do-dont

一套客户/服务器模式的备份系统代码,跨平台,支持linux,AIX, IRIX, FreeBSD, Digital Unix (OSF1), Solaris and HP-UX.

        Do-s and Dont-s

ALWAYS
------

        - configure a Startup-Info-Command on the client side
          to save the crucial information for the emergency
          restore after a hard crash losing the filename logfiles
          i.e. the filelists and probably more ...
	- use more than 1 tape (per cartridge set), i recommend
	  at least 3
	- also save the server side's .../var directory after the
	  regular clients' backup, cause it contains crucial files
	  for recovery, i.e. configure the server as client, too.
	  If not only the .../var directory of the server is saved,
	  but also other data residing on the server, the .../var
	  directory should be saved separately after any client's
	  backup. Use the full_backup command with a different
	  client ID supplied with option -W and a different var
	  directory given with option -V and supply the path to
	  the real server's var directory as argument. Example:
	  /path/to/full_backup -W serv-var -V /tmp/foo .../server/var
	  The directory /tmp/foo must already exist. In fact any
	  directory can be used as argument for -V, it may even be
	  removed afterwards. This saving of the server's var
	  directory serves for being able to restore it in case of
	  disaster, i.e. complete server crash. Remember, that the
	  files in this var directory are crucial for restore of
	  any client, especially the file cartridge_order. If this
	  saving is done, you are able to do disaster recovery even
	  without any minimum restore information. See HOWTO Q24
	  what to do exactly in case of disaster.
        - run a verify after the first backup with afbackup to 
          make sure, that everything is working correctly
        - write the numbers of the cartridges onto their cases
          or use adhesive labels
	- use a host alias name for the backup server, so it can
	  easily be moved to another machine
        - be happy and forgive me the bugs
	- try to keep things simple. Making things complicated
	  unnecessarily will strike back mercilessly sooner or
	  later, according to Murphy sooner, if not ASAP ... or
	  even AYAP


        ... to be continued


NEVER
-----

        - kill the following client-side afbackup-related programs
          with signal -9 (== -KILL): full_backup, incr_backup,
          update_indexes. It may take a while until the program has
          cleaned up but it does clean up and it will terminate.
          If this is taking more than 10 minutes, then you might
          THINK of kill -9. If you can see a process afclient or
          afbackup, that is a subprocess of one of the named ones,
          kill this one brutally. This is quite safe as it doesn't
          maintain persistent data. But first watch the processor
          time consumed by the processes ! kill -9 on the server is
          save for all programs, but the client side is a different
          story ! Be patient !
        - use fewer cartridges than with a capacity of at least
          two times the size required for one full backup including
          subsequent incremental backups. Otherwise you will receive
          mails from the backup system complaining, that no more
          space is available on the configured tapes, requesting
          to mark tapes for reuse i.e. overwrite or increase the
          number of available tapes.
	- change the process and/or unprocess program, if not
	  enough hidden files with the same name like the indexes
	  (with a leading dot) have been created by afbackup-3.2.7
	  or higher. These files contain the command to unprocess
	  the related index. If it is missing, the program in the
	  configuration file is tried, that might fail, if it has
	  been modified.
          See also FAQ Q27 for more information.

        ... to be continued