Re: Reducing ClogControlLock contention

On 26 August 2015 at 11:40, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:

> On Tue, Aug 25, 2015 at 2:21 PM, Simon Riggs <simon(at)2ndquadrant(dot)com>
> wrote:
>
>> On 22 August 2015 at 15:14, Andres Freund <andres(at)anarazel(dot)de> wrote:
>>
>
>
>> TransactionIdSetPageStatus() calls TransactionIdSetStatusBit(), which
>>> writes an 8 byte variable (the lsn). That's not safe.
>>>
>>
>> Agreed, thanks for spotting that.
>>
>> I see this as the biggest problem to overcome with this patch.
>>
>
> How about using 64bit atomics or spinlock to protect this variable?
>

Spinlock is out IMHO because this happens on every clog lookup. So it must
be an atomic read.

I'm wondering if its worth making this work on 32-bit systems at all. The
contention problems only occur on higher end servers, so we can just
disable this optimization if we aren't on a 64bit server.

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