Hello again,
Yes, new databases are created rarely, but this is not solution for my
issue. My users can create databases and our system is clustered with
LinuxDirector (ipvs) from Ultramonkey project (http://www.ultramonkey.org/),
so we have only one virtual sql serwer. The NDBCluster can be used to make
n-way replication between n APIs. If my user is making connection to
database server he would never know which API is he connecting, because this
connetion will be switched by LinuxDirector using specific switching
algorithm. After his creation of database he will make another connection to
use that database, and LinuxDirector will switch him to another API. In
effect he wont be able to use his newly created database. But if NDBCluster
would propagate info about creation of database through all APIs, everything
would be just fine.
Best Regards,
Pawel Panek
> New databases are typically created very rarely.
>
> You only need to execute "create database ..." on each of the cluster
> servers. The tables are in the cluster as long as you have created
> them on any api node and will be available to selects from the other
> nodes as soon as you create the database on the other database
> servers.
>
> You can verify your tables are in your cluster by using ndb_show_tables.
>
> The tables will not show up in "show tables" until you select from
> them. This is also a known/open bug that will eventually be fixed.
> The only practical application of this shortcoming is from a
> development perspective. Applications selecting from the tables will
> work just fine. Only "show tables" has an issue because it doesn't
> query the cluster for a metadata listing for the current database
> (yet).
>
>On Thu, 12 Aug 2004 14:34:20 +0200, Paweł Panek <pawel.panek@stripped>
wrote:
> > Hi all!
> >
> > I meet problem described in post from 11 Jul 2004
> > (http://lists.mysql.com/cluster/108). Now is 12 Aug and this problem is
> > still there. I have onle polite question: when can we expect full
> > replication in MySQL Cluster?
> >
> > It is very important to heve replicated newly created database and its
> > tables through all APIs in the Cluster. Making it by hand is sensless,
> > because IMO automated and full replication (or distribution) of data
should
> > be the main idea of clustered system.
> >
> > If I'm wrong please feel free to kick my butt off.
> >
> > Best Regards,
> > Pawel Panek
> >
> > PS
> > Sorry for my english.
> > --
> > Image Electronics http://www.inet.pl/
> > Plac Rodla 8, 70-419 Szczecin, Poland
> > e-mail: Pawel.Panek@stripped
> > tel: +4891 4324200 fax: +4891 4882175
> >
> > --
> > MySQL Cluster Mailing List
> > For list archives: http://lists.mysql.com/cluster
> > To unsubscribe:
http://lists.mysql.com/cluster?unsub=1
> >
> >
