[pgpool-general: 1856] Re: weird error with postgres_fdw & pgpool

[Tatsuo Ishii]

No. Because I'm not familiar with postgres_fdw, I just wanted to make
sure what the cause of the problem is. I cannot promise I can solve
your problem, but I'm sure without knowing the cause, it's absolutely
impossible to solve the problem.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp

> I'm not sure that I understand the question. Are you asking me to set
> client_idle_limit to 0?
> 
> On Wed, Jun 19, 2013 at 6:15 PM, Tatsuo Ishii <ishii at postgresql.org> wrote:
>> Lonni,
>>
>> Is it possible for postgres_fdw to create a persistent connection from
>> db11 to db12 for certain pupose(maybe connection pooling)? If so, the
>> connection will be treated as a normal conection from client by pgpool
>> and client_idle_limit will be applied.
>> --
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> Japanese: http://www.sraoss.co.jp
>>
>>> I'm in the process of testing out a beta of postgres-9.3 using
>>> pgpool-3.2.4 and the (new) postgres_fdw.   I have two different
>>> postgres-9.3 servers setup (db10 & db11), and pgpool is setup (db12)
>>> pointing to one of them (db10).  I can connect with psql to both
>>> servers directly, and also via pgpool, with no problems.  My intent
>>> was to be able to run SQL queries while logged into db11 through
>>> pgpool on db12 to the server on db10, using a postgres_fdw.
>>>
>>> I setup the postgres_fdw server:
>>> CREATE SERVER db10 FOREIGN DATA WRAPPER postgres_fdw OPTIONS (host
>>> 'db12', dbname 'nightly', port '9999');
>>> CREATE USER MAPPING FOR lfriedman SERVER db10 OPTIONS (user 'lfriedman');
>>> CREATE FOREIGN TABLE nppsmoke (id bigint NOT NULL, date_created
>>> timestamp without time zone not null default now(), last_update
>>> timestamp without time zone not null default now(), build_type text
>>> NOT NULL,current_status text NOT NULL, info text NOT NULL, cudacode
>>> text NOT NULL, gpu text NOT NULL, subtest text NOT NULL, os text NOT
>>> NULL, osversion text NOT NULL, arch text NOT NULL, cl integer NOT
>>> NULL, dispvers text NOT NULL, branch text NOT NULL, pass integer,fail
>>> integer, oldfail integer, newfail integer, failureslog text, totdriver
>>> boolean NOT NULL default 'true',ddcl integer NOT NULL, buildid
>>> integer, testdcmd text NOT NULL, pclog text, filtercount integer,
>>> filterlog text, error integer) SERVER db10 ;
>>>
>>> After which, I verified that I could issue SQL queries from inside
>>> psql while connected to db11, against the foreign table, nppsmoke that
>>> was created above:
>>> #########
>>> nightly=# select last_update from nppsmoke where id='86901' ;
>>>      last_update
>>> ---------------------
>>>  2013-06-12 12:53:23
>>> (1 row)
>>> #########
>>>
>>> However, I'm seeing some strange errors when my psql session exceeds
>>> the pgpool client_idle_limit:
>>> #######
>>> nightly=# select last_update from nppsmoke where id='86901' ;
>>> ERROR:  connection terminated due to client idle limit reached
>>> CONTEXT:  Remote SQL command: START TRANSACTION ISOLATION LEVEL REPEATABLE READ
>>> nightly=# select last_update from nppsmoke where id='86901' ;
>>> ERROR:  unknown error
>>> CONTEXT:  Remote SQL command: START TRANSACTION ISOLATION LEVEL REPEATABLE READ
>>> ########
>>>
>>> And the only way that I've found to run the SQL queries against the
>>> foreign table is to exit psql and connect again from scratch.  This is
>>> definitely different behavior than I experience for local tables.  For
>>> queries issued after the client_idle_limit, I always see:
>>> #####
>>> ERROR:  connection terminated due to client idle limit reached
>>> ERROR:  connection terminated due to client idle limit reached
>>> The connection to the server was lost. Attempting reset: Succeeded.
>>> #####
>>>
>>> after which I can re-run the query without exiting & reconnecting with psql.
>>>
>>> This seems like a bug to me, as I'd expect the behavior to be
>>> consistent regardless of what is on the remote end.  Thoughts?