Thread: "set primary keys..." is missing when using hight values fortransactions / scaling factor with pgbench

I have strange issue with pgbench where it fails to execute step to create primary keys when I specify scaling factor / transactions to some reasonable high value - eg. 8k.

I am doing

$ pgbench -i -s 8000 sampledb

$ pgbench -c 10 -j 2 -t 8000 sampledb

when I fails I see

=====
vacuum...
starting vacuum...end.
====

and good case

vacuum...
set primary keys...
done.
starting vacuum...end.

Only see this with a bit reasonable values for transactions / scaling factor ( eg. 4k,8k )

I suspected that there could be issue with memory/disk allocated to machine, but for above case it will generate cca 150 GB of data and I have disk with 350 GB  of free space, and allocating 30GB of memory for postgresql machine ( tried with even more resources )
I am using pgbench (PostgreSQL) 9.6.5 .
If you have any idea, please share , thank you
Elvir



On 27 June 2018 at 00:21, Elvir Kurić <elvirkuric@gmail.com> wrote:
> I have strange issue with pgbench where it fails to execute step to create
> primary keys when I specify scaling factor / transactions to some reasonable
> high value - eg. 8k.

The primary keys are only created in -i mode, which can't be used in
conjunction with the options you've mentioned.

pgbench will perform a vacuum before an actual test run, so perhaps
that's what you're seeing.  You may also have noticed it also didn't
perform the create tables and data population too without -i.


--
 David Rowley                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services


thank you David.

I first run initialize step

$  pgbench -i -s 8000 sampledb

and then run step

$ pgbench -c 10 -j 2 -t 8000 sampledb

if I change -s/-t to lower value , eg, 100 above commands will show

--
set primary keys...
done.
-- -
I am not getting it ,why it fails when I rise -t/-s to 8000 - with same commands.

Do you suggest that above is not correct way?




On Tue, Jun 26, 2018 at 2:41 PM, David Rowley <david.rowley@2ndquadrant.com> wrote:
On 27 June 2018 at 00:21, Elvir Kurić <elvirkuric@gmail.com> wrote:
> I have strange issue with pgbench where it fails to execute step to create
> primary keys when I specify scaling factor / transactions to some reasonable
> high value - eg. 8k.

The primary keys are only created in -i mode, which can't be used in
conjunction with the options you've mentioned.

pgbench will perform a vacuum before an actual test run, so perhaps
that's what you're seeing.  You may also have noticed it also didn't
perform the create tables and data population too without -i.


--
 David Rowley                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services