On 2020-02-27 12:26:36 -0800, Andres Freund wrote:
> Hi,
>
> On 2015-09-25 20:35:45 +0200, Fabien COELHO wrote:
> >
> > Hello Tatsuo,
> >
> > > Hmmm... I never use -C. The formula seems ok:
> > >
> > > tps_exclude = normal_xacts / (time_include -
> > > (INSTR_TIME_GET_DOUBLE(conn_total_time) /
> > > nthreads));
> >
> > Hmmm... it is not:-)
> >
> > I think that the degree of parallelism to consider is nclients, not
> > nthreads: while connection time is accumulated in conn_time, other clients
> > are possibly doing their transactions, in parallel, even if it is in the
> > same thread, so it is not "stopped time" for all clients. It starts to
> > matter with "-j 1 -c 30" and slow transactions, the cumulated conn_time in
> > each thread may be arbitrary close to the whole time if there are many
> > clients.
>
> I think this pretty much entirely broke the tps_exclude logic when not
> using -C, especially when -c and -j differ. The wait time there is
> actually per thread, not per client.
>
> In this example I set post_auth_delay=1s on the server. Pgbench
> tells me:
> pgbench -M prepared -c 180 -j 180 -T 10 -P1 -S
> tps = 897607.544862 (including connections establishing)
> tps = 1004793.708611 (excluding connections establishing)
>
> pgbench -M prepared -c 180 -j 60 -T 10 -P1 -S
> tps = 739502.979613 (including connections establishing)
> tps = 822639.038779 (excluding connections establishing)
>
> pgbench -M prepared -c 180 -j 30 -T 10 -P1 -S
> tps = 376468.177081 (including connections establishing)
> tps = 418554.527585 (excluding connections establishing)
>
> which pretty obviously is bogus. While I'd not expect it'd to work
> perfectly, the "excluding" number should stay roughly constant.
>
>
> The fundamental issue is that without -C *none* of the connections in
> each thread gets to actually execute work before all of them have
> established a connection. So dividing conn_total_time by / nclients is
> wrong.
>
> For more realistic connection delays this leads to the 'excluding'
> number being way too close to the 'including' number, even if a
> substantial portion of the time is spent on it.
Not suggesting it as an actual fix, but just multiplying the computed
conn_time within runThread() by the number of connections the thread
handles (for the non -C case) results in much more reasonable output.
pgbench -M prepared -c 180 -j 30 -T 10 -P1 -S
before:
tps = 378393.985650 (including connections establishing)
tps = 420691.025930 (excluding connections establishing)
after:
tps = 379818.929680 (including connections establishing)
tps = 957405.785600 (excluding connections establishing)
pgbench -M prepared -c 180 -j 180 -T 10 -P1 -S
before:
tps = 906662.031099 (including connections establishing)
tps = 1012223.500880 (excluding connections establishing)
after:
tps = 906178.154876 (including connections establishing)
tps = 1012431.154463 (excluding connections establishing)
The numbers are still a bit bogus because of thread startup overhead
being included, and conn_time being computed relative to the thread
creation. But they at least make some basic sense.
diff --git i/src/bin/pgbench/pgbench.c w/src/bin/pgbench/pgbench.c
index 1159757acb0..3bc45107136 100644
--- i/src/bin/pgbench/pgbench.c
+++ w/src/bin/pgbench/pgbench.c
@@ -6265,6 +6265,16 @@ threadRun(void *arg)
INSTR_TIME_SET_CURRENT(thread->conn_time);
INSTR_TIME_SUBTRACT(thread->conn_time, thread->start_time);
+ /* add once for each other connection */
+ if (!is_connect)
+ {
+ instr_time e = thread->conn_time;
+ for (i = 0; i < (nstate - 1); i++)
+ {
+ INSTR_TIME_ADD(thread->conn_time, e);
+ }
+ }
+
/* explicitly initialize the state machines */
for (i = 0; i < nstate; i++)
{
Greetings,
Andres Freund
