[Postgres-xl-general] Why Recursive limited to replicated tables?

Petr Chmelar petr.chmelar at greycortex.com
Fri Sep 23 10:44:42 PDT 2016


Thanks Krzysztof,

however this doesn't solve Mahesh's question, or does it?

I didn't wrote it clearly, but I understand it differently - we need to 
execute a query on a distributed table (from the point of XL cluster) 
when we know all the data is in a single node. Probably the planner can 
be the same as the plain PostgreSQL has, which I doubt, but the node can 
definitely look at the table it has the complete data (same as 
distributed table) for the particular node only, similarly to plain 
PostgreSQL. Of course this can break things, but it's far more efficient.

Do you agree with me, Mahesh? Do you think we should implement such feature?

Cheers,

Petr


On 23.9.2016 18:11, Krzysztof Nienartowicz wrote:
> I think this should be already achievable by defining groups or using
> "to node/to group" syntax on replicated table.
>
> http://files.postgres-xl.org/documentation/sql-altertable.html
>
> [
>    DISTRIBUTE BY { REPLICATION | ROUNDROBIN | { [HASH | MODULO ] (
> column_name ) } } |
>    DISTRIBUTED { { BY ( column_name ) } | { RANDOMLY } |
>    DISTSTYLE { EVEN | KEY | ALL } DISTKEY ( column_name )
> ]
> [ TO { GROUP groupname | NODE ( nodename [, ... ] ) } ]
>
> just use 'to node' on replicated table.
>
> Cheers
>
>
> On Fri, Sep 23, 2016 at 5:20 PM, Petr Chmelar
> <petr.chmelar at greycortex.com> wrote:
>> Hi XL-ers,
>>
>> at the moment, I'm considering (starting the work on) such a feature
>> that some data or even tables might not be distributed or replicated at
>> all - sitting in a single node only. In this way, you can connect
>> directly to the datanode and many other possibilities without the need
>> to get a global TID from the global monitor. You may also use this
>> feature to create and modify (unlogged?) data while the global monitor
>> or some other nodes are unavailable.
>>
>> If you find this feature useful, mail me, please.
>>
>> Cheers,
>>
>> Petr
>>
>>
>> On 23.9.2016 17:04, Krzysztof Nienartowicz wrote:
>>> Because the planner does not have any knowledge about your data
>>> distribution and cannot treat a table as a local one.
>>>
>>> Cheers
>>>
>>> On Fri, Sep 23, 2016 at 3:59 PM, Mahesh Viraktamath <yuva670 at gmail.com> wrote:
>>>> Hello All,
>>>>
>>>> I can't understand as to why should RECURSIVE be only limited to replicated
>>>> table? For example, as table (on which I run the RECURSIVE) is distributed
>>>> in such a way that all the required data will reside in the same datanode.
>>>> So, shouldn't it be easy to execute it if all the data (expected by the
>>>> REVURSIVE) is on the same datanode.
>>>>
>>>>
>>>> Thanks,
>>>> Mahesh
>>>>
>>>> ------------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Postgres-xl-general mailing list
>>>> Postgres-xl-general at lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/postgres-xl-general
>>>>
>>> ------------------------------------------------------------------------------
>>> _______________________________________________
>>> Postgres-xl-general mailing list
>>> Postgres-xl-general at lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/postgres-xl-general
>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> Postgres-xl-general mailing list
>> Postgres-xl-general at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/postgres-xl-general





More information about the postgres-xl-general mailing list