cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Dejanovski <a...@thelastpickle.com>
Subject Re: Why compacting process uses more data that is expected
Date Thu, 05 Jan 2017 07:53:37 GMT
Indeed, nodetool compactionstats shows uncompressed sizes.
As Oleksandr suggests, use the table compression ratio to compute the
actual size on disk.

It would actually be a great improvement for ops if we could add a switch
to compactionstats in order to have the compression ratio applied
automatically.

On Thu, Jan 5, 2017 at 7:22 AM Oleksandr Shulgin <
oleksandr.shulgin@zalando.de> wrote:

> On Jan 4, 2017 17:58, "Jean Carlo" <jean.jeancarl48@gmail.com> wrote:
>
> Hello guys
>
> I have a table with 34Gb of data in sstables (including tmp). And I can
> see cassandra is doing some compactions on it. What surprissed me is that
> nodetool compactionstats says he is compacting  138.66GB
>
>
> root@node001 /root # nodetool compactionstats -H
> pending tasks: 103
> *   compaction type        keyspace              table
> completed       total         unit   progress*
>         Compaction             keyspace1           table_02   112.74 GB
> 138.66 GB   bytes     81.31%
> Active compaction remaining time :   0h03m27s
>
> So My question is, from where those 138.66GB come if my table has only
> 34GB of data.
>
>
> Hello,
>
> I believe that output of compactionstats shows you the size of
> *uncompressed* data. Can you check (with nodetool tablestats) your
> compression ratio?
>
> --
> Alex
>
> --
-----------------
Alexander Dejanovski
France
@alexanderdeja

Consultant
Apache Cassandra Consulting
http://www.thelastpickle.com

Mime
View raw message