On Fri, Feb 24, 2017 at 8:49 AM, Amit Kapila <amit.kapila16@gmail.com> wrote:
>> IIUC, I think that we need to have the number of half-dead pages in meta page.
>
> Don't you think we need to consider backward compatibility if we want
> to do that?
Yeah, that would be an on-disk format break.
I think this thread is pretty short on evidence that would let us make
a smart decision about what to do here. I see three possibilities.
The first is that this patch is a good idea whether we do something
about the issue of half-dead pages or not. The second is that this
patch is a good idea if we do something about the issue of half-dead
pages but a bad idea if we don't. The third is that this patch is a
bad idea whether or not we do anything about the issue of half-dead
pages.
Unfortunately, we have no evidence at all that would let us figure out
which of those three things is true. The original post didn't include
any relevant benchmarks or test results. Simon's reply, which
suggested that the problem of half-dead pages, didn't include any
benchmarks or test results. In fact, in neither place were any tests
suggested, even hypothetically, which would help us decide what to do.
I had a hunch when I saw this thread that it was a good idea, and
Simon has a hunch that this btree page recycling thing needs to be
fixed first, and he might be right. Or we might both be wrong. Or
... who knows, really?
I think we need to come up with some set of tests to figure out what
actually works well in practice here. Theories are a good starting
point, but good vacuum behavior is really important, and a patch that
changes it ought to be backed up by at least some experimental
evidence.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company