Re: reconstructing database

From: akanksha kulkarni <akanksha_kul(at)yahoo(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: reconstructing database
Date: 2004-10-06 16:27:35
Message-ID: 20041006162736.45531.qmail@web52905.mail.yahoo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Hi Tom,

I am still not able to recover my old postgres
database.

As postmaster was not starting due to invalid
checkpoints, I ran pg_resetxlog utility . After that
postmaster started, but I am not getting complete
data.
I made pg_dump of database but that comes out to be
very small file. I am sure most of the data is not
there.

Also most of the tables are showing empty when I
connect to database.

I have old /data directory with me.

Is there any other way to extract data from raw data
files in /data/base.

I have found one of your posting related to this in
DBForums which is as below. Can you explain a little
bit more about 0x55 stuff.

Or if you know of some tool/script/postgres recovery
specialist(person) to extract data from raw /data
files, PLEASE let me know.

Thanks for your help.

********* your reply to mlw's posting
**********************
Re: Recovery tools

--------------------------------------------------------------------------------

mlw writes:
> Just suppose that all the log files are gone, and
the only thing left is
> some of the files in the /data directory. Is there
any way to scan this
> data and dump it to a file which could
subsequently be used with a "COPY
> FROM STDIN" on a new database?

There aren't separate tools, and I'm not sure there
could or should be.
What I'd do in that situation is:
* pg_resetxlog to get a minimally valid xlog
* if clog is missing, gin up files containing 0x55
everywhere
(to make it look like every transaction has committed
--- or
put 00 everywhere if you'd rather assume that recent
transactions all aborted)
* start postmaster, look around, fix problems until I
can pg_dump.

AFAICS, you can make tools that work at the page/item
level (like
pg_filedump, see http://sources.redhat.com/rhdb/), but
there is hardly
any scope for doing anything intermediate between that
and a full
postmaster. There's no hope of decoding the contents
of a tuple without
access to the table's tuple descriptor, which means
you need most of the
system catalog mechanisms; plus you'd need the I/O
routines for the
datatypes involved. Might as well just use the
postmaster as your data
inspection tool.

regards, tom lane
***************************************************

--- Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> > 5. I then tried uninstalling new version by giving
> > command gmake clean. I also moved the new data
> > directory to temp directory.
>
> "gmake clean" is not going to do an uninstall.
>
> > 6. I reinstalled old version of postgres from RH
> > package manager.I copied old data directory of
> > previous install of older version of postgres and
> then
> > tried running postmaster but it didn't ran saying
> > there is newer version of postgres.
>
> I think you had both sets of executables installed
> at once and somehow
> you were picking up the wrong one. There is nothing
> wrong with your old
> data directory, it's the executables you need to get
> straight.
>
> regards, tom lane
>


_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com

Browse pgsql-admin by date

  From Date Subject
Next Message simon 2004-10-06 21:36:01 Re: Multiple WALs for PITR ?
Previous Message David.Jacques 2004-10-06 14:41:40 Adaptation of instnwnd.sql to Pgsql