Re: Concurrency bug in UPDATE of partition-key

On Mon, Jun 11, 2018 at 3:02 PM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> On Thu, Jun 7, 2018 at 1:53 PM, Amit Khandekar <amitdkhan(dot)pg(at)gmail(dot)com> wrote:
>> On 7 June 2018 at 11:44, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>>> On Tue, Jun 5, 2018 at 8:03 PM, Amit Khandekar <amitdkhan(dot)pg(at)gmail(dot)com>
>>> wrote:
>>>
>>> I think this will allow before row delete triggers to be fired more than
>>> once. Normally, if the EvalPlanQual testing generates a new tuple, we don't
>>> fire the triggers again.
>>
>> If there are BR delete triggers, the tuple will be locked using
>> GetTupleForTrigger(). So the subsequent EvalPlanQual testing won't be
>> run, since the tuple is already locked due to triggers having run.
>>
>> But that leads me to think : The same concurrency issue can occur in
>> GetTupleForTrigger() also. Say, a concurrent session has already
>> locked the tuple, and GetTupleForTrigger() would wait and then return
>> the updated tuple in its last parameter newSlot. In that case, we need
>> to pass this slot back through ExecBRDeleteTriggers(), and further
>> through epqslot parameter of ExecDelete(). But yes, in this case, we
>> should avoid calling this trigger function the second time.
>>
>> If you agree on the above, I will send an updated patch.
>>
>
> Sounds reasonable to me.
>

Try to add a test case which covers BR trigger code path where you are
planning to update.

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