Re: [HACKERS] Speedup twophase transactions

On 27 January 2017 at 11:01, Nikhil Sontakke <nikhils@2ndquadrant.com> wrote:
> On 27 January 2017 at 15:37, Simon Riggs <simon@2ndquadrant.com> wrote:
>> On 27 January 2017 at 09:59, Nikhil Sontakke <nikhils@2ndquadrant.com> wrote:
>>>>> But, I put the recovery process and the checkpointer process of the
>>>>> standby under gdb with breakpoints on these functions, but both did
>>>>> not hit CreateRestartPoint() as well as CheckPointGuts() when I issued
>>>>> a promote :-|
>>>>
>>>> No end-of-recovery checkpoints happen at promotion since 9.3. You can
>>>> still use fallback_promote as promote file to trigger the pre-9.2 (9.2
>>>> included) behavior.
>>>
>>> Ok, so that means, we also need to fsync out these 2PC XIDs at promote
>>> time as well for their durability.
>>
>> Why?
>
> The xact_redo code will add prepared transactions to the
> KnownPreparedList in memory. Earlier it used to create the on-disk 2PC
> file.
>
> At standby promote, the surviving (yet uncommitted) prepared
> transactions from KnownPreparedList need to be persisted, right?

I don't see why, so please explain or show the error that can be
caused if we don't.

-- 
Simon Riggs                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services