Concurrent database query benchmark

[Alex Averbuch]

Thanks Jens,
That was the clue I needed.
The example I shared was a slightly simplified version of the actual
benchmark, which populates the database as below.
I was missing the *tx.close()* line.

*    void populateDb( Database db )*
*    {*
*        Transaction tx = db.beginTx();*
*        try*
*        {*
*            for ( int i = 0; i < RECORD_COUNT; i++ )*
*            {*
*                db.createRecord();*
*                if ( i % TX_SIZE == 0 )*
*                {*
*                    tx.success();*
*                    tx.close();             //<---- THIS WAS MISSING*
*                    tx = db.beginTx();*
*                }*
*            }*
*            tx.success();*
*        }*
*        finally*
*        {*
*            tx.close();*
*        }*
*    }*



On Tue, Jun 7, 2016 at 6:17 PM, Jens Wilke <jens.wilke at headissue.com> wrote:

> On Tuesday 07 June 2016 17:24:09 Alex Averbuch wrote:
> > When inspecting the database in *TxState.setUp()* I find that
> > *System.identityHashCode(db)* is always the same, as is the path to the
> > database, so I assume it is the same store -- a good thing.
> > However, *db.getRecordById( id )* then fails due to empty database -- a
> bad
> > thing.
> >
>
> I don't know what database you are using here. Is tx.success() really
> committing your initial population?
>
> Probably you see the content in one thread because the transaction joins
> your still open transaction for population,
> but the transactions in the other threads see nothing, of course.
>
> The general approach for the benchmark looks good!
>
> Cheers,
>
> Jens
>
> --
> "Everything superfluous is wrong!"
>
>    // Jens Wilke - headissue GmbH - Germany
>  \//  https://headissue.com
>