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

On Tue, Feb 13, 2018 at 12:41 PM, amul sul <sulamul(at)gmail(dot)com> wrote:
> On Tue, Feb 13, 2018 at 11:32 AM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>>
>> Your change appears fine to me. I think one can set both block number
>> and offset as we do for HeapTupleHeaderIsSpeculative, but the way you
>> have done it looks good to me. Kindly include it in the next version
>> of your patch by adding the missing comment.
>>
>
> Thanks for the confirmation, updated patch attached.
>

+# Concurrency error from GetTupleForTrigger
+# Concurrency error from ExecLockRows

I think you don't need to mention above sentences in spec files.
Apart from that, your patch looks good to me. I have marked it as
Ready For Committer.

Notes for Committer -
1. We might need some changes in update-tuple-routing mechanism if we
decide to do anything for the bug [1] discussed in the nearby thread,
but as that is not directly related to this patch, we can move ahead.
2. I think it is better to document that for update tuple routing the
delete+insert will be replayed separately on replicas. I leave this
to the discretion of the committer.

[1] - https://www.postgresql.org/message-id/CAAJ_b94bYxLsX0erZXVH-anQPbWqcYUPWX4xVRa1YJY%3DPh60ZQ%40mail.gmail.com

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com