I have been developing my app in a code-first approach atop an exisiting database.
Only now do I have a need to run a migration. I don't want to destroy this database when my model changes because it is very large and it has nice sample data in it.
How can I run a migration without the framework telling me that there are pending changes to apply? (The code runs anyway as I do my migrations from Rails-tho I would like to do them from here)
I do not want to set up automatic migrations because I am working on a big database with lots of seeded data that I do not want to delete/recreate. I also want to have control over what is made, deleteded and when.
This is also needed for when I take it to production, I'd like to roll out the changes via Migration instead of manually. How can I migrate by adding in/removing the fields I want and not have EF care about what it is I do?
If I know how it knows which one it is on (like Rails) can I trick her into thinking that she can run the migrations I want?
I thought that setting the initializer by:
Database.SetInitializer<MyDbContext>(new CreateDatabaseIfNotExists<MyDbContext>
would take acre of it, but it does not.
To answer the question in the title, because there's a sys table __MigrationHistory which tells EF that there is a difference in your tables vs what the database has.
As far as how to do it (from package manager console):
Enable-Migrations
In the configuration class set AutomaticMigrations = false;
Set your Database.SetInitializer<Context>(null) so it doesn't DropCreate or Update
AddMigration <name> to queue any pending changes to a change model
Update-Database will call the MigrationName.Up method to alter the database with any changes (sans losing data).
There's a Table "__MigrationHistory" that EF uses to store Migration Name / Order. You can backup this table in your dev environment, then delete these records. When you deploy to production, you run the migrations. Another option is use Database compare (dev / prod) and get scripts to change your tables / data.
Related
I inherited an ASP.NET MVC application using Entity Framework 6.2. The application was originally built using Code First and I can see migrations that have been applied to the database. The structure of the database seems accurate.
However, a ton of relationships have been added to fix referential integrity issues and these were done directly on the database. There are a bunch of them. Can someone help me to update the models from the database in a code first implementation? Ideally it would remain code first, not a conversion, but I need to do a one time update.
I'm more familiar with newer EF Core and I thought it would as easy as running a scaffold-dbcontext command with a -force and some cleanup. But no such luck in this older version. Is there something similar?
Depending on what a 'ton' means and whether you need to keep the old migrations you could either just rebuild the entire model by doing code first from existing database or you could save the generated code into a folder and do a manual reconstruction:
Run code first from existing database into a new folder.
Map new fields and navigation properties into existing models.
Fix up any fluent code (modelBuilder).
Generate a migration which should include those changes.
A good test would be restoring a pre-change version of the database and running the migration against it. Then perhaps use a schema compare utility.
Apply the migration to databases. If the database already includes the changes, comment out the Up() code before applying. This will update the stored model so the next migration does not repeat these changes.
We have a project running EF6 and have now activated Migrations. The problem we have is that we ha a lot of Data that needs to be added to the database when it is created. This worked well until we needed to add data in the migration part so it is part becomes part of the SQL script. Because when we try and create a new database on our development machines the migration steps are done before the Initializer's seed method and some of the data in the migration step depends on that seed data.
Is there some way to just get the initial database, without any migrations, plus the Initializer's seed data? Or do we need to move the critical data to the InitialCreate mirgations Up method?
The production database will cannot be wiped at this stage and thus we need to add all new production data to the migration step.
Edit: In the end we just moved the effected data to the Up method in migrations, not the way I wanted to do it but we where short on time and this solved the problem.
Migrations also have a Seed method you can override that takes place after the Up() method. You can use the AddOrUpdate command in case data may exist. http://www.asp.net/web-api/overview/data/using-web-api-with-entity-framework/part-3
Using Entity Framework 5, we're using Code-First Migrations in our application. Every developer has his own database on which they work.
I have accidently emptied mine: There is nothing in there anymore, no tables, not even the migration history table.
So, I've tried to update the database again through the PM console by executing update-database. It immediatley gives me an error that a table does not exist, while it should be created in my inital code migration.
What is interesting is that the PM console also shows what migrations are being applied, which does not contain the inital create code migration, thus not creating any tables at all, and ofcourse failing at later migrations.
I tried executing update-database -targetmigration:initialcreate which gives me the message that that code migration does not exist, while it is a direct copy/paste from the cs file, so the id must be correct (note: this works for other migrations).
I also tried update-database -targetmigration:0 and update-database -targetmigration: $InitialDatabase which both give me 'Target database is already at version 0'.
I've also tried deleting the database altogether and let EF create it for, didnt work either, it keeps skipping the initialcreate migration.
So how do I get Code Migrations to execute my initalCreate code migration?
Working with EF migrations in a team scenario is not ideal to say the least. The best practice my team follows is to never commit migrations. Migrations are personal and apply only to your particular database instance. If everyone commits their own migrations you end up with a mess, quick.
While not directly related to your question, some may wonder how you deal with production migrations. Simply, you don't. Your Release Manager, or whoever will actually push the release live, should generate SQL to apply all the changes at once, and then hand this off to your DBA, or whoever manages the production database.
That said, in the scenario that you describe, where you database has been emptied. The best fix is to delete all migrations in your Migrations folder. Even and especially, the initial migration (they don't matter, anyways, because you shouldn't be holding on to them outside of your personal codebase). Then generate a new migration, which will trigger EF to compare your current database state (empty) with the state of the app and essentially create a new initial migration based on the current state of your app. Then, you can apply this migration.
I have a MVC 3 project were I use code based migrations together mith automatic migrations(this works).
When I install this project on a new server the database is created by CreateDatabaseIfNotExist initializer, cause I´m using the seed method of this. After executing this I have a __MigrationHistory table with one entry. The model hash of this entry is exactly the same like the last one from my developement server. On my development server I have an entry for each of my code based migration in the __MigrationHistory table.
Now the problem is that when I try to run the migrations on the new server, I expected them to say to me "nothing to do, cause model hash is same", but instead of this the migrations seems to look only for the MigrationId in the database and try to execute every migration whose MigrationId is missing. Of course this leads to Exceptions, cause the migration tries to add database structures already there.
I think this should be a very common scenario, so is there a kind of workaround for this? My workaround for the moment is to copy all contents from the __MigrationHistory of the development system to the new server, but this is very tricky, due to the dealing with the modelhash as varbinary. Is there a better solution or did I understand some logical things wrong?
I just installed Entity Framework Migrations, added a property to a class, and gave EF Migrations a whirl.
My development database was promptly updated. So far, so good.
Now, I want to create a change script for this initial use of Migrations for the production database. Note there was a pre-existing database because I applied this to an existing project.
The migrations I have are:
PM> Get-Migrations
Retrieving migrations that have been applied to the target database.
201204102238194_AutomaticMigration
201203310233324_InitialCreate
PM>
I thought I could get a delta script using the following:
Update-Database -SourceMigration:201203310233324_InitialCreate -TargetMigration:201204102238194_AutomaticMigration -script
However, that gives me the error:
'201204102238194_AutomaticMigration' is not a valid migration.
Explicit migrations must be used for both source and target when
scripting the upgrade between them.
Just to see what would happen, I reversed the two parameters (backward migration) and did get the script I would expect after adding the -force flag (new columns dropped).
How can I get a script for this first migration?
The right way to start using EF migrations with an existing database is to start with adding an empty migration that contains the metadata of the current database.
I think that you have to roll back to a model that is compatible with the initial database schema. Then run the following command:
add-migration InitialSchema -IgnoreChanges
That should give you an initial migration, that does nothing, but contains the metadata of the current model. You can of course add migrations later with -IgnoreChanges if you've expanded your code model to cover more of the tables already present in the database.
Once you have that initial migration step in place, the scripting would work.
Generally I would not recommend to use automatic migrations unless you only ever intend to only use automatic migrations. If you want some kind of control over the changes to the database (including scripting them) then code-based migrations is the way.