Database upgrade Pg11 to Pg12 issues

2021-02-03 Thread Simon Windsor

Hi

I have upgraded many Pg databases from 9 to 10 to 11 using the Ubuntu 
pg_upgradecluster command (wrapper to pg_update) without issues, however 
today I upgraded a DB from 11 to 12.


Using the syntax

*/pg_upgradecluser -k -m upgrade 11 main/*

using the latest 12.5 binaries all appeared to work well, and for a 330G 
DB this took 2 minutes. On starting the DB using


systemctl start postgresql@12.service

the DB started, but the screen showed the default directory was 
*//var/lib/postgresql/11/main/*. Checking the open database with


/*show data_directory; */gave /*/var/lib/postgresql/12/main*/

Then using /*lsof | grep '11/main'*/ gave no files, but/*lsof | grep 
'12/main'*/ gave over 30K files, Checking the directory sizes suggests 
that 11/main is 300G and 12/main is 40G


So all appears OK, with files hard linked between 11/main and 12/main, 
and I assume the issue relates to the PGDATA/postgresql.auto.conf file 
being just copied from the original 11/main with data_directory being 
set to /var/lib/postgresql/11/main.


If I where to run /*pg_dropcluster 11 main*/ to remove the old database 
and conf files, will this destroy my running Pg12 database with hard 
linked files in 11/main and 12/main? In theory it shouldn't and files 
from 11/main and 12/main will all be under 12/main, I just have doubts.


Is there an easy way to tidy this up?

Simon


--
Simon Windsor

Eml: simon.wind...@cornfield.me.uk


“There is nothing in the world that some man cannot make a little worse and 
sell a little cheaper, and he who considers price only is that man's lawful 
prey.”



Re: Error XX000 After pg11 upgrade

2019-08-15 Thread Simon Windsor

Hi

The Full system used to in an Oracle DB and was ported to Postgres 9.5 
about 2+ years ago, and partitioned using inheritance tables.


Since then pg_upgrade has been used to upgrade to pg10 (with apt upgrade 
to take to 10.5 occasionally).


Last week, pg_upgrade was againn used to upgrade to pg11.4.

Since then, large bulk inserts of configuration changes are failing with 
this Error, but adhoc and small changes are working ok.


The actual error is reported by a Java process

Caused by: org.postgresql.util.PSQLException: ERROR: cache lookup failed 
for type 22079

  Where: SQL statement "insert into configObjectsFull_2019 values (new.*)"
PL/pgSQL function configobjectsfull_insert_trigger() line 28 at SQL 
statement
    at 
org.postgresql.core.v3.QueryExecutorImpl.receiveErrorResponse(QueryExecutorImpl.java:2310)

    at

The DB Tableand Trigger are attached.

As you can see, the id and timestamp are set on insert, and are used to 
determine the partition used.


This issue started after the upgrade to pg11, pg10 and pg9 had no problems.

Any ideas would be appreciated.

Simon

On 15/08/2019 16:31, Tom Lane wrote:

Simon Windsor  writes:

I have just upgraded a Db from pg 10.5 to pg11.4 and almost immediately we
are seeing errors like
ERROR,XX000,"cache lookup failed for type 22079"
When inserting into a a partitioned table. About 30% of inserts are
reporting this error.

Hmm ... can you show the full schema (eg, psql \d+ output) for the
partitioned table and its children?  Is there any pattern to the
failing inserts, eg do they all resolve as inserts to the same
partition(s)?


... And an insert trigger uses the current date to dertmine where the
object is stored.

In other words, this isn't actual partitioning as introduced in v10,
but a hand-rolled equivalent?


How do I determine what type 22079 refers to?

Presumably, the problem is that that type OID *doesn't* refer to anything
any more.  You should be asking "where is this dangling reference coming
from?".  It's possibly hiding in the partitioning expression(s) of
this partitioned table, but there's no way to tell with this amount
of info.

How did you do the upgrade exactly?

    regards, tom lane


--
Simon Windsor

Eml: simon.wind...@cornfield.me.uk
Tel: 01454 617689
Mob: 0755 197 9733


“There is nothing in the world that some man cannot make a little worse and 
sell a little cheaper, and he who considers price only is that man's lawful 
prey.”



objects.sql
Description: application/sql


Error XX000 After pg11 upgrade

2019-08-15 Thread Simon Windsor
Hi

I have just upgraded a Db from pg 10.5 to pg11.4 and almost immediately we
are seeing errors like

ERROR,XX000,"cache lookup failed for type 22079"

When inserting into a a partitioned table. About 30% of inserts are
reporting this error.

How do I determine what type 22079 refers to? I have checked pg_attributes
and pg_classes without success? What would be causing this error?

The basic table layout is

objects
objects_2010
objects_2011
-
objects_2018
objects_2019

And an insert trigger uses the current date to dertmine where the object is
stored.

Any ideas or suggestions would be appreciated.

Simon
-- 
Simon

Simon Windsor
Eml: simon.wind...@cornfield.org.uk
Tel: 01454 617689
Mob: 07960 321599