Thank for you replies.
1) I can not create raw each X time , since it will not allow me to get a
complete list of currently active records ( this is the only reason i keep
this raw initially ).
2) As for compaction i thought that only raw ids are cached and not columns
itself. I have completed compaction and it indeed cleared most of data (
75% of complete table ,and there are some other raws ).
It looks like it did not cleared all the data from disk ( compaction should
set table into single file , while other files still left ) , but showed in
log that data cleared , also after compaction this raw became responsive.

Most of our data is wrotten once and kept in history , however we have also
counter columns which are not working good ( had troubles with that ) and
several places where we use create / delete approach.Now i understand why
our data grows so much , it just does not clears old data at all , but
marks it as tombstone...
What NOSQL database would you recommend for such usages ( write onces read
many mixed with counter columns mixed with read/write oftenly data)?

Thanks and best regards
Yulian Oifa




On Mon, Apr 7, 2014 at 5:57 PM, Lukas Steiblys <lu...@doubledutch.me> wrote:

>   Deleting a column simply produces a tombstone for that column, as far
> as I know. It's probably going through all the columns with tombstones and
> timing out. Compacting more often should help, but maybe Cassandra isn't
> the best choice overall for what you're trying to do.
>
> Lukas
>
>  *From:* Yulian Oifa <oifa.yul...@gmail.com>
> *Sent:* Sunday, April 6, 2014 11:54 AM
> *To:* user@cassandra.apache.org
> *Subject:* Transaction Timeout on get_count
>
>    Hello
> I am having raw in which approximately 100 values is written per minute.
> Those columns are then deleted ( it contains active records list ).
> When i am trying to execute get_count on that raw i get transaction
> timeout , even while the raw is empty.
> I dont see anything in cassandra log on neither node , pending tasks are
> zero.
> What could be a reason for that , and how can it be resolved?
> Best regards
> Yulian Oifa
>

Reply via email to