I am building a multi tenant application with NodeJS and MSSQL on the backend which will have one instance running. I think about using schema based multi tenancy architecture, meaning single database with separate schema for each tenant `[Tenant1].[DataTable1], [Tenant2].[DataTable1]`.
My question is what is the better way to implement this to handle correct "routing" of SQL queries from different tenants?
I came with four solutions:
Use Connection Pools - each tenant user will have a separate SQL connection with a different database user (who has default schema and access to it only).
Use single SQL connection but add EXECUTE AS USER = 'Tenant1' .. REVOKE to each query (database user has default schema and access to it only).
Or have shared database with single schema
Row-Level Security, but it seems that it could have a performance impact.
Add Tenant_table and insert tenant_id into each row.
Related
I need to restore the database from a different environment where different logins are used but I want to keep permission to this database for already existing login on this server.
Now, after restoring my login on the targeted server it lost mapping for this database (and owner role)
I want to restore everything inside the database and same time I don't want to map the database and add owner role to the login.
Is it even possible?
I've tried different ways of restore - deleting before restore, as well as delete but not to close existing connections. The result is the same, I need to map new restored DB again and add the owner role.
Normally database users have to be mapped to instance logins to work, but if you want to make your DB more instance independent and control your database from the inside of it then the solution for you would be a Contained Databases.
From MS BOL:
A contained database is a database that is isolated from other
databases and from the instance of SQL Server that hosts the database.
SQL Server 2017 helps user to isolate their database from the instance
in 4 ways.
Much of the metadata that describes a database is maintained in the database. (In addition to, or instead of, maintaining metadata in the
master database.)
All metadata are defined using the same collation.
User authentication can be performed by the database, reducing the databases dependency on the logins of the instance of SQL Server.
The SQL Server environment (DMV's, XEvents, etc.) reports and can act upon containment information.
How to use an external SQL database with Keycloak?
I have a sql database that contains my clients from my application.
If you have a pre-existing database with your own custom table structure for identity and access (most likely, you rolled your own user and role tables and probably some other tables as well), you can implement Keycloak's "User Storage SPI" in order to connect your database to Keycloak.
To do this, see the docs.
Now I can use beeline to create role and user, every user have his/her own database using admin role to grant, but I need to create a new database that all users can create tables in the public database, how to create database like that?
Users can not create tables in other databases and can only create tables under their own databases.
You need an authorization provider that supports Hive. Apache Sentry is a popular one.
You will need to create a Sentry role that allows access to certain databases only.
Then assign a (Linux) group to that role.
Any user in above group will get access privileges to certain databases only (because of step 1.)
I am transitioning a project from Advantage Database Server to MS SQL server. In Advantage, you can password protect an individual table, which is also encrypted. As such, you cannot open, view, update, etc. the table without the password. I place my project's registration information in this table, so I don't want any user to be able to look at its contents.
I cannot find a similar function in SQL server. Encrypting the data is insufficient. So my question is: is there a way to password protect a table in SQL Server.
In SQL server you can link various access roles to the users. These roles can be applied to tables, views, stored procedures etc. The best thing to do is to create views on the database, and let the users access specific views, rather than giving permissions on all DB objects.
Alternatively, you can deny permissions on a specific table to a user or a role.
Here are two articles on MSDN that will get you started:
GRANT Object Permissions
DENY Object Permissions
I have MS Access as a front end and PostgreSQL as back end for my database. So I set up the database in PostgreSQL and linked the tables to MS Access using the ODBC drivers. Everything works great, I can update the tables in MS Access and the record will appear in Postgres database.
Since I can still see the linked tables in MS Access, I feel like it is possible for some users to go in and manually modify the tables without filling out proper forms. Is it possible to HIDE the tables or lock the tables so that Access users cannot modify the raw data at all? If not, what can I do to secure the integrity of the database.
Thanks!
I would recommend looking at Postgres privileges as a way to lock the tables down.
In short, you could have your backend run as one user that has full access permissions on the tables in question, and when the users login to the app, they would be connected to Postgres using a user whose privileges are considerably more locked down (say, read only if you just want to be able to do SELECTs to surface data).
For example, you could run the following SQL against your Postgres server:
REVOKE ALL ON accounts FROM joe;
GRANT SELECT ON accounts TO joe;
Which would first remove all privileges from the user joe for the table accounts, and then allow only SELECT priveleges for that table.
You could do something similar for all the tables you wish to lock down. You'll also need to do the same for the sequences used by those tables.
You may wish to create a special readonly user which has only read access across the board, and use those credentials to surface the Postgres data for the users to access.
When you need to alter data, your backend could specifically use a power user of sorts which has much greater access.
Here's a link which details creating a readonly Postgres user (for purposes of backups in this case, but the general concept and the SQL commands should apply (just ignore the stuff about pg_dump).
If you aren't concerned about users' ability to modify the data in those tables via the up other than in the ways that are authorized, but are only concerned about them using, say, psql to go in and update them, then you probably don't need a readonly user, but can simply lock the tables down and have the backend use that user with sufficient access.