incubator-cassandra-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dragos cernahoschi <dragos.cernahos...@gmail.com>
Subject Re: CASSANDRA-1472 (bitmap indexes)
Date Wed, 24 Nov 2010 10:43:35 GMT
I've tried to reproduce my test data and the failing queries with stress.py.

So, I've slightly modified the stress.py and added 2 more indexes for
insertion. The indexrangeslice query is also performed on 3 indexes. The
insert is done using an uniform distribution of values.

Then:

1. python contrib/py_stress/stress.py -r -C 32 -x keys
2. python contrib/py_stress/stress.py -C 32 -o indexedrangeslice -t 3

The queries fails as in the attachment: not on the first query but on the
3rd, 4th ... not allways the same.

Dragos

On Mon, Nov 22, 2010 at 9:39 PM, Jonathan Ellis <jbellis@gmail.com> wrote:

> Let's start wth the low-hanging fruit: can you give steps to reproduce
> queries that fail right away?
>
> On Wed, Nov 17, 2010 at 10:37 AM, dragos cernahoschi
> <dragos.cernahoschi@gmail.com> wrote:
> > Back. I've tested the keys index pagination once again. 0.7 head. Smaller
> > data set: 1 million rows. It seems there are still some issues:
> >
> > 1. *test*: query on one column, count: 1000, expected number of distinct
> > results: 48251
> >    *result*: 5 pages of 1000 results, than, after the 6th page, the
> results
> > begin to repeat, I would expect that repetition begins after the 48251-th
> > row
> >
> > 2. *test*: query on 3 columns, count: 10 (count 100, count 1000 failed
> with
> > time out)
> >    *result*: 1 page of 10 results, than second page => time out
> >
> > 3. There are queries with combinations of 2, 3 columns that fail right
> away
> > with time out (count 10, 100).
> >
> > Dragos
> >
> >
> > On Mon, Nov 15, 2010 at 2:29 PM, Jonathan Ellis <jbellis@gmail.com>
> wrote:
> >
> >> On Mon, Nov 15, 2010 at 5:57 AM, dragos cernahoschi
> >> <dragos.cernahoschi@gmail.com> wrote:
> >> > I've tested 0.7-beta3 branch index feature without the 1472 patch. The
> >> > queries on more than one column works better than the patched version,
> >> but
> >> > definitely not correctly.
> >>
> >> Please test 0.7 branch head, as you can see from the CHANGES there
> >> have been a lot of fixes.
> >>
> >> > 1.
> >> > 2.
> >> > 4.
> >>
> >> Should be fixed in head.
> >>
> >> > 3. Is there any example on the pagination feature? (without knowing
> the
> >> > expected number of rows).
> >>
> >> Same way you paginate through range slices or columns within a row,
> >> set start to the last result you got w/ previous query.
> >>
> >> > Will the get_indexed_slices return an empty list when there is no more
> >> > results?
> >>
> >> No, all queries are start-inclusive.
> >>
> >> --
> >> Jonathan Ellis
> >> Project Chair, Apache Cassandra
> >> co-founder of Riptano, the source for professional Cassandra support
> >> http://riptano.com
> >>
> >
>
>
>
> --
> Jonathan Ellis
> Project Chair, Apache Cassandra
> co-founder of Riptano, the source for professional Cassandra support
> http://riptano.com
>

Mime
View raw message