CHAPTER 12 Managing BACKUP.UNET Catalogs

Previous Chapter : Next Chapter : Top


Table of Contents


BACKUP.UNET uses a number of tables to control and maintain the networked backup environment. These tables together make up the BACKUP.UNET catalog.


Considerations For Using Catalogs

The tables are used by the BRP and/or IOP daemons. The majority of the tables are maintained locally and are therefore different on each host. Three tables Sys, SysPool, and Pool are global tables maintained by the syncro servers and as such, should be identical on every host.

When the daemons are started, the tables are checked for consistency and the global tables are synchronized with the syncro server. Any modification to global tables is first made to the syncro server hosts and then to the other hosts when they access the tables. For more information on the syncro server hosts, see Chapter 14, BACKUP.UNET Daemon Reference.

A number of utility programs exist that can be used to synchronize, repair, and edit these tables in the event a problem should arise. For more information on these utilities, see page 12 6, BACKUP.UNET Table Repair Procedures and Chapter 13, BACKUP.UNET Utility Reference.

Disk Space Considerations

The BACKUP.UNET catalog can take up a large amount of disk space. Space consumed varies, based on the number of files that are backed up. A good rule of thumb is 112 bytes per file for a full backup and 40 bytes per file for an incremental backup. For example, if BACKUP.UNET were to back up 1000 files (five full and 20 incremental backups), the total catalog size would be about 1.5 Mbytes.


Catalog File Structure

Each BACKUP.UNET table in the $NBK/nbk-TABLES' sub directory consists of three files:

For example, 'Pool.dat', 'Pool.lok', and 'Pool.idx' make up the pools table.


Contents of the nbk-TABLES Directory

This directory holds the CISAM tables used by BACKUP.UNET. These tables make up the BACKUP.UNET catalog. The catalog contains online data information that describes backed up data, and information on volume, device, and system management.

The 'nbk-TABLES' directory contains two kinds of information:

WARNING: Do not edit the contents of these tables. Changing the contents of these tables can crash the system. If you think these tables need to be changed, contact MTI Customer Service for assistance.

Table 12-1 Global Tables

---------------------------------------------------
Table Name  Description                     Used By  
---------------------------------------------------
Sys         System table. Containseach      BRP/IOP  
            system in BACKUP.UNET.                   

Pool        Pools table. Contains each      BRP/IOP  
            available pool.                          

SysPool     System to Pool table. Relates   BRP/IOP  
            which system services which              
            pool.                                    
---------------------------------------------------

.

Table 12-2 Local Tables

-------------------------------------------------------------
Table Name  Description                               Used By  
-------------------------------------------------------------
BFI         Backup File Instance table. Stores        BRP      
            the backup file instance and is the                
            basis of the catalog. Stores the                   
            INODE information and which vol                    
            ume of a dump the file is on.                      

Link        Link table. This table points to the      BRP      
            BFI record where the data is stored                
            if a file is linked.                               

Overflow    Overflow table. If the filename is        BRP      
            too long for the BFI, the remaining                
            part of the name is stored in this                 
            table.                                             

Dumpseq     Dump Sequence table.Contains a            BRP      
            list of files logically associated with            
            a backup, by directory and in alpha                
            betical order. BRP                                 
             Event                                             

Event       Event table. Records the history of       BRP      
            all backups that have been run.                    

DumpVol     Dump Volume table. Relates the            BRP      
            volumes to the backup, and which                   
            segment was used.                                  

Vol         Volume table. Records all the details     IOP      
            of a backup volume. IOP                            
             Dev Device table. Contains the                    
            devices available on this host.                    

Dev         Device table. Contains the devices        IOP      
            available on this host.                            

PoolDev     Pool to Device table. Associates          IOP      
            devices to pools.                                  

Operator    Operator table. Contains the users        BRP      
            who are operators and administra                   
            tors for this host.                                

Configure   Configuration table. Maintains con        BRP      
            figuration cards used by the com                   
            mand line interface.                               
-------------------------------------------------------------

Controlling Catalog Storage Space

BACKUP.UNET stores information about its backups in online tables. These tables contain data and information about backup volumes used in a backup cycle. To manage catalog storage space you can offline the backup information or use the nbk.sqztbl utility.

Offlining Catalog Information

Between the time a backup volume is created and it is recycled (if ever), you may want to free disk space used by backed up data files. This process is called offlining. You can offline backup information either automatically or manually.

Onlining Catalog Information

To recreate the data information in the tables once it has been offlined, the volumes must be explicitly onlined. Onlining is the process of recreating the the backup data information for a specified backup from the appropriate backup volumes. You must online the backup data information before you can restore from it.


BACKUP.UNET Table Repair Procedures

Three BACKUP.UNET utilities, NBKdb_index, NBKdb_fixup, and dbcheck help you repair or rebuild damaged or corrupted table index files.

Checking Table Integrity

To view the contents of a table or check the validity of a table that you suspect may be out of sync, use the dbcheck utility and specify the -i option.

Rebuilding a Corrupted Index

To rebuild an index or regenerate empty tables (both data, '.dat' and index, '.idx' portions) use the NBKdb_build utility (refer to page 13-7 for more information). During BACKUP.UNET installation, an index is built automatically.

If an index (.idx) file is damaged or corrupted to the extent that it cannot be used, or cannot be repaired with the dbcheck utility, rebuild the index using the following steps:

  1. Bring down the BRP and IOP daemons.
  2. Change to/access the $NBK/nbk-TABLES directory
  3. Remove the '.idx' file (for example, 'Pool.idx').
  4. Use the NBKdb_index utility to write a new table index file (for example, 'Pool.idx').
  5. Use the dbcheck utility with the -ly option to rebuild the contents of the newly created index file.
  6. If the table that was corrupted is either BFI or Event, use NBKdb_fixup to set the index pointer.
  7. Check the table permissions. Both '.idx' and '.dat' should be owned by root and the BACKUP.UNET group, and have permissions 664.

Repairing a Faulty Index

If a host stops abruptly or the daemons are arbitrarily killed, then the index (.idx) may be out of sequence with the data (.dat). Use the dbcheck utility with the -y option, specifying the table to be repaired.

Note: If an '.idx' file is damaged or corrupted to the point where it cannot be used, or cannot be repaired with the dbcheck utility, you can rebuild the file using the NBKdb_index utility (refer to page 12 6).

  1. Check the $NBK/nbkLOG file to verify that an error was detected on daemon initialization.
  2. Ensure that all daemons have stopped and no odd processes still exist (for example, dds, iod, or frp). You can use the UNIX ps command.
  3. Use the UNIX ipcs command to list any shared memory and semaphores currently active in the system. Remove any semaphores and shared memory using the UNIX ipcrm command.
  4. Ensure you are in the $NBK/nbk-TABLES directory. Use the dbcheck utility with the -y option and specify the name of the table index to be rebuilt. You can repeat this command for each table, but usually the BFI table is the table that needs repair.
  5. If you specified a tablename of either BFI or Event for the dbcheck utility, run the NBKdb_fixup utility.