Re: BUG #14530: Logical Decoding Slowness

On Fri, Feb 10, 2017 at 6:20 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:

> On 2017-02-10 06:33:30 +0000, Huan Ruan wrote:
> > >
> > >
> > > Could you give a bit more detail abou thtis one - it indeed looks like
> > > different profile, but it's hard to nail down without more lines...
> > >
> > >
> > First screen of unexpanded output:
> >
> > …
> >
> > + 94.84% 0.00% postgres postgres
> > [.] ExecMakeTableFunctionResult
> > + 94.84% 0.00% postgres postgres
> > [.] pg_logical_slot_get_changes_guts
> > + 94.80% 0.00% postgres postgres
> > [.] LogicalDecodingProcessRecord
> > + 94.45% 0.03% postgres postgres
> > [.] ReorderBufferCommit
> > + 94.19% 0.01% postgres postgres
> > [.] ReorderBufferCleanupTXN
> > + 94.11% 0.01% postgres postgres
> > [.] ReorderBufferReturnChange
> > + 94.08% 93.40% postgres postgres
> > [.] AllocSetFree
> >
> > + 2.33% 0.00% perf [kernel.kallsyms]
> > [k] system_call_fastpath
>
> Ok, so this is likely a known issue. If you're in the position to try,
> there's a set of patches to address this:
> http://archives.postgresql.org/message-id/20ca0c69-e8c3-
> cb50-4456-f03908c672ec%402ndquadrant.com
>
> Andres
>
>
Hi Andres,

I've been working with Huan on this. I tried those patches, and can confirm
that they resolve the subtransaction issue.

Thanks,
Nick