Sql-server – Access a DB through Another DB ONLY

The basic concept is to use GRANT/DENY Schema Permissions. You can efficiently manage permissions by creating a role and then adding members to it.

Below is an example that will explain you in detail

use master
go
--Create Logins
CREATE LOGIN UserA WITH Password='UserA123';
go
CREATE LOGIN UserB WITH Password='UserB123';

use AdventureWorks2008R2
go
--Create Database Users
CREATE USER UserA;
go
CREATE USER UserB;
go
--Create the Test Schemas
CREATE SCHEMA SchemaA AUTHORIZATION UserA
go
CREATE SCHEMA SchemaB AUTHORIZATION UserB
go

-- create test tables
create table schemaA.TableA (fname char(5))
go
insert into schemaA.TableA (fname) values ('Kin-A')
go

create table SchemaB.TableB (fname char(5))
go
insert into SchemaB.TableB (fname) values ('Kin-B')
go

Now test :

--Test for UserA in SchemaA
EXEC('select * from schemaA.TableA') AS USER = 'UserA'
go
--Kin-A

-- Test for UserB in SchemaB == this should fail
EXEC('select * from SchemaB.TableB') AS USER = 'UserA'
go
--Msg 229, Level 14, State 5, Line 1
--The SELECT permission was denied on the object 'TableB', database 'AdventureWorks2008R2', schema 'SchemaB'.

Now create Stored Procedures :

CREATE PROCEDURE SchemaB.proc_SelectUserB
AS
    select * from schemaA.TableA;
go
create procedure schemaA.proc_SchemaA
as 
    select * from schemaA.TableA

Now Grant execute permissions to UserA on schemaB's SP

GRANT EXECUTE ON OBJECT::[SchemaB].[proc_SelectUserB] TO [UserA] 
go

Test it .. to see if UserA is able to run SP from schemaB. This will PASS

EXECUTE AS LOGIN='UserA';
    Exec SchemaB.proc_SelectUserB;
    revert;
go
--- Kin-A

But UserA wont be able to see data from SchemaB

EXECUTE AS LOGIN='UserA';
    select * from SchemaB.TableB
revert;
go

--- Msg 229, Level 14, State 5, Line 3
--- The SELECT permission was denied on the object 'TableB', database 'AdventureWorks2008R2', schema 'SchemaB'.

Alternatively you can use DATABASE ROLE and just add users to it for better manageability of permissions:

EXEC sp_addrole 'SchemaBUsesSchemaAProc'
go
EXEC sp_addrolemember 'SchemaBUsesSchemaAProc','UserA';
go

Below statement will make sure that UserA is able to see schemaA and NOT schemaB. The good thing is that you can just add users to SchemaBUsesSchemaAProc role and they will inherit all the permissions granted to that role.

GRANT SELECT ON SCHEMA::SchemaA TO SchemaBUsesSchemaAProc;
go

If you only want to allow UserA to execute SP's which are owned by SchemaB then below statement will do the job:

GRANT EXECUTE ON OBJECT::[SchemaB].[proc_SelectUserB] TO [SchemaBUsesSchemaAProc] 
go

This way, UserA is not able to see SchemaB's tables, but still can execute procs from SchemaB.

Below will explain the permission hierarchy :

I would recommend to use log shipping but only if you do not want to prevent delete operations. Log shipping keeps the target database the same as the source database. Log shipping is a relatively simple mechanism, it just backups and copies over the transaction log and applies it to the target database. It is a bit simpler to setup and to maintain than Replication.

Replication also keeps the source database completely in sync with the target database. With transactional or snapshot replication there is no possibility to exclude delete operations. The only exception is that in merge replication exclusion of DELETE statements is still possible. However, this possibility will be removed in future versions of SQL Server (see the first note in https://msdn.microsoft.com/en-us/library/ms146947.aspx). So that is nr. 1 of your worries I think. Some more considerations can be found here: https://technet.microsoft.com/en-us/library/ms151206(v=sql.105).aspx.

Performancewise there also can be a big impact. This depends a lot on the database and the operations on it. So test it in a test environment to see if the performance loss is acceptable. Personally, I'm not a huge fan of merge replication as I have experienced many problems with it.

Summarizing, if you want to prevent delete operations I would recommend to write your own synchronization mechanism. If you don't want to do that, and can live with the issues mentioned above, merge replication will also be usable.