RE: Memory leak in ODBC driver

Just wanting to let you know that your fixes to the ODBC driver memory
leak have corrected this bug.
We retrieved about 12 million rows over about 17 hours of automated
testing starting with one user and ending up with 100 users and the
memory used in the Win2000 did not increase except when we added a new user.

This is great!

We are now running a TPC-C benchmark using 1 to 100 users. The test looks
good so far. This will take about 17 hours to complete and will test
updating and retrieval. We will let you know how that turns out.

If you ever need to have us run these tests to help debug any problems or
test any performance issues, please let us know and we will be happy to
help out.

Thanks again

Terry Carlin
VP QA and Lab
Great Bridge LLC.

>>>>>>>>>>>>>>>>>> Original Message <<<<<<<<<<<<<<<<<<

On 3/9/01, 10:37:10 PM, "Hiroshi Inoue" <Inoue(at)tpf(dot)co(dot)jp> wrote regarding
RE: [BUGS] Memory leak in ODBC driver:

> > -----Original Message-----
> > From: pgsql-bugs(at)postgresql(dot)org
> >
> > Terry Carlin (terry(at)greatbridge(dot)com) reports a bug with a severity of 2
> > The lower the number the more severe it is.
> >
> > Short Description
> > Memory leak in ODBC driver
> >
> > Long Description
> > As part of Great Bridge QA testing, we run the AS3AP, TPC-C, and TPC-D
> > benchmarks using 1 to 100 users. Our benchmarking tool is Benchmark
> > Factory. It is a WinNT/Win2000 application. It uses ODBC to
> > connect to PostgreSQL as there is no native driver for PostgreSQL
> > yet. Benchmark Factory will run as many transactions as it can
> > during a given timeframe.
> >
> > We have found that there is a memory leak in the PostgreSQL ODBC
> > drivers. The system memory grows 100k to 300k every five seconds
> > during the test. We are hitting from 500 to 1100 transactions per
> > second. After we get through several hundred thousand
> > transactions, we run out of swap space on the Win2000 box. The
> > Win2000 box has about 1.2 gig of swap space.
> >

> I've just committed a change to CVS.
> It would resolve the memory leak a little.

> regards,
> Hiroshi Inoue