Re: pg_clog - Mailing list pgsql-admin

From Edoardo Ceccarelli
Subject Re: pg_clog
Date
Msg-id 2c5b22d10612181026i21d37dear75230439b7649a7a@mail.gmail.com
Whole thread Raw
In response to Re: pg_clog  ("Shoaib Mir" <shoaibmir@gmail.com>)
List pgsql-admin
Thanks for the help, I will probably try this but I would like to know how all of this was originated:

I am posting here the evolution of my db, hoping that somebody can help:

- I upgraded from 7.x to 8.1.5 about 20 days ago
- the db is about 2 / 3 giga and the application that works on it has not been changed across pg versions(apart from the jdbc driver)
- db is in production, with an average of 2000 web sessions per day launching several queries and everything was working perfectly

last night I noticed that the usual LARGE OBJECT "cleaning" wasn't been executed with the new version due to a password problem
the java code that does the LO cleaning is this:

                        db1 = DriverManager.getConnection("jdbc:postgresql://mydatabase", "zzzzzzzzz", "xxxxxxxxx");
                        Statement stat = db1.createStatement();
                        ResultSet rs = stat.executeQuery("select loid, imgoid from pg_largeobject LEFT JOIN images on pg_largeobject.loid=images.imgoid group by loid, imgoid");
                        // query semplice: select * from pg_largeobject LEFT JOIN images on pg_largeobject.loid= images.imgoid

                        db2 = DriverManager.getConnection("jdbc:postgresql://mydatabase", "zzzzzzzzz", "xxxxxxxxx");

                        // fetch the large object manager
                        LargeObjectManager lom = ((org.postgresql.PGConnection) db2).getLargeObjectAPI();
                        while (rs.next()) {
                            count++;
                            imgoid = rs.getInt("imgoid");
                            loid = rs.getInt("loid");

                            if (imgoid == 0) {
                                ccount++;
                                lom.delete(loid);
                            }
                        }
                        stat.close();


after I have activated this code I started noticing these problems (database won't reindex, dump doesn't work)
Does anybody knows if this way of dealing with LO's has changed from 7.x to 8.1.5?
Maybe this is not related at all with clog file problems?

any help apreciated
Edoardo


On 12/18/06, Shoaib Mir <shoaibmir@gmail.com> wrote:
You can try creating the missing clog file as 256K of zeros, once that is done you can now try going through the pg_dump process again.

This suggestion might help you as well --> http://archives.postgresql.org/pgsql-admin/2003-02/msg00263.php

---------------------
Shoaib Mir
EnterpriseDB ( www.enterprisedb.com)

On 12/18/06, Edoardo Ceccarelli <eddy@axa.it> wrote:
Hi,

just encountered this error trying to dump my db:
any ideas?

Thank you


pg_dump: ERROR:  could not access status of transaction 1629514106

DETAIL:  could not open file "pg_clog/0612": No such file or directory

pg_dump: SQL command to dump the contents of table "annuncio400" failed: PQendcopy() failed.

pg_dump: Error message from server: ERROR:  could not access status of transaction 1629514106

DETAIL:  could not open file "pg_clog/0612": No such file or directory

pg_dump: The command was: COPY public.annuncio400 (testo, telef1, pref1, rubric, tiprec, uscita, telef2, pref2, giornouscita, meseuscita, annouscita, prezzo, annofabbr, locali, codedi, codcliente, email, tipoannuncio, ka) TO stdout;



pgsql-admin by date:

Previous
From: "Shoaib Mir"
Date:
Subject: Re: pg_clog
Next
From: Joshua Colson
Date:
Subject: Where in the source is the "database system identifier" generated?