Re: pgsql: Generic Messages for Logical Decoding - Mailing list pgsql-committers

From Fujii Masao
Subject Re: pgsql: Generic Messages for Logical Decoding
Date
Msg-id CAHGQGwEXrJbouHH9XaJH18J2A+0Uq6Ek9Z=BXsiOCK=9RczR+w@mail.gmail.com
Whole thread Raw
In response to Re: pgsql: Generic Messages for Logical Decoding  (Andres Freund <andres@anarazel.de>)
Responses Re: pgsql: Generic Messages for Logical Decoding
List pgsql-committers
On Thu, Apr 7, 2016 at 12:06 AM, Andres Freund <andres@anarazel.de> wrote:
>
>
> On April 6, 2016 5:00:54 PM GMT+02:00, Fujii Masao <masao.fujii@gmail.com> wrote:
>>On Wed, Apr 6, 2016 at 6:08 PM, Simon Riggs <simon@2ndquadrant.com>
>>wrote:
>>> Generic Messages for Logical Decoding
>>>
>>> API and mechanism to allow generic messages to be inserted into WAL
>>that are
>>> intended to be read by logical decoding plugins. This commit adds an
>>optional
>>> new callback to the logical decoding API.
>>
>>When I specify very long text message, pg_xlogdump failed to dump
>>correctly the WAL file which should contain that inserted text message.
>>Isn't this a bug?
>>
>>You can reproduce the problem by the following steps.
>>
>>=# SELECT pg_xlogfile_name(pg_switch_xlog());
>>     pg_xlogfile_name
>>--------------------------
>> 000000010000000000000003
>>(1 row)
>>
>>=# SELECT pg_xlogfile_name(pg_current_xlog_location());
>>     pg_xlogfile_name
>>--------------------------
>> 000000010000000000000003
>>(1 row)
>>
>>=# SELECT pg_logical_emit_message(true, 'test',
>>repeat('0123456789ABCDEFG', 1024*1024));
>> pg_logical_emit_message
>>-------------------------
>> 0/510CD40
>>(1 row)
>>
>>=# SELECT pg_xlogfile_name(pg_current_xlog_location());
>>     pg_xlogfile_name
>>--------------------------
>> 000000010000000000000005
>>(1 row)
>>
>>=# insert into t values(0,0);
>>INSERT 0 1
>>postgres=# SELECT pg_xlogfile_name(pg_current_xlog_location());
>>     pg_xlogfile_name
>>--------------------------
>> 000000010000000000000005
>>(1 row)
>>
>>The WAL record of pg_logical_emit_message() should be stored in
>>000000010000000000000004 and 000000010000000000000005.
>>The WAL record of last insertion should be stored in
>>000000010000000000000005.
>>But the results of pg_xlogdump were wrong as follows.
>>
>>$ pg_xlogdump data/pg_xlog/000000010000000000000004
>>rmgr: Standby     len (rec/tot):     24/    50, tx:          0, lsn:
>>0/04000028, prev 0/030146F8, desc: RUNNING_XACTS nextXid 863
>>latestCompletedXid 862 oldestRunningXid 863
>>
>>$ pg_xlogdump data/pg_xlog/000000010000000000000005
>>pg_xlogdump: FATAL:  could not find a valid record after 0/5000000
>
> If you specify a file it only looks at records in that file. Try with -s.

In my example, the WAL record of INSERT that I executed last should be in
000000010000000000000005. But pg_xlogdump could not display that.
The output of pg_xlogdump was:

$ pg_xlogdump data/pg_xlog/000000010000000000000005
pg_xlogdump: FATAL:  could not find a valid record after 0/5000000

ISTM that if a WAL file starts with the latter half of LOGICAL MESSAGE
WAL data, pg_xlogdump treats it as invalid and gives up dumping the
remaining WAL data in the file.

Regards,

--
Fujii Masao


pgsql-committers by date:

Previous
From: Fujii Masao
Date:
Subject: Re: pgsql: Use GRANT system to manage access to sensitive functions
Next
From: Noah Misch
Date:
Subject: pgsql: Standardize GetTokenInformation() error reporting.