Обсуждение: Full backup

Hello :)

    It is important to me to be able to have a full backup of
everything needed by PostgreSQL in case of data loss. Things like hardware
problems or accidental erasure should be kept in mind.

    In order to do this, I've been doing a .tar.gz for the entire
pgsql home, and that has many drawbacks as you all probably know.

    It would be very nice, in my opinion, if the pg_dumpall tool could
include all system tables in the dump, possibly checking for superuser
access before doing that. It seems that only pg_shadow is added.

    Also, it seems that pg_dumpall is unable to pass the -F flag to
the sucessive calls to pg_dump.

    I apologise if this topic has already been raised.

[]'s                                    HHHHHHH . HHHHHHHHHHHHH          //
                                       HHHHHHH .|i HHHHRIMHARIHH       ,':
  F e l i p e                         HHHH --+~  ``\ HHHHHHHHHHHH      ; :
                                      HH  _`;      ``\_ HHHHHHHHH      ; ;
       (H r i m h a r i)             HHH`' `~'`.       ````---..___.  ,';
                                     HHHHHHHHHH \                   \//'
              C a r a s s o          HHHHHHHHHHH `|                   \
                                     HHHHHHHHHHH ,   \-\______  /~|   ;
                                      HHHHHHHHHH ;:\ :` HHH  / /   ;  :`
                                      HHHHHHHHH ;:` ;: HHHH |_ \    -\\
  hrimhari@rdc.puc-rio.br              HHHHHH +/,  _;:` HHHH _//     //
  wwwusers.rdc.puc-rio.br/hrimhari/     HHHH '  `  ~  ` HHHH ~ `    ~^-
  /"\
  \ /  Campanha da fita ASCII - contra mail html e reply nao editado
   X   ASCII ribbon campaign - against html mail and non-edited reply
  / \

Felipe Carasso writes:

>     It would be very nice, in my opinion, if the pg_dumpall tool could
> include all system tables in the dump, possibly checking for superuser
> access before doing that. It seems that only pg_shadow is added.

There is hardly ever a reason to save the full content of the system
catalogs, because the "standard" content would be added back by initdb and
the local modifications are saved by pg_dump in the form of CREATE
SOMETHING commands.  At least that *should* be the case.  Is there a
particular case you are having trouble with?

>     Also, it seems that pg_dumpall is unable to pass the -F flag to
> the sucessive calls to pg_dump.

That would be nice, but it's tricky to work out how to make sure you can
restore everything in one piece.  As it is, you could have a piece of text
file from pg_dumpall proper and a bunch of compressed archives or some
such created by the pg_dumps.  Maybe pg_dumpall needs to be taught about
the several formats that pg_dump now offers.

--
Peter Eisentraut   peter_e@gmx.net   http://funkturm.homeip.net/~peter