Re: [Postgres-xc-general] JDBC driver for 0.9.4

[Lionel F. <lio...@gm...>]

Hi,

> > Have you setup correctly data_node_hosts and data_node_ports on Coordinator?
Coordinator shows :
num_data_nodes = 2
num_coordinators = 1
data_node_hosts = '10.114.12.26,10.114.12.14'
data_node_ports = '25001,25001'

(However, coordinator does not start with these options

#data_node_users = pgxc
#coordinator_users = pgxc

generated through pgxc_config, so I removed them. Any importance ?)

I see them connected through 'netstat -apn' on both sides, and node 2
connected with gtm.

> Have you used -i option on datanode to be sure that it can accept TCP/IP
> remote connections and not only local ones?

Yes , launch option is -i -p 25001

> Perhaps it is a problem with pg_hba.conf.
It's set to trust on local segment network, and had errors with it
that I corrected, so assuming it's working now :) psql working
properly on creating database remotely.

> gtm_host and gtm_port are correctly set for both Coordinator and datanodes.

They are :
Node 2 :
datanode/postgresql.conf:gtm_host = '10.114.12.26'
datanode/postgresql.conf:gtm_port = 16680

Node 1
coord/postgresql.conf:gtm_host = '10.114.12.26'
coord/postgresql.conf:gtm_port = 16680
datanode/postgresql.conf:gtm_host = '10.114.12.26'
datanode/postgresql.conf:gtm_port = 16680

Of course, gtm running on 16680. And confirmed they are TCP-connected
through netstat.

> Is pgxc_node_id set differently for your datanodes?
Yes, but doc is not crystal clear on that; are pgxc_nodes to be all
different between coord & data nodes ? Meaning

coord=1, data1=2, data2=3
or
coord=1, data1=1,data2=2 ?

(already tried some combinations...)

On first JDBC query, coordinator shows an "EOF on client connection".

Thanks (and thanks to Benny for the fix, will compile in parallel the
newest build)

Lionel Frachon



2011/5/27 Michael Paquier <mic...@gm...>:
> This may be a setting problem.
> Have you setup correctly data_node_hosts and data_node_ports on Coordinator?
> If your coordinator cannot connect to a backend datanode it may be this
> problem.
>
> Have you used -i option on datanode to be sure that it can accept TCP/IP
> remote connections and not only local ones?
> Perhaps it is a problem with pg_hba.conf.
> Have you set up this file for both nodes correctly?
>
> If a node complains about GTM no accessible, you may need also to check if
> gtm_host and gtm_port are correctly set for both Coordinator and datanodes.
>
> You have also to take care about the node IDs used in each configuration
> file.
> Is pgxc_node_id set differently for your datanodes?
>
> On Fri, May 27, 2011 at 2:47 PM, Lionel Frachon <lio...@gm...>
> wrote:
>>
>> Hi Michael,
>>
>> I believe this is the kind of query generated by BenchmarlSQL on the
>> "load data" stage; trying to load a 2-node cluster yesterday
>> (1gtm+coord+datanode, 1 datanode), only one is loaded, and the other
>> one complains saying "No GTM snapshot available" and thus not loading
>> anything. Table content is not accessible either from the second node.
>> I tried to switch the preferred_node and master node settings, to try
>> to force the writes on second node first, but without positive result.
>> However, a std psql query distributes the work properly when issued
>> through coordinator.
>>
>> I'll try to revert today to the previous build (05/18) to see if
>> there's any behaviour difference.
>>
>> Any clue appeciated ^^
>>
>> Lionel Frachon
>>
>> 2011/5/26 Michael Paquier <mic...@gm...>:
>> > Btw, if I were you I'd avoid to use multiple-value INSERT with JDBC.
>> > Those
>> > are queries like:
>> > INSERT INTO table_ex VALUES (1),(2),(3),(4);
>> > I noticed today during tests that it wasn't working properly.
>> > However, single value INSERT and normal queries work fine.
>> > --
>> > Michael Paquier
>> > http://michael.otacoo.com
>> >
>
>
>
> --
> Michael Paquier
> http://michael.otacoo.com
>