Re: [HACKERS] Restrict concurrent update/delete with UPDATE of partition key

On Sat, Apr 7, 2018 at 9:08 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> Hi Tom, All,
>
> On 2018-04-06 14:19:02 +0530, amul sul wrote:
>> Thanks for the reminder -- fixed in the attached version.
>
> Tom, this seems to be the best approach for fixing the visibility issues
> around this. I've spent a good chunk of time looking at corruption
> issues like the ones you feared (see [1]) and I'm not particularly
> concerned. I'm currently planning to go ahead with this, do you want to
> "veto" that (informally, not formally)?
>
> I'll go through this again tomorrow morning.
>
> [1] https://postgr.es/m/20180405014439.fbezvbjrmcw64vjc@alap3.anarazel.de
>
>
>> v9:
>> Its the rebase version of Andres Freund patch v8[1] with the
>> following additional changes:
>> 3. Argument changing_part of heap_delete renamed to ChangingPart to be
>> consistent with ExecDelete
>
> FWIW, I'd left it as it was before because the two functions have a bit
> different coding style, and the capitalization seemed more fitting in
> the surrounding context.
>
>> +test: partition-key-update-1
>> +test: partition-key-update-2
>> +test: partition-key-update-3
>
> Can you give these more descriptive names please (or further combine them)?
>

As I explained above further combining might not the good option and about
the descriptive names I have following suggestions but I am afraid of
the length of
test case name:

+test: concurrent-partition-key-update.out

This test does the serialization failure check.

+test: concurrent-partition-key-update-and-insert-conflict-do-nothing-1
+test: concurrent-partition-key-update-and-insert-conflict-do-nothing-2

Both are testing partition key update behaviour with the insert on
conflict do nothing case.

Attached is the patch does the renaming of this tests -- need to apply
to the top of v10 patch[1].

Regards,
Amul

1] https://postgr.es/m/CAAJ_b94X5Y_zdTN=BGdZie+hM4p6qW70-XCJhFYaCUO0OfF=aQ@mail.gmail.com