You can do

OdbcCommand cmd = new OdbcCommand("SET partialResultsMode", "true");

before running the your query.

For more information on SET Command see SET Statement - Teiid 8.7 (draft)

The pg driver logic typically only sends well know parameters pg when connecting.  Something Teiid specific that is applicable via a set statement would need to be done after the connection as Ramesh shows or as part of an initialization string if the client supports it.

Use "SET partialResultsMode true" without the =.

Currently that is not possible.

@Steve: I believe we need to go with your suggestion of adding the execution properties to vdb.xml as supported properties as this seems to be much more useful to configure them at vdb level.

Ramesh..

When you are deploying the VDB (I am guessing you are using dynamic VDB) there should be at least one datasource available, to read in the metadata. After that it should be fine if one or both sources are down. I am suspecting when you deployed the VDB both your sources are down, is it?

Also for multi-source to work both schema need to be identical, they can not have different tables names with different schema names in both databases.

Ramesh..

> if the second has table PC90 which the first doesn't have ,when we execute query "select * from PC90 " ,it will occur the error above.

Yes, the current assumption is that the metadata is the same across multi-source sources, so metadata is only loaded from 1 source.

> is there other methods to trigger the vdb reload to recognise the new table ?

A redeploy through the admin deploy method, or an admin restart with the restartVDB method.

The ability of the data virtualization does end with multi-source or querying the single sources under a VDB.  If you choose to import metadata from each individual source, then the metadata can be different in each of the source. Then you can create VIRTUAL models (schemas) using JBDS. In virtual models you can define Virtual Tables or Views that can be derived from different tables from different sources.

For example you TableOne in ModelOne, and TableTwo in ModelTwo, then you can create a virtual model and have a CombinedTable that UNIONS both the TableOne and TableTwo. Then you can simply issue "select * from CombinedTable". Look at quick starts Data Federation Example

Also, you do not have to create virtual models before you can integrate data from two separate models, you can do what you showed above like select id,name from  (select id ,name from model1.test union all select id ,name from model2.test) e" from client. However, if you would like to create abstraction/logical layer between your client code and physical sources, virtual models is the way to go.

Ramesh..

> how can i import metadata from each individual source in a multi-source model?

You currently cannot in a built-in way.  The underlying assumption is that the sources are the same.  To do what you want, there are a couple of possibilities:

1. Use different models for each source and manually create the multi-source views:

create view test_ms as select id ,name, 'model1' as SOURCE_NAME from model1.test union all select id ,name, 'model2' from model2.test

Then you can issue queries against test_ms similar to the multi-source scenario: 

"select * from test_ms" - goes to both sources

"select * from test_ms where source_name = 'model1'" - goes to just the model1 source

2. Use a multi-source model, and single source models.

Using the importer schema/table pattern or excludeTables settings you could have a model that represents just the tables that are present in all sources, then add additional models that represent just the single source table from each source.  There could be some downside here in that the optimizer may not be able to fully optimize some scenarios that utilize both the multi-source and single-source models.

3. Still use the multi-source concept, but create a custom translator that will return a super set of the metadata rather than just the metadata for the first available source.  This requires extending a translator or using a delegating translator that would look at the Model on the MetadataFactory to determine the sources and interrogate each one.  However this will not work very well as the optimizer will assume that a multi-source single source table is available on each source, so you may generally not get the results you expect unless the query is specifically targeted at the source containing all of the tables in the query.