On 06.09.2020 21:56, Andres Freund wrote:
>
> Hm, that is interesting / odd. If you record a profile with call graphs
> (e.g. --call-graph dwarf), where are all the LWLockAttemptLock calls
> comming from?
>
Attached.
> I assume the machine you're talking about is an 8 socket machine?
>
> What if you:
> a) start postgres and pgbench with numactl --interleave=all
> b) start postgres with numactl --interleave=0,1 --cpunodebind=0,1 --membind=0,1
> in case you have 4 sockets, or 0,1,2,3 in case you have 8 sockets?
>
TPS for -c 100
--interleave=all
1168910
--interleave=0,1
1232557
--interleave=0,1,2,3
1254271
--cpunodebind=0,1,2,3 --membind=0,1,2,3
1237237
--cpunodebind=0,1 --membind=0,1
1420211
--cpunodebind=0 --membind=0
1101203
>> And which pgbench database scale factor you have used?
> 200
>
> Another thing you could try is to run 2-4 pgench instances in different
> databases.
I tried to reinitialize database with scale 200 but there was no
significant improvement in performance.
