Re: Freezing without write I/O

On Mon, Sep 2, 2013 at 3:16 PM, Jeff Davis <pgsql(at)j-davis(dot)com> wrote:
> On Fri, 2013-08-30 at 20:34 +0200, Andres Freund wrote:
>> I have a quick question: The reason I'd asked about the status of the
>> patch was that I was thinking about the state of the "forensic freezing"
>> patch. After a quick look at your proposal, we still need to freeze in
>> some situations (old & new data on the same page basically), so I'd say
>> it still makes sense to apply the forensic freezing patch, right?
>>
>> Differing Opinions?
>
> The Freeze Forensically patch is nice because (if I understand it
> correctly) it allows us to freeze at the same time as we mark
> PD_ALL_VISIBLE, which avoids the potential extra page write.

The patch itself doesn't actually make that change, but it removes one
major objection to such a change.

> But that's
> not such a big advantage if we don't ordinarily have to write again for
> freezing, anyway.

That was my thought as well.

> However, there are still some cases where we'd be able to preserve the
> forensic information. If nothing else, that might help debug this patch,
> if necessary. There might also be cases where we can freeze more eagerly
> to avoid the case where very old (but unfrozen) and very new tuples mix
> on the same page. Perhaps Robert has some thoughts here, as well.

I basically agree. I think if we adopt Heikki's patch forensic
freezing becomes much less important, but we might find there's still
a reason to do it.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company