Restoring Deleted Databases on CloudSQL - google-app-engine

I know GCP's CloudSQL supports instance backups of some kind. Also noted that they make it really easy to create and delete databases from the CloudSQL instance details -> Databases tab.
I am concerned about how easy it is to delete a database, especially since mistakes happen. Is it possible at all to restore an accidentally-deleted db? I cannot find any docs on how one would go about such a scenario. Or better yet, how to prevent such accidental deletes?

Just for clarification, I'm discussing deleting databases on an instance, and not the instances themselves.
Even with a regular SQL instance, deleting a database is very easy. As long as you have the correct permissions, all it takes is the following:
DROP DATABASE db_name
So if you want to prevent someone from deleting a database, just don't give them the DROP permission on the database.
Cloud SQL follows the same logic. If you check out the docs for deleting a database, you'll see it requires the sqlservice.admin scope. Avoid granting users this IAM permission, and they won't be able to delete an instance.

Related

Is there a way to create a constraint that prevents updates in a SQL Server 2012 database table?

I am creating an application that will track hours for employees. Ideally, HR has asked that certain tables not be modified once data is commited. This is done easily enough from the front-end and stored procedures. However, it would be great to be able to prevent it from the server itself through constraints so that folks that have access to the back-end data can't change any values in the selected tables (unless they are sneaky enough to know how to disable the constraints).
If you trust your SQL Server admins then it’s possible. Have your admin to create users that don’t have datawriter permissions for those tables or schema.
So, application would write the data into database and users who have access to those tables would only be able to read the data.
If you don’t want admins to have the ability to modify data that’s not possible. There is no way to prevent it but there is a way to detect it if it happens. Check out this article for details on details how to this is done in third party application and see if it helps.
Use Server Side security roles to give only the HR Group data-write privileges.

How to configure database permissions for a Django app?

I'm looking for links, or an answer here, on to how to properly configure the database permissions to secure a Django app? To be clear, I'm looking specifically for material dealing with grants on the database, not permissions within the Django framework itself.
From the django docs:
https://docs.djangoproject.com/en/dev/topics/install/
If you plan to use Django’s manage.py syncdb command to automatically create database tables for your models (after first installing Django and creating a project), you’ll need to ensure that Django has permission to create and alter tables in the database you’re using; if you plan to manually create the tables, you can simply grant Django SELECT, INSERT, UPDATE and DELETE permissions. On some databases, Django will need ALTER TABLE privileges during syncdb but won’t issue ALTER TABLE statements on a table once syncdb has created it. After creating a database user with these permissions, you’ll specify the details in your project’s settings file, see DATABASES for details.
I've just tested initial setup with MySQL. For python manage.py migrate at least you need following grants for simple operation (if yo use db-preparation):
CREATE, ALTER, INDEX
SELECT, UPDATE, INSERT, DELETE
And, by the way - security matters. You can reduce attack impact by limiting your system exposure. In this case - you can restrict 'DROP' - which is fairly huge plus. If you leave some tricky hole with ability to SQL-inject - you probably reduce the damage. I will research in the future if it will not do any harm to remove DELETE keyword - that would limit potential threats as well. Just because we all leave bugs from time to time :)
I usually:
grant all privileges on my_db.* to my_user#localhost identified by 'my_user_pass'
grant all privileges on test_my_db.* to my_user#localhost identified by 'my_user_pass'
I suppose if there were a bug in django, you might be opening your database up to terrible things, but you'd have other problems if there were that big of a security hole in django.
django minimally needs select, insert, update, and delete, to operate. If you're using test or syncdb at all, you'll also need to be able to create tables, and indexes (and maybe the file permission for loading sql fixtures).
So, for a mysql db, I'd guess the optimal set of permissions might be select, insert, update, delete, create, index, and file. If you wanted to get real nitty-gritty, you could selectively grant these permissions as appropriate on the table level (rather than the db level).
Personally, I find grant all ... easier to type.
What's the purpose of configuring permissions on DB level? If your server is compromised then the attacker will be able to do anything with your database (because he has the login/pass) and permissons won't help. If your server is secured then permissions are useless.
Permissions can make sense if your DB server is available from the outer world, but it is not a good idea to do so.

SQL Server Login Configuration and Automation

What's a good way to manage and maintain SQL Server logins, server roles, and individual access rights across multiple databases that exist in multiple environments? What are your best practices?
Some info about my situation:
SQL Server 2005
We have N amount of "client" databases with identical schemas (in theory, at least)
We have a central "admin" database that references each client database and can hold configuration values
This "admin/client" pattern is duplicated across multiple environments (dev/qa/stage/prod)
Some users, like testers, need different rights based on evironment
We frequently have to pull client db backups from one environment to restore on another for development or testing purposes
We keep our stored procedures and scripts in source control and deploy in a build cycle
Right now my organization is chaotic and we don't follow good security practices. We have no formal DBA. However, if we got any more complex it would be a constant hassle to maintain it all the time. I could see migrating to a new server or recovering from disaster being extremely time consuming if we where to attempt configuring it directly through the management studio IDE.
First, to make restoring a database to a different server easier, make sure that your logins all have the same SID on all of your servers by using the sp_help_revlogin stored procedure from Microsoft to script the login on the first server you create it on and then use the script to create the login on your other servers. This keeps the database user mapped to the login correctly when you restore the database.
Having different permissions at the database level depending on the environment is going to be a hassle to a point no matter how you role this out. I have a stored procedure in master that gets called on my Dev Server as a part of my restore process that performs the additional GRANT's on the database to give the developers access to make changes. That's the best I have been able to come up with to solve similar problems.
A way to make the rights easier would be to create rolls in the database called Dev, QA, Test, Prod and grant the correct rights to those roles. Then as you restore the databases to each environment just drop the developers in the correct role.
We use active directory groups and enforce windows authenticated logins. From within SQL Server we can then define access based on the AD group the user is in by creating a single SQL Server login per AD group. Not sure if this is any better or worse than DB roles, but it means the roles are managed outside each database.
Propagating access to databases is then either a manual operation or a short SQL script to ensure the logins in the database point to a valid SQL Server login (which in turn is an AD group).
Generally this works well for the general case. We can use DB roles then to assign the builtin roles (e.g, db_datareader) to each AD group
Rarely someone needs some specific access to a database outside this model. We either end up opening it up to the group as a whole if it's not going to be invasive or critical or we'll end up creating a per-user account that has to be managed separately. We endevour to keep these to an absolute minimum, and clean them up every now and then so they're not abused/forgotten about.

How to have a "master-structure" database with "children-data" databases in SQL SERVER 2005?

I have been googling a lot and I couldn't find if this even exists or I'm asking for some magic =P
Ok, so here's the deal.
I need to have a way to create a "master-structured" database which will only contain the schemas, structures, tables, store procedures, udfs, etc, everything but real data in SQL SERVER 2005 (if this is available in 2008 let me know, I could try to convince my client to pay for it =P)
Then I want to have several "children" of that master db which implement those schemas, tables, etc but each one has different data.
So when I need to create a new stored procedure or something like that, I just create it on the master database (and of course it's available on its children).
Actually I have several different databases with the same schema and different data. But the problem is to maintain congruency between them. Everytime I create a script to create some SP or add some index or whatever, I have to execute it in every database, and sometimes I could miss one =P
So let's say you have a UNIVERSE (would be the master db) and the universe has SPACES (each one represented by a child db). So the application I'm working on needs to dynamically "clone" SPACES. To do that, we have to create a new database. Nowadays I'm creating a backup of the db being cloned, restoring it as a new one and truncate the tables.
I want to be able to create a new "child" of the "master" db, which will maintain the schemas and everything, but will start with empty data.
Hope it's clear... My english is not perfect, sorry about that =P
Thanks to all!
What you really need is to version-control your database schema.
See do-you-source-control-your-databases
If you use SQL Server, I would recommend dbGhost - not expensive and does a great job of:
synchronizing 2 databases
diff-ing 2 databases
creating a database from a set of scripts (I would recommend this version).
batch support, so that you can upgrade all your databases using a single batch
You can use this infrastructure for both:
rolling development versions to test, integration and production systems
rolling your 'updated' system to multiple production deployments (especially in a hosted environment)
I would write my changes as a sql file and use OSQL or SQLCMD via a batchfile to ensure that I repeatedly executed on all the databases without thinking about it.
As an alternative I would use the VisualStudio Database Pro tools or RedGate SQL compare tools to compare and propogate the changes.
There are kludges, but the mainstream way to handle this is still to use Source Code Control (with all its other attendant benefits.) And SQL Server is increasingly SCC friendly.
Also, for many (most robust) sites it's a per-server issue as much as a per-database issue.
You can put things in master like SPs and call them from anywhere. As far as other objects like tables, you can put them in model and new databases will get them when you create a new database.
However, in order to get new tables to simply pop up in the child databases after being added to the parent, nothing.
It would be possible to create something to look through the databases and script them from a template database, and there are also commercial tools which can help discover differences between databases. You could also have a DDL trigger in the "master" database which went out and did this when you created a new table.
If you kept a nice SPACES template, you could script it out (without data) and create the new database - so there would be no need to TRUNCATE. You can script it out from SQL or an external tool.
Little trivia here. The mssqlsystemresource database works as you describe: is defined once and 'appears' in every database as the special sys schema. Unfortunately the special 'magic' needed to get this working is not available to the user databases. You'll have to use deployment techniques to keep your schema in synk. That is, apply the changes to every database as the other answers already suggested.
In theory, you could put a trigger on your UNIVERSE.sysobjects table (assuming SQL Server), and then you could enumerate master.dbo.sysdatabases to find all the child databases. If you have a special table that indicates it's a child database, you can reference child.dbo.sysobjects to find it.
Make no mistake, it would be difficult to implement. But it's one way you could do it.

Which database implementations allow sandboxing users in separate databases?

Can anyone tell me if there are RDBMSs that allow me to create a separate database for every user so that there is full separation of users' data?
Are there any?
I know I can add UID to every table but this solution has its own problems (for example per user database schema changes are impossible).
Doesnt MySQL, PostgreSQL, Oracle and so on and so on allow you to do that?. There's the grant statements to control ACLs
I would imagine most (all?) databases allow you to create a user which you could then grant database level access to? SQL server certainly does.
Another simple solution if you don't need the databases to be massive or scalable, say for teaching SQL to students or having many testers work against their own database to isolate problems is SQLite, that way the whole database is a single file (per user), and each user cannot possibly screw up or interfere with other users.
They can even mail you the databases, or install them anywhere, say at home and at work with no internet required.
MS SQLServer2005 is one which can be used for multiple users.An instance can be created
if you have any, run the previlegs and use one user per instance
Oracle lets you create a separate schema (set of tables, indexes, functions, etc) for individual users. This is good if they should have separate different tables. Creating a new user could be a very expensive operation as you would be making new tables. Updating is a nightmare as well, as you need to update the model for each user.
If you want everyone to have the same set of tables, but only able to view their own records then you could use Fine Grain Access Control or Virtual Private Database features to do this.

Resources