Re: In-placre persistance change of a relation

At Fri, 08 Jan 2021 17:52:21 +0900 (JST), Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com> wrote in
> At Fri, 08 Jan 2021 14:47:05 +0900 (JST), Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com> wrote in
> > This version RelationChangePersistence() is changed not to choose
> > in-place method for indexes other than btree. It seems to be usable
> > with all kind of indexes other than Gist, but at the mement it applies
> > only to btrees.
> >
> > 1: https://www.postgresql.org/message-id/CA+TgmoZEZ5RONS49C7mEpjhjndqMQtVrz_LCQUkpRWdmRevDnQ@mail.gmail.com
>
> Hmm. This is not wroking correctly. I'll repost after fixint that.

I think I fixed the misbehavior. ResetUnloggedRelationsInDbspaceDir()
handles file operations in the wrong order and with the wrong logic.
It also needed to drop buffers and forget fsync requests.

I thought that the two cases that this patch is expected to fix
(orphan relation files and uncommited init files) can share the same
"cleanup" fork but that is wrong. I had to add one more additional
fork to differentiate the cases of SET UNLOGGED and of creation of
UNLOGGED tables...

The attached is a new version, that seems working correctly but looks
somewhat messy. I'll continue working.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center