Beta2 Vacuum and pg_dump failures and mangled databases - Mailing list pgsql-hackers

From Frank Joerdens
Subject Beta2 Vacuum and pg_dump failures and mangled databases
Date
Msg-id 3A5F3945.DC213B19@joerdens.de
Whole thread Raw
Responses Re: Beta2 Vacuum and pg_dump failures and mangled databases  (Frank Joerdens <frank@joerdens.de>)
Re: Beta2 Vacuum and pg_dump failures and mangled databases  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
First I tried to dump out a database like:

frank@limedes:~ > pg_dump mpi > dump.mpi
getTables(): relation 'institute': 6 Triggers were expected, but got 0

The database mpi does contain a table 'institute' and a few foreign key constraints. Then
I tried to dump another database, as in:

postgres@limedes:~ > pg_dump intranet > dumptest
getTables(): relation 'institute': 6 Triggers were expected, but got 0

The database intranet has _no_ table 'institute' and no foreign key constraints.

Then I had a look via psql at intranet and it turns out that it shows up as the database
mpi mangled into the database intranet, contentwise; i.e. it doesn't only show the tables
that are in intranet but also those that belong to mpi?! Then I look at _any_ of the
databases in this Postgres installation, they show up as mangled together with mpi?! When
I try to vacuum any of those databases, I always get:

[ . . . stuff that looks normal . . . ]Index pg_class_oid_index: Pages 2; Tuples 138: Deleted 45. CPU 0.00s/0.00u sec.
NOTICE:  Index pg_class_oid_index: NUMBER OF INDEX' TUPLES (138) IS NOT THE SAME AS HEAP'
(205).       Recreate the index.
NOTICE:  Index pg_class_relname_index: Pages 4; Tuples 138: Deleted 44. CPU 0.00s/0.00u
sec.
NOTICE:  Index pg_class_relname_index: NUMBER OF INDEX' TUPLES (138) IS NOT THE SAME AS
HEAP' (205).       Recreate the index.
ERROR:  Cannot insert a duplicate key into unique index pg_class_relname_index

However, if I use another client, i.e. not psql, but a web app, then I do still have
access to the contents of, for instance, the intranet database.

Restarting the server didn't make a difference.

Does this make any sense to anyone?

Regards, Frank


pgsql-hackers by date:

Previous
From: The Hermit Hacker
Date:
Subject: Re: Beta4 for GiST? (Was: Re: AW: Re: GiST for 7.1 !! )
Next
From: Thomas Lockhart
Date:
Subject: Re: FWD: bizarre behavior of 'time' data entry