Re: BUG #6041: Unlogged table was created bad in slave node

On Wed, Jun 1, 2011 at 7:28 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Thu, May 26, 2011 at 12:34 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Alvaro Herrera <alvherre(at)commandprompt(dot)com> writes:
>>> Excerpts from Euler Taveira de Oliveira's message of jue may 26 12:00:05 -0400 2011:
>>>> I think we should emit the real cause in those cases, if possible (not too
>>>> much overhead). The message would be "Unlogged table content is not available
>>>> in standby server".
>>
>>> I guess what it should do is create an empty file in the slave.
>>
>> Probably it should, because won't the table malfunction after the slave
>> is promoted to master, if there's no file at all there?  Or will the
>> process of coming live create an empty file even if there was none?
>
> Coming live creates an empty file.
>
>> But Euler is pointing out a different issue, which is usability.  If the
>> slave just acts like the table is present but empty, we are likely to
>> get bug reports about that too.  An error telling you you aren't allowed
>> to access such a table on slaves would be more user-friendly, if we can
>> do it without too much pain.
>
> I looked into this a bit.  A few observations:
>
> (1) This problem is actually not confined to unlogged tables;
> temporary tables have the same issue.  For example, if you create a
> temporary table on the master and then, on the slave, do SELECT * FROM
>  pg_temp_3.hi_mom (or whatever the name of the temp schema where the
> temp table is) you get the same error.  In fact I suspect if you took
> a base backup that included the temporary relation and matched the
> backend ID you could even manage to read out the old contents (modulo
> any fun and exciting XID wraparound issues).  But the problem is of
> course more noticeable for unlogged tables since they're not hidden
> away in a special funny schema.

Seems like you're trying to fix the problem directly, which as you
say, has problems.

At some point we resolve from a word mentioned in the FROM clause to a
relfilenode.

Surely somewhere there we can notice its unlogged before we end up
down in the guts of smgr?

--
 Simon Riggs                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services