Re: Error with index on unlogged table

On Thu, Dec 10, 2015 at 9:07 PM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Thu, Dec 10, 2015 at 8:56 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:
>>> So, do we go for something like the patch you attached in
>>> 20151208125716(dot)GS4934(at)alap3(dot)anarazel(dot)de for master and 9.5, and for
>>> ~9.4 we use the one I wrote in
>>> CAB7nPqSxErpZJ+BZ-mfopzFZP5pAbiE9jWBUcJy6qaYertt4uw(at)mail(dot)gmail(dot)com?
>>
>> I'm more thinking of using something like my patch for all branches. Why
>> would we want to go for the more complicated approach in the more
>> distant branches?
>
> That's not what I think it meant: I don't wish to do the complicated
> approach either anymore. I sent two patches on the mail mentioned
> above: one for master/9.5 that used the approach of changing WAL, and
> a second aimed for 9.4 and old versions that is close to what you
> sent. It looks that you did not look at the second patch, named
> 20151209_replay_unlogged_94.patch that does some stuff with
> XLOG_HEAP_NEWPAGE to fix the issue.
>
>>> Note that in both cases the patches are not complete, we need to fix
>>> as well copy_relation_data(at)tablecmds(dot)c so as the INIT_FORKNUM pages
>>> are logged all the time.
>>
>> Agreed. It's probably better treated as an entirely different - pretty
>> ugly - bug though. I mean it's not some issue of a race during replay,
>> it's entirely missing WAL logging for SET TABLESPACE of unlogged
>> relations.
>
> Okidoki.

In short: should I send patches for all those things or are you on it?
It seems that we are on the same page: using the simple approach, with
XLOG_FPI that enforces the flushes for 9.5/master and
XLOG_HEAP_NEWPAGE that does the same for ~9.4.

For the second issue, I would just need to extract the fix from one of
the patches upthread.
--
Michael