cqlsh strange inconsistency

[Onur Yalazı <on...@8digits.com>]

Hello,

I have a table with a status field and a secondary index on that table. 
Every row has 1 or -1 status values.

I have these queries run from cqlsh with the results:

consistency all;
select count(*) from table; /** 1901
Select count(*) from table where status=1; /** 27
select count(*) from table where status=-1; /** 267

This is really strange. If I select status from the table, I see 1901 
rows with status 1 or -1.
I rebuild the index and the results did not change. But If I update 
every row with the correct status value, cassandra starts reporting 
correct counts.

What is happening? How can this situation happen? Do you have any ideas?

Cassandra version: 2.1.3
cqlsh version: 4.1.1

Re: cqlsh strange inconsistency

[Robert Coli <rc...@eventbrite.com>]

On Thu, Oct 1, 2015 at 5:37 AM, Onur Yalazı <on...@8digits.com> wrote:

> What is happening? How can this situation happen? Do you have any ideas?

 ...

>  This is really strange. If I select status from the table, I see 1901
> rows with status 1 or -1.

I rebuild the index and the results did not change. But If I update every
> row with the correct status value, cassandra starts reporting correct
> counts.


There is an endless trickle of such issues with secondary indexes. This is
one of the reasons I do not personally recommend their use.

If you have a repro test case, file it at issues.apache.org and let the
list know the issue URL?

=Rob