Konstantin,

I've found that dropping a column family creates a snapshot quietly.
So I gave it a "./nodetool clearsnapshot" command and the stale files were
deleted.

On Tue, Jul 19, 2011 at 6:26 PM, Konstantin Naryshkin
<konstant...@a-bb.net>wrote:

> "2. Trying to reduce disk occupation I deleted CF which used 90% of
> available
> space. After issuing a "drop column family User;" command
>  no *User*.db files were deleted. "nodetool compact" haven't helped too.
> How
> can that deletion be triggered?"
>
> You have to wait for a garbage collect (or do a rolling restart). Those are
> the only times that Cassandra deletes files. There is a better way being
> developed (https://issues.apache.org/jira/browse/CASSANDRA-2521).
>
> Konstantin
>
> ----- Original Message -----
> From: "Nikolai Kopylov" <kopy...@gmail.com>
> To: user@cassandra.apache.org
> Sent: Tuesday, July 19, 2011 1:56:15 PM
> Subject: deletion questions
>
> Hello, I'm trying to use Cassandra 0.8.1 as a media storage where
> the stored items have some TTL.
> So that I need to somehow delete and purge obsoleted items. I've played
> with
> deletion columns
> from db by using Thift  CassandraClient::remove call, but faced with the
> problem that sstable files
> become too big and I had to give Cassandra up to 100% more disk space only
> for the compation process
> to succeed. This is a point because I'm calculating the total disks
> capacity
> and the data stored in column is not less than 8GB.
> So for now I should build 16TB volumes, not less.
> According to the above I can outline two questions.
> 1. Is it possible to reorganize compaction process so that the compaction
> result would be not one but several files with the size upper-limited
> with some parameter with default value of, say, 4GB?
> 2. Trying to reduce disk occupation I deleted CF which used 90% of
> available
> space. After issuing a "drop column family User;" command
>  no *User*.db files were deleted. "nodetool compact" haven't helped too.
> How
> can that deletion be triggered?
>
> --
> WBR, Nikolay
>

Reply via email to