| Author |
Message |
![[Post New]](/templates/default/images/icon_minipost_new.gif) 4 May 2010 14:35:50
|
Wilfried_Hartman
New member
Joined: 4 May 2010 14:21:58
Messages: 1
Offline
|
Hello
I'd like to use GridSQL to execute distributed queries on several PostgreSQL servers.
I ran some test with GridSQL but I still have some questions. My coordinator is currently managing two nodes. These two nodes have been "initialized" (DB schema creation) through the coordinator with some "create table" instructions. In a few months, when I'll decide to add a new node, how will I initialize it ? I can't use 'create table' instructions in the coordinator because the 2 first nodes will already have these tables... So, is there another way to initialize the third node than "manually" filling the metadata database ? Is it possible to synchronize a new node from an existing one ? Is it possible to do it without rebooting gs-server ?
Thank you for your time.
|
|
|
 |
![[Post New]](/templates/default/images/icon_minipost_new.gif) 10 May 2010 14:04:20
|
Mason_S
Senior member
Joined: 1 Apr 2008 09:03:08
Messages: 380
Offline
|
The Advanced Server version of GridSQL makes it slightly easier to add nodes. Once you add nodes though, it copies over the replicated tables, but does not repartition your data. You would have to either dump the data, recreate and reload, or create a new table, do an INSERT SELECT, and then rename the table to the new name.
With open source GridSQL you will have to recreate your database and reload. I also suggest you plan for growth. That is, you may want to create additional logical nodes, then if you want more hardware resources, move some of these off onto new servers.
Regards,
Mason
|
|
|
 |
![[Post New]](/templates/default/images/icon_minipost_new.gif) 26 Oct 2010 11:45:12
|
harry_zhang_2
New member
Joined: 13 Jul 2010 09:34:04
Messages: 1
Offline
|
Mason_S wrote:The Advanced Server version of GridSQL makes it slightly easier to add nodes. Once you add nodes though, it copies over the replicated tables, but does not repartition your data. You would have to either dump the data, recreate and reload, or create a new table, do an INSERT SELECT, and then rename the table to the new name.
With open source GridSQL you will have to recreate your database and reload. I also suggest you plan for growth. That is, you may want to create additional logical nodes, then if you want more hardware resources, move some of these off onto new servers.
Regards,
Mason
hi,Mason_S
you said: With open source GridSQL you will have to recreate your database and reload. I also suggest you plan for growth. That is, you may want to create additional logical nodes, then if you want more hardware resources, move some of these off onto new servers.
what is your mean?
your mean is first i create a database have 32 logical nodes in 4 servers , then when i have more hardware resources, i should migration those server OS system to new server???
thank you !
|
|
|
 |
![[Post New]](/templates/default/images/icon_minipost_new.gif) 26 Oct 2010 17:53:53
|
Mason_S
Senior member
Joined: 1 Apr 2008 09:03:08
Messages: 380
Offline
|
Yes, you could create additional partitions already and migrate them over to new servers later.
Regards,
Mason
|
|
|
 |
|
|
|
|