[Postgres-xl-developers] Addition of datanode master fails

Pavan Deolasee pavan.deolasee at gmail.com
Sun Jan 18 21:49:42 PST 2015

On Sun, Jan 18, 2015 at 1:23 PM, Ramanuj Hari <hari.ramanuj at gmail.com>

> Hi,
> On further test, for Behaviour 3, I have perform below command for each
> table:
> ALTER TABLE af_device_l2 TO NODE (datanode1, datanode2), DISTRIBUTE BY
> after that it seems its working. Now data is redistributed properly and
> inserts are happing to both datanode.
I'm glad you found that.

> Wondering why 'add datanode master ..." command should have included above
> 'alter table ..' in it's steps.
The decision to not automatically redistribute existing tables to include
new node is because we allow tables to be distributed on only a subset of
nodes. So the system would not know if the new node is to be used for
existing tables, at least not for the tables which are distributed on a
subset of nodes. Another important reason is that redistribution is an
expensive operation. In fact, in the current XL release it will grab an
exclusive lock on the table under redistribution, thus cutting access to
all other sessions. So we would want the DBA to control that activity.

> Also was looking at :
> http://files.postgres-xl.org/documentation/add-node-datanode.html
> last step is mentioned as 'Redistribute existing data by using ALTER TABLE
> REDISTRIBUTE. The new datanode is now ready.' I didn't find any command
> called 'ALTER TABLE REDISTRIBUTE', could you please help if I am missing
> something.

Yes. I think thats a typo or leftover from some previous implementation.
Thanks for pointing that out. Will take care of that in the docs.


Pavan Deolasee
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.postgres-xl.org/private.cgi/postgres-xl-developers-postgres-xl.org/attachments/20150119/a157c26b/attachment.htm>

More information about the Postgres-xl-developers mailing list