Re: pg_restore out of memory - Mailing list pgsql-general

From Miguel Ramos
Subject Re: pg_restore out of memory
Date
Msg-id 1470127375.1032.17.camel@miguel.ramos.name
Whole thread Raw
In response to pg_restore out of memory  (Miguel Ramos <org.postgresql@miguel.ramos.name>)
List pgsql-general
Greetings,

To all those who helped out with this problem, I'm sorry it took me so
long to respond. For the record, the matter is solved, at least for us,
but we had to repeat a lot of things to make sure.


First, the "out of memory" problem repeated itself when restoring that
single table, after doing a schema-only restore and removing all
foreign key constraints.

Second, the other weirdness on this table was a couple of constraints
to make sure that the three arrays on each row were of the same length.
But that was not the problem.


It turned out that the backup file was the problem.

Plus, I'm convinced that this was our mistake, someone simply didn't
see an error message during backup or some copy of the file.
Also, we should have already repeated the backup and maybe we didn't.


We did a new custom dump from within latest pgadmin III, and that one
restored just fine (using the same old 9.1.8 pg_restore).

It could have been a bug in pg_dump 9.1.8 since the good backup was
from a newer version. But this didn't seem so likely and because
repeating a full backup would force us to keep the original database on
the server for a few more days (dumps take so long) we didn't do it.

We did however a dump of that single table with pg_dump 9.1.8 and that
one also restored just fine.



Our immediate practical problem is over.
The only complaint would be that the "out of memory" message is
unfriendly.
If you would find useful that we make some additional tests, or some
observation of the file, we would be glad to return the help.


Best Regards,

--
Miguel Ramos


pgsql-general by date:

Previous
From: "hari.prasath"
Date:
Subject: Upserting all excluded values
Next
From: Branden Visser
Date:
Subject: Re: Re: Query planner using hash join when merge join seems orders of magnitude faster