Hi,
Boehm, Matthew wrote:
> Morning all,
>
> I ran in to a situation this morning where all my nodes dropped off the
> cluster because they were out of free space on their server. To my
> astonishment, /var/lib/mysql/cluster/ndb_X_fs was using up all 115GB of
> space on each node.
>
>
>
> Surprisingly, I have no TABLESPACE tables and no blob columns or the
> like. Everything is integer, varchar, char or datetime.
>
>
>
> I do, however, have a RedoBuffer of 96MB, FragementLogFileSize of 96MB
> and 800 NoOfFragmentLogFiles for 10GB DM and 2GB IM.
>
>
>
> 800 FragLogFiles * 96MB each = 75GB, still less than 115GB. What else
> can contribute to disk utilization? The actual data and index memory?
> That adds another 12GB to the 75, bringing the total to 87GB. I still
> have 28GB unaccounted for.
>
There is also a factor of 4 to include here, there are 4 parallel log
parts that work in parallel.
Cut from manual:
> The default parameter value is 16, which by default means 16 sets of 4 16MB files
> for
a total of 1024MB. Beginning with MySQL Cluster NDB 6.1.1, the size of
the individual
log files is configurable using the FragmentLogFileSize parameter; more
information
about this parameter can be found here. In scenarios requiring a great
many updates,
the value for NoOfFragmentLogFiles may need to be set as high as 300 or
even higher
to provide sufficient space for REDO logs.
=> 300GB.
Rgrds Mikael
>
>
> Can anyone shed some light? I'm hoping this discussion will help others
> in their planning of rollouts.
>
>
>
> Cheers,
> Matthew
>
>
