postgres-xc-general Mailing List for Postgres-XC

On Thu, Jan 5, 2012 at 1:35 AM, Koichi Suzuki <koi...@gm...> wrote:

> Hi,
>
> Running nodes while some of others don't work makes sense only for
> read only transaction.   Update transaction in such situation will end
> up with inconsistent database state.   This is basically why XC does
> not allow such situation.
>
> Please understand that node removal, which will be an issue beyond
> V1.0, should be done in consistent state, that is, all the node should
> be healthy at the removal operation.   If any of them fails, we need
> backup to fail over such nodes by streaming replication or other
> means.
>
> This is what we are assuming.
>
> I welcome any other ideas and proposals of features.   If it comes
> with a patch, this will accelerate the development.
>
> Regards;
> ----------
> Koichi Suzuki
>
>
Hi,
Thank's all for your answers.
I understand your point of view and the postgres-xc situation.
Sorry, but i can help you to write some patch ;) my C is not "fluent" but
i'll continue to follow the evolution of postgress-xc.
Regards,

-- 
Magorn

Hi,

Running nodes while some of others don't work makes sense only for
read only transaction.   Update transaction in such situation will end
up with inconsistent database state.   This is basically why XC does
not allow such situation.

Please understand that node removal, which will be an issue beyond
V1.0, should be done in consistent state, that is, all the node should
be healthy at the removal operation.   If any of them fails, we need
backup to fail over such nodes by streaming replication or other
means.

This is what we are assuming.

I welcome any other ideas and proposals of features.   If it comes
with a patch, this will accelerate the development.

Regards;
----------
Koichi Suzuki



2012/1/4 Ashutosh Bapat <ash...@en...>:
> Hi Magorn,
> Your expectation is right, but as of now, Postgres-XC is not so intelligent
> to understand that it can get the data from the node, which is up and thus
> choose the node which is running over the one which is shut down. In your
> case, it mostly chooses the node which is shut down thus giving the error.
>
> I think it will be a good feature to have.
>
> Regarding your question related to shutting down a server without disturbing
> rest of the cluster - As of now, Postgres-XC is not dynamic, in the sense,
> we can not add or remove nodes at will, neither can we boot or shut them
> down at will. That's something on the list of desirable feature.
>
> On Wed, Jan 4, 2012 at 7:17 PM, Magorn <ma...@gm...> wrote:
>>
>> Hi,
>>
>> I have question about Postgres-XC. I have test it with 2 node, i create
>> "distribute by replication" table. I do some insert.
>> I do SELECT and i see all my records. That's fine :)
>>
>> Next i shutdown one datanode, and i do SELECT that it fail with error :
>> ERROR:  Failed to get pooled connections .
>>
>> Is it normal ? And if is it, when cluster have a lot of node, how stop
>> some node for changer server for example without bloc all the cluser ?
>>
>> Perhaps i miss something about feature of postgres-xc.
>>
>> Regards,
>>
>> --
>> Magorn
>>
>>
>> ------------------------------------------------------------------------------
>> Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex
>> infrastructure or vast IT resources to deliver seamless, secure access to
>> virtual desktops. With this all-in-one solution, easily deploy virtual
>> desktops for less than the cost of PCs and save 60% on VDI infrastructure
>> costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox
>> _______________________________________________
>> Postgres-xc-general mailing list
>> Pos...@li...
>> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general
>>
>
>
>
> --
> Best Wishes,
> Ashutosh Bapat
> EntepriseDB Corporation
> The Enterprise Postgres Company
>
>
> ------------------------------------------------------------------------------
> Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex
> infrastructure or vast IT resources to deliver seamless, secure access to
> virtual desktops. With this all-in-one solution, easily deploy virtual
> desktops for less than the cost of PCs and save 60% on VDI infrastructure
> costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox
> _______________________________________________
> Postgres-xc-general mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general
>

On Wed, Jan 4, 2012 at 10:47 PM, Magorn <ma...@gm...> wrote:

> Hi,
>
> I have question about Postgres-XC. I have test it with 2 node, i create
> "distribute by replication" table. I do some insert.
> I do SELECT and i see all my records. That's fine :)
>
> Next i shutdown one datanode, and i do SELECT that it fail with error :
> ERROR:  Failed to get pooled connections .
>
> Is it normal ? And if is it, when cluster have a lot of node, how stop
> some node for changer server for example without bloc all the cluser ?
>
> Perhaps i miss something about feature of postgres-xc.
>
As said before by Ashutosh, you are missing nothing. For the time being XC
is made to run with a constant number of Coordinators and Datanodes.
There are already basics to have support for dynamic addition and deletion
of nodes, and we are working on strengthening the cluster setup to maintain
global consistency. However, such features as the one you mentionned is
something different that we may target after 1.0 is out.

In case you are also interested in the feature regarding node selection of
a replicated table that fails, you could also give a try and propose a
feature spec to write a patch.
-- 
Thanks,

Michael Paquier
http://michael.otacoo.com