On Sun, Feb 22, 2009 at 11:55 PM, Jordan Tomkinson <jordan@moodle.com> wrote:
> Hi list,
>
> We are running postgresql 8.3.5 and are trying to stress test our LMS.
> The problem is when our stress tester (Jmeter) inserts around 10,000 rows
> (in 3 hours) over 2 tables (5000 rows each table) the CPU of the sql server
> hits 100% over all 4 cores for all future inserts.
>
> I have tried numerous things to get the cpu back down but so far the only
> thing that works is deleting the 10,000 rows Jmeter inserted.
>
> For more information on the problem along with a time stamped list of test
> results and outcomes please see
> http://spreadsheets.google.com/pub?key=pu_k0R6vNvOVP26TRZdtdYw
Can you post the jmeter files? OR create a SQL test case? I haven't
had this problem myself, so I'm guessing something in your method or
something in your schema is setting something strange off. OR the
background writer is busy writing all the changes out after the fact
while the database is breathing from the heavy run. 10,000 rows over
three hours isn't really a whole lotta work unless those are really
wide rows.
Oh, what is an LMS?