the other database partitions can be backed up at the same time.
• In a partitioned database system, you should also keep a copy of the db2nodes.cfg file with any backup copies you take, as protection against possible damage to this file.

If you have tables that contain DATALINK columns, also see Backup Utility Considerations.

To use tape devices, DB2 users on SCO UnixWare 7 must specify BUFFER to be 16. The default value of BUFFER is 1024 pages. If BUFFER is set to zero, the database manager configuration parameter backbufsz must be set to 16.

Planning to Use Tapes When Backing Up

When you back up your table space or database, you must correctly set your block size and your buffer size. This is particularly true when you are using a variable block size (for example, on AIX when the block size has been set to zero).

There is a restriction on the number of FIXED block sizes that can be used when backing up. This restriction exists because DB2 writes out the back up image header as a 4 KB block. The only fixed block sizes DB2 supports are 512, 1024, 2048, and 4096 bytes. If you are using a fixed block size, you can specify any buffer size for the backup. However, you may find that your back up will not complete successfully if the fixed block size is not one of those mentioned here.

If your database data is large, using the fixed block sizes mentioned above will mean you back up will take a long time. You may want to consider using a variable block size.

When using a variable block size, you must also specify a buffer size in the BACKUP command that is less than, or equal to, the maximum limit for the tape device you are using. The buffer size must be equal to the maximum block size limit of the device being used if you would like the best performance.

You should be aware that restoring from a back up image where the block size is variable may return an error. If this happens, you might need to rewrite the image using an appropriate block size. An example of doing this in AIX follows:

   tcl -b 0 -Bn -f /dev/rmt0 read > backup_filename.file
   dd if=backup_filename.file of=/dev/rmt0/ obs=4096 conv=sync

This dumps the backup image to a file called "backup_filenam.file" The "dd" command then dumps the image back onto the tape using a block size of 4096.

The complication with this method of correction occurs if the image is too large to dump to a file. One possible solution with such a large image, is to use the "dd" command to dump the image from one tape device to another. This will work as long as the image does not span more than one tape. When using two tape devices, the "dd" command is:

   dd if=/dev/rmt1 of=/dev/rmt0 obs=4096

If using two tape devices is not possible, you may be able to dump the image to a raw device using the "dd" command, and then dump the image from the raw device to tape. The difficulty using this method is that the "dd" command must keep track of the number of blocks dumped to the raw device. The number of blocks dumped needs to be specified when the image is moved back to tape. When the "dd" command is used to dump the image from the raw device to tape, the command dumps the entire size of the raw device to tape. The "dd" command cannot tell how much of the raw device is used to hold the image.

When using the BACKUP command you will need to know the maximum block size limit for the tape device you are using. Here are some examples:

Device	Attachment	Block Size Limit	DB2 Buffer Size Limit (in 4 KB pages)
8 mm	scsi	131 072	32
3420	s370	65 536	16
3480	s370	65 536	16
3490	s370	65 536	16
3490E	s370	65 536	16
7332 (4 mm)*	scsi	262 144	64
3490e	scsi	262 144	64
3590**	scsi	2 097 152	512
3570 (magstar MP)		262 144	64

Notes:

1. * The 7332 does not implement a block size limit. 256 KB is simply a suggested value. Block size limit is imposed by the parent adapter.

2. ** While the 3590 does support a 2 MB block size, you might experiment with lower values (like 256 KB) provided the performance is adequate for your needs.

3. Check your device limit with the device documentation and/or with the device vendor.

Invoking the BACKUP Command

The following considerations are useful when running the BACKUP command:

• You must start the database manager (DB2START) before running the BACKUP command or API. When using the Control Center, you do not need to explicitly start the database manager.
• When using the command, API, or task under Control Center, you must specify a database alias name, not the database name itself.
• To reduce the amount of time required to complete a backup:
  • Increase the value of the PARALLELISM parameter.

    Using this parameter can dramatically reduce the amount of time required to complete the backup. The PARALLELISM parameter defines the number of processes or threads that are started to read data from the database. Each process or thread is assigned to back up a specific table space. When it completes backing up the table space, it requests another. You should note, however, that each process or thread requires both memory and CPU overhead: for a heavily loaded system, you should leave the PARALLELISM parameter at its default value of 1.

  • Increase the backup buffer size.
  • Increase the number of buffers.

    If you use multiple buffers and I/O channels, you should use at least twice as many buffers as channels to ensure that the channels do not have to wait for data. The size of the buffers used will also contribute to the performance of the backup operation. The ideal backup buffer size should be a multiple of the extent size for the table space(s).

    If you have multiple table spaces with different extent sizes, specify a value that is a multiple of the largest extent size.

    You may specify the number of pages to use for each backup buffer when you invoke the BACKUP command. The minimum number of pages is 16. If you do not specify the number of pages, each buffer will be allocated based on the database manager configuration parameter backbufsz. If there is not enough memory available to allocate the buffer, an error will be returned.

    See Chapter 32, Configuring DB2 for more information on this configuration parameter.

  • Use multiple target devices.
• In OS/2, when backing up a database to removable media, such as tape, the database manager writes information to media volume 1. Once the first media is in the drive, do not remove the media unless the operating system backup facility prompts you for media 2.
• You cannot back up a database that is not in a usable state except for a database in the backup pending state.
  • If a database is in a partially restored state due to a system crash during any stage of restoring the database, you must successfully restore the database before you can back it up.
  • If a database was created with a previous release of the database manager and the database has not been migrated, you must migrate the database before you can back it up.

    See Appendix B, Planning Database Migration, for information about migrating a database.

  • If any of the table spaces in a database is in an "abnormal" state, you cannot back up the database, unless it is in the backup pending state.
• If a system crash occurs during a critical stage of backing up a database, you cannot successfully connect to the database until you re-issue the BACKUP command.
• The BACKUP command provides a concurrency control for multiple processes that are making backup copies of different databases. The control keeps the backup target device open until the entire backup process has ended.

  If an error occurs during a backup process and the open container cannot be closed, other backup processes to the same target drive may receive access errors. To correct any access errors, you must completely exit the backup process that caused the error and disconnect from the target device.

• If you are using the BACKUP command for concurrent backup processes to tape, ensure that the processes do not target the same tape.

Backup Images Created by BACKUP

Backup images are created at the target specified when you call the BACKUP command:

• In the directory for disk or diskette backups
• At the device specified for tape backups
• At a Tivoli Storage Manager (TSM) server
• At another vendor's server
• For OS/2, through the use of a user exit.

The recovery history file is updated automatically with summary information whenever you carry out a backup or restore of a full database. This file can be a useful tracking mechanism for restore activity within a database. This file is created in the same directory as the database configuration file. For more information on the recovery history file, see Recovery History File Information.

In UNIX-based environments, the file name(s) created on disk will consist of a concatenation of the following information, separated by periods; on other platforms a four-level subdirectory tree is used:

Database alias

A 1-to-8 character database alias name that was supplied when the backup command was invoked.

Type

Type of backup taken, where: "0" is for full database. "3" is for a table space backup. "4" is for a backup generated by the LOAD...COPY TO command.

Instance name

A 1-to-8 character name of the current instance of the database manager that is taken from the DB2INSTANCE environment variable.

Node number

The node number.

Catalog node number

The node number of the database's catalog node.

Time stamp

A 14-character representation of the date and time the backup was performed. The timestamp is in the format yyyymmddhhnnss, where:
yyyy is the year (1995 to 9999)
mm is the month (01 to 12)
dd is the day of the month (01 to 31)
hh is the hour (00 to 23)
nn is the minutes (00 to 59)
ss is the seconds (00 to 59)

Sequence number

A 3-digit sequence number used as a file extension.

In UNIX-based operating systems, the format would appear as:

Database alias.Type.Instance name.NODEnnnn
.CATNnnnn.timestamp.number

On other operating systems, the format would appear as:

Database alias.Type\Instance name.NODEnnn
\CATNnnn\yyyymmdd\hhmmss.number

For example in UNIX-based environments, a database named STAFF on the DB201 instance may be backed up on disk to a file named:

STAFF.0.DB201.NODE0000.CATN0000.19950922120112.001

For tape-directed output, file names are not created; however, the above information is stored in the backup header for later verification purposes.

Notes:

1. If you want to use tape media for database backup and restore operations, a tape device must be available through the standard operating system interface. On a large partitioned database system, however, it may not be practical to have a tape device dedicated to each database partition server. You can connect the tape devices to one or more TSM servers, so that access to these tape devices is provided to each database partition server.

2. On a partitioned database system, you can also use products that provide virtual tape device functions, such as REELlibrarian 4.2 or CLIO/S. You use these products to access the tape device connected to other nodes (database partition servers) through a pseudo tape device. Access to the remote tape device is provided transparently, and the pseudo tape device can be accessed through the standard operating system interface.

Displaying Backup Information

There is a backup utility to display information regarding the backup images that exist. The name of this utility is db2ckbkp and it allows you to:

• Test the integrity of a backup image and determine whether it can be restored or not.
• Display the information about the back that is stored in the backup header.

You must have read permissions on the backup images you specify when using this utility.

To simply verify the existence of a backup image, you can use the utility as follows:

   db2ckbkp STAFF.0.DB201.NODE0000.CATNOOOO.19950922120112.001

The output from this utility is similar to:

   [1] Buffers processed:  ##
 
   Image Verification Complete - successful.

Refer to the Command Reference for additional information on this utility.

Restoring a Database

The following sections provide requirements and other considerations that apply to the RESTORE command:

• "Planning to Use the RESTORE Command"
• "Invoking the RESTORE Command"
• "Redefining Table Space Containers During RESTORE"
• "Restoring to an Existing Database"
• "Restoring to a New Database".

Figure 54. Restoring a Database Using a Backup Image

Planning to Use the RESTORE Command

You should consider the following:

• You must have SYSADM, SYSCTRL, or SYSMAINT, authority to restore to an existing database from a full database backup. To restore to a new database, you must have SYSADM or SYSCTRL authority.
• You can only use this command if the database has been previously backed up with the BACKUP command.
• If you use the Control Center, you cannot restore backups that were taken previous to the current version of DB2.
• In OS/2, the RESTORE command can call a user exit program only if a user exit program was used to backup the database.
• You can choose at the time of the restore which type of restore is to be carried out. You can select from the following types:
  • A full restore of everything from the backup
  • A restore of only the recovery history file
  • A subset of the table spaces in the backup.
• The RESTORE command can use the Tivoli Storage Manager (TSM) utility, and any restrictions of that utility should also be considered. (See Tivoli Storage Manager.)
• Another vendor storage management product may also be used if that product was used to store the original backup.
• A database restore requires an exclusive connection: that is, no applications can be running against the database when the task is started. Once it starts, it prevents other applications from accessing the database until the restore is completed. A table space restore can be done online.
• The database may be local or remote.
• If the WITHOUT DATALINK option is not specified, and the DB2 Data Links Manager containing the DATALINK data is not available, then the restore operation will fail.

  If the option is specified and the DB2 Data Links Manager containing the DATALINK data is not available, all table spaces containing tables with DATALINK values on the unavailable server are placed in the RESTORE PENDING state. These table spaces must be restored again when the Data Links server becomes available.

• To use tape devices, DB2 users on SCO UnixWare 7 must specify BUFFER to be 16. The default value of BUFFER is 1024 pages. If BUFFER is set to zero, the database manager configuration parameter backbufsz must be set to 16.

If you have tables that contain DATALINK columns, see both Restore and Rollforward Utility Considerations and Restoring Databases from an offline Backup without Rolling Forward.

Invoking the RESTORE Command

The following considerations are useful when running the RESTORE command:

• The database manager must be started before restoring a database.
• The database to which you restore the data may be the same one as the data was originally backed up from, or it may be different. You may restore the data to a new or an existing database.
• During the restore procedures, you have the ability to optionally select to use multiple buffers to improve the performance of the restore procedure. The multiple internal buffers may be filled with data from the backup media.

  You may specify the number of pages to use for each restore buffer when you invoke the RESTORE command. The value you specify must be a multiple of the number of pages that you specified for the backup buffer. The minimum number of pages is 16. If you do not specify the number of pages, each buffer will be allocated based on the database manager configuration parameter restbufsz. If there is not enough memory available to allocate the buffer, an error will be returned.

  See Chapter 32, Configuring DB2 for more information on this configuration parameter.

• The TAKEN AT parameter of the RESTORE DATABASE command requires the timestamp for the backup. The timestamp can be exactly as it was displayed after the completion of a successful BACKUP command, that is in the format yyyymmddhhmmss.

  You can also specify a partial timestamp. For example, assume that you have two different backups with the timestamps 19971001010101 and 19971002010101. If you specify 19971002 for TAKEN AT, the 19971002010101 backup is used.

  If TAKEN AT is not specified, there must only be one backup on the source media.

• The backup copy of the database to be used by the RESTORE command can be located on a fixed disk, a tape or a location managed by the Tivoli* Storage Manager (TSM) utility or another vendor storage management product. See Tivoli Storage Manager for information on TSM.

  If you use TSM and do not specify the TAKEN AT parameter, TSM retrieves the latest backup copy.

  Under OS/2, the backup copy of the database could also be located on diskette or through a user exit.

  Under supported Windows operating systems, the backup copy of the database could also be located on diskette.

• On Windows operating systems, and on OS/2, you must specify only the device letter when using the "TO target-directory" clause. If a longer path is specified, an error is returned.
• Once the RESTORE command starts, the database is not usable until the RESTORE command completes successfully.
• If a system failure occurs during any stage of restoring a database, you cannot connect to the database until you reuse the RESTORE command and successfully complete the restore.
• If the code page of the database being restored does not match a code page available to an application; or, if the database manager does not support code page conversions from the database code page to a code page that is available to an application; then the restored database will not be usable.

Redefining Table Space Containers During RESTORE

During a backup of a database, a record is kept of all the table space containers in use by the table spaces that are backed up. During a RESTORE, all containers listed in the backup are checked to see if they currently exist and are accessible. If one or more of the containers is inaccessible because of a media failure (or for any other reason), the RESTORE will fail. In order to allow a restore in such a case, the redirecting of table space containers is supported during the RESTORE. This support includes adding, changing, or removing of table space containers.

There are cases in which you want to restore even though the containers listed in the backup do not exist on the system. An example of such a case is where you wish to recover from a disaster on a system other than that from which the backup was taken. The new system may not have the necessary containers defined. In order to allow a RESTORE in this case, the redirecting of table space containers at the time of the RESTORE to alternate containers is supported.

In both situations, this type of RESTORE is commonly referred to as a redirected restore.

You can redefine table space containers through the restore task from within the Control Center. You can also use the REDIRECT parameter of the RESTORE command to specify the redirection. If you are using the Control Center, one way of performing a redirected restore is to use the Containers page of the Restore Database notebook. This page provides function that you can use to add new containers, change the path of an existing container, or remove a container. If, during the process of the restore database operation an invalid container path is detected, the Control Center will prompt you to either change the container path, or remove the container.

Notes:

1. Directory and file containers are automatically created if they do not exist. No redirection is necessary unless the containers are inaccessible for some other reason. The database manager does not automatically create device containers.

2. The ability to perform container redirection on any RESTORE provides considerable flexibility in managing table space containers. For example, even though we do not directly support adding containers to SMS table spaces, you could accomplish this by simply specifying an additional container on a redirected restore. Similarly, you could move a DMS table space from file containers to device containers.

3. Redirected restore is also supported through a number of APIs. Although you could write a program to perform redirected restore for a specific case, these APIs are primarily intended for developers who want to produce a general purpose utility.

Restoring to an Existing Database

You may restore a backup copy of a full database backup to an existing database. To restore to an existing database, you must have SYSADM, SYSCTRL, or SYSMAINT authority. The backup image may differ from the existing database in its alias name, its database name, or its database seed.

A database seed is a unique identifier of a database that remains constant for the life of the database. This seed is assigned by the database manager when the database is first created. The seed is unchanged following a restore of a backup even if the backup has a different database seed. DB2 always uses the seed from the backup.

When restoring to an existing database, the restore task performs the following functions:

• Delete table, index, and long field contents for the existing database, and replace them with the contents from the backup.
• Replace table space table entries for each table space being restored.
• Retain recovery history file unless the one on disk is damaged. If the file on the disk is damaged, the database manager will copy the file from the backup.
• Retain the authentication for the existing database.
• Retain the database directories for the existing database that define where the database resides and how it is cataloged.
• When the database seeds are different:
  • Delete the logs associated with the existing database
  • Copy the database configuration file from the backup
  • If newlogpath is specified on the RESTORE command, set it to the log path (which is specified by the logpath parameter) in the database configuration file.
• When the database seeds are the same:
  • Retain the current database configuration file, unless the file is corrupted, in which case this file will be copied from the backup.
  • Delete the logs if the image is of a non-recoverable database.
  • If newlogpath is specified on the RESTORE command, set it to the value of the logpath database configuration parameter. Otherwise, copy the current log path to the database configuration file. Validate the log path: If it cannot be used by the database, change the database configuration to use the default log path.

Restoring to a New Database

As an alternative to restoring a database to a database that already exists, you may create a new database and then restore the backup of the data. To restore to a new database, you must have SYSADM or SYSCTRL authority.

Note:

The code pages of the backup and the target database must match. If they do not, first create the new database specifying the correct code page, then restore it.

When you restore to a new database, the RESTORE command will perform the following functions:

• Create a new database, using the database name and database alias name that was specified by the target database alias parameter. (If this target database alias was not specified, the RESTORE command will create a database with the name and alias the same as the source database alias parameter.)
• Restore the database configuration file from the backup.
• If newlogpath is specified on the RESTORE command, set it to the value of the logpath parameter in the database configuration file. Validate the log path: If it cannot be used by the database, change the database configuration to use the default log path.
• Restore the authentication type from the backup.
• Restore the database comments from the backup for the database directories.
• Restore the recovery history file for the database.