Re: about ndb fs size

[Mikael Ronström]

Leo Cardia wrote:
> when i insert 44Mb of sqldump file to table using
> ndb cluster engine
> the ndb fs size is very bigger than sql dump file
>

This is most likely due to that you have lots of big VARCHAR fields not filled very much.
MySQL Cluster has only fixed size rows in 4.1 and 5.0. We have implemented variable
sized record support for 5.1 and it is also integrated with the work on fields on disk.

The size on disk is 3 times the size of data in memory since we checkpoint main memory
entirely in 3 versions before overwriting an old version.
Fields on disk will only be stored in one copy on disk.
In addition the REDO log will use 512 MByte as deafult. There is also an UNDO log that take
up some space if there are lots of updates.

Rgrds Mikael

> # ls -alh cq.sql
> -rw-r--r-- 1 root root 44M 24 Feb 22:00 cq.sql
>
> # du -sh /root/mysqldata/ndb_2_fs/
> 20M /root/mysqldata/ndb_2_fs/
>
> # ./mysql -u root -p nas < cq.sql
> Enter password:
> ERROR 1114 (HY000) at line 58138: The table
> 'zetyx_member_table' is full
>
> # du -sh /root/mysqldata/ndb_2_fs/
> 2.0G /root/mysqldata/ndb_2_fs/
>
> ( one server instance nodes / sql, ndb, mgm node
> at one)
>
> so i checked about (22MB) 58138 record using over
> 500MB's memory
> and using 2GB's filesystem
>
> this is a normal?
>
> we have 4GB of myisam database.
> when putting 4GB db to ndb then we requires 800GB
> Memory, 4TB's hdd disk?
>
>
> NoOfReplicas=1
> DataMemory=500M
> IndexMemory=100M
> NoOfFragmentLogFiles=4
> MaxNoOfConcurrentOperations=5000
> MaxNoOfSavedMessages=12
> MaxNoOfAttributes=100
> MaxNoOfOrderedIndexes=64