Re: help:about mysql cluster partition and parallel query

[Mikael Ronström]

Hi,

fu zhibo Wrote:
-------------------------------------------------------
> hello,Mikael and every one
> After installing Mysql NDB cluster 5.0.27 in
> RHEL3.0, We configurate two node groups(no
> replication among a node group).
>
> I have some questions as follows:
>
> 1.It inserts data to two node group randomly? Does
> it scan all node groups or nodes simultaneously in
> full table scan and order index scan when
> executing a query?
>

Lots of questions in one:
1) Insert data is done by using a hash function
2) Scan all nodes in parallel in full table scan
3) Scan all nodes in parallel in index scan unless
partition pruning can show only one partition needs
to be scanned in which case only one node is scanned.

> 2.How does mysql cluster know to fetch a row from
> which node when using parimary index?record node
> id in parimary index file? and Where is parimary
> index file stored?
> such as a query:select * from s_pri_uni_idx where
> sno=2;
>

It hashes on the primary key fields, this will give
a partition id, this partition id is distributed
in a number of nodes known and it will direct the
search to one of those nodes.

Primary index is partitioned over all nodes in the
cluster.

Rgrds Mikael

> thanks!
> my name:zhibo.fu
>
> You can reply to email:[e-mail]