35
votes

I am using EF Core 2.1

This was my initial model definition.

public class Customer //Parent
{
    public int Id { get; set; }

    public string Name { get; set; }

    public string Email { get; set; }

    public BankAccount BankAccount { get; set; }

}


public class BankAccount
{
    public int Id { get; set; }

    public string Branch { get; set; }

    public string AcntNumber { get; set; }

    public DateTime CreatedDate { get; set; }

    public int CustomerId { get; set; }

    public Customer Customer { get; set; }

}

But I realized having Id & CustomerId both is overhead as its One-to-One relation, I can update my BankAccount model definition as below.

public class BankAccount
{
    public int Id { get; set; }

    public string Branch { get; set; }

    public string AcntNumber { get; set; }

    public DateTime CreatedDate { get; set; }

    public Customer Customer { get; set; }

}

While in DbContext class defined the principal entity as below.

HasOne(b => b.Customer).WithOne(c => c.BankAccount).HasForeignKey<BankAccount>(f => f.Id);

While running the update-database I am getting the below error.

System.InvalidOperationException: To change the IDENTITY property of a column, the column needs to be dropped and recreated.

However, ideally I should not but just get rid of this error, I deleted the column, constraints and as well table and then the complete database as well. But still the same error.

12
what do you mean by '...I realized having Id & CustomerId both is overhead as its One-to-One relation'? Id is the primary key and CustomerId is a foreign key so what's the overhead here? - Elyas Esna
@ElyasEsna, One Customer One BankAccount, so the Id in Customer entity could be used as FK & primary key in BankAccount entity - Kgn-web
seems to be the issue github.com/aspnet/EntityFrameworkCore/issues/7444 , which is an open issue in EF core. - DevilSuichiro
To future answerers: people keep piling up "solutions" to this question, all amounting to roughly the same thing and none of them really suitable for a production environment that doesn't allow data loss and has foreign keys to the primary key. If you have a solid solution taking that into account, please post it, otherwise please think twice before adding more noise - Gert Arnold

12 Answers

30
votes

I ran into the same problem, and I solved it by two steps and two migrations:

Step 1

  1. Drop the identity column.
  2. Comment the ID in BankAccount and add a new one (i.e., BankAccountId as
    identity, add migration and update - this drops id).
  3. Add a new column as identity.

Step 2

  1. Drop the newly added column and re-add the previous one. Comment BankAccountId and un-comment ID.
  2. Add migration and update (this drops the BankAccountId and adds Id as identity).
8
votes

I had this problem when I tried to change a model from public byte Id {get; set;} to public int Id {get; set;}. To face the issue, I did the following things:

  1. Remove all the migrations until the creation of the target model with Remove-Migration -Project <target_project> in the Package Manager Console
  2. Delete the actual database
  3. If you have some migrations in the middle that you have not created, (for example they came from another branch), copy the migrations files and also the ModelSnapshot file and paste them in your branch (overwrite them carefully!).
  4. create a new migration with add-migration <migration_name> in the Package Manager Console
  5. update the database with update-database in the Package Manager Console

I can solve it in this way because my code was not in a production environment. Maybe you have to face another complex issues if the model is already in there.

5
votes

This error occurs when you try to alter or modify a table that already exists when you want to change schema or the table that already exists, which EF core Doesn't support it yet it needs manual action. here is what you can do about this:

  • Comment related code in migration file to avoid this error.
  • or Remove migration files and create a fresh one.
  • Delete upstream migration and let migration generate new code.
3
votes

I had to:

  1. Comment the table totally from code
  2. Run a migration that clears it from DB
  3. Uncomment table again with corrected mapping
  4. Run migration again

Done

2
votes

In my opinion running the EF Migrations against anything but your development database is asking for trouble as you are naturally limited by the fact that EF migrations will sometimes flatly refuse to work when altering the structure of you objects (changing primary keys and changing foreign keys being the most often encountered).

For many years I have used tools to ensure DB schema is included in version control (complementary to EF migrations). Do your developments to change your dev database (where the data is not important), create multiple migrations but then use the tools to roll these up into a DB deployment script.

Here’s a summary of what I would do in this case: -

  1. Remove (comment out) all references to the old class BankAccount
  2. Create Migration and apply to dev database
  3. Re-Add the BankAccount class with it’s corrected definition
  4. Create Migration and apply to dev database
  5. Use a DB comparison tool (my preference is APEX SQL Diff, but there are other in the marketplace) to create a deployment script that rolls up both migrations.
  6. Test this script on your staging environment (where you should have some data)
  7. If test is good apply to Production

The reality is if you have production data that you want to radically change the structure of with a code first approach it will probably end badly for you unless you understand and address the data migration from one structure to the other.

2
votes

In my case, table SharedBalances is renamed to Balances and it's identity column SharedBalancesId is renamed BalanceId. SQL commands are executed on SQL Server. You can also try migrationBuilder.Sql(my_sql_command_here)

I created the migration and got the same error.

Rename the column and the table using TSQL command:

EXEC sp_RENAME 'SharedBalances.SharedBalanceId', 'BalanceId', 'COLUMN';

EXEC sp_RENAME 'SharedBalances', 'Balances';

-- Caution: Changing any part of an object name could break scripts and stored procedures.

Comment the RenameTable command in your migration:

/*
migrationBuilder.RenameTable(
    name: "SharedBalances",
    newName: "Balances");
*/

Comment the AddPrimaryKey command in your migration:

/*
migrationBuilder.DropPrimaryKey(
    name: "PK_SharedBalances",
    table: "Balances");
migrationBuilder.AddPrimaryKey(
    name: "PK_Balances",
    table: "Balances",
    column: "BalanceId");
*/

Update occurences of the table name DropForeignKey commands in your migration:

From this....

        migrationBuilder.DropForeignKey(
            name: "FK_SharedBalances_Users_OwnerUserId",
            table: "SharedBalances");

        migrationBuilder.DropPrimaryKey(
            name: "PK_SharedBalances",
            table: "SharedBalances");

To this:

        migrationBuilder.DropForeignKey(
            name: "FK_SharedBalances_Users_OwnerUserId",
            table: "Balances");

        migrationBuilder.DropPrimaryKey(
            name: "PK_SharedBalances",
            table: "Balances");

Now your migration will work. This is how it happened:

2
votes

The question has a 2 part answer:

TL;DR.: Let the EF do it for you.

First: Let the EF do the relations

In short, the identity property is what the DB uses to manage a dedicated ID column, so it does not depend on anything outside itself to identify each row, therefore the bankAccount class needs it's own Id field to have the identity property; now, if you try to tell the EF manually how to do the relations like you do with the line

    HasOne(b => b.Customer).WithOne(c => c.BankAccount).HasForeignKey<BankAccount>(f => f.Id);

what you do is to override the inside logic of the EF itself and in this case you are telling the EF that the bank account Id is the field that references the costumer, which is not the case, and so EF tries to drop the IDENTITY from the id field in the bank account model, but that is only because you told it that the bank accound Id should be the same as the Customer ID. I think I understand what you are trying to say with the one to one relation, but what happens when the customer tries to open another bank account? D: Instead, you should keep the int CustomerId field and delete the Customer Customer one, not only because it's cleaner but also because the EF will recognize the relation between the two as long as there is a class called Customer with a field Id.

   public int CustomerId { get; set; } //keep

   public Customer Customer { get; set; } //delete

So then, the customer gets to open as many accounts with the bank as they please, the table does not suffer, and the EF knows what to do.

It will automatically generate the appropriate foreign key and add it to the migration file like so:

    migrationBuilder.AddForeignKey(
            name: "FK_BankAccount_Customer_CustomerId",
            table: "BankAccount",
            column: "CustomerId",
            principalTable: "Customer",
            principalColumn: "Id",
            onDelete: ReferentialAction.Cascade);

That will reflect a one to many relation, which is fine and dandy.

Now, to answer the question, if you still want to drop the property:

Second: Force the EF to drop the Identity property

(but it will not solve the foreign key problem of the original question).

First, just to recap: to change the identity property, the DB manager (mysql, sqlserver, etc.) will always ask you to drop and recreate the table because that field is the heart of the table. So you need to trick the EF to do a workaround for you.

  1. Add a second Id element to the model class with an obvious name like duplicateId.

    public class BankAccount
    {
        public int Id { get; set; }
        public int duplicateId { get; set; }
        ...
    }
    
  2. THIS IS THE TRICK ;)
    In the class there you implemented the DbContext interface, add the following method where you tell the ef which field you want the primary key to be:

    protected override void OnModelCreating(ModelBuilder modelBuilder)
    {
        modelBuilder.Entity<BankAccount>().HasKey(x => new { x.duplicateId });
    }
    
  3. Add a new Migration with $ dotnet ef migrations add ModelIdChange1, since this changes the primary key of the table and the migration Up method should look lile this:

    protected override void Up(MigrationBuilder migrationBuilder)
    {
        migrationBuilder.DropPrimaryKey(
            name: "PK_BankAccount",
            table: "BankAccountTableName");
    
        migrationBuilder.AddColumn<int>(
            name: "duplicateId",
            table: "BankAccountTableName",
            type: "int",
            nullable: false,
            defaultValue: 0)
            .Annotation("SqlServer:Identity", "1, 1");
    
        migrationBuilder.AddPrimaryKey(
            name: "PK_BankAccount",
            table: "BankAccountTableName",
            column: "duplicateId");
    ...
    }
    
  4. Then do the database update with $ dotnet ef database update (these commands may vary, use whatever syntax you already used before).

  5. (OPTIONAL) If you are need to preserve the original IDs, check that they got preserved or simply do a dirty update on the table to copy the data from the Id field into the duplicateId.

  6. Now the original Id field is free to be deleted or updated, so go ahead and just delete de original field from the model:

    public class BankAccount
    {
        public int duplicateId { get; set; }
        ...
    }
    

If you are still trying to force the original command that links the BankAccount Id with the Customer's Id, it should work if you run the command at this step, but please don't.

  1. And add a new Migration with $ dotnet ef migrations add ModelIdChange2, and then do the database update with $ dotnet ef database update, which deletes the original Id column and leaves duplicateId as the primary key.

  2. Now, the model looks almost the Original one but with a new identity column, you can leave it like that or just rename the field back from duplicateId to Id in the BankAccount class like this:

    public class BankAccount
    {
        public int Id { get; set; }
        ...
    }
    

    and do $ dotnet ef migrations add ModelIdChange3, and then do the database update with $ dotnet ef database update.

1
votes

I ran into the same problem ( In my case I didn't have any data in the tables ), and I solved it in this way ( It's not the proper way, but it worked for me ):

  1. I've deleted manually migrations from the EFCore project. I removed those lines which have been added from the file _ContextModelSnapshot as well. ( I had one migration which has been applied and one which has been created but it wasn't applied, as I was getting an error - Change the IDENTITY property of a column, the column needs to be dropped and recreated )
  2. I've deleted manually the tables which have been created in the database ( by the first migration)
  3. I've deleted the row in the table _EFMigrationHistory, that one which related to the Migration I wanted to remove.
  4. Re-run VS
  5. Add-Migration NewOneCleanMigration
  6. Update-Database
1
votes
  1. Table has no important data
  • Rename the table entity.ToTable("BankAccount2")
  • Add Run migration Add-Migration BankAccountTempChanges
  • Update the database Update-Database
  • Rename back the table entity.ToTable("BankAccount")
  • Add Run migration Add-Migration BankAccountOk
  • Update the database again Update-Database
  1. Table has data not to be lost
  • Apply solution from @Hani answer
0
votes

please follow this step:

1-please do all change of identity column in sql server(not in your code first entity framework)

2-comment identity column changes in the migration (.cs file)

3-update-database

enjoy that

0
votes

For those who are lazy like me: You want to change the datatype of a primary key column "Id" from int to Guid in a table called "Translations" as my case was.

  1. Your generated migration in that case is
migrationBuilder.AlterColumn<Guid>(
     name: "Id",
     table: "Translations",
     type: "uniqueidentifier",
     nullable: false,
     oldClrType: typeof(int),
     oldType: "int")
     OldAnnotation("SqlServer:Identity", "1, 1");

You can delete or comment that out

  1. From the update-database error System.InvalidOperationException: To change the IDENTITY property of a column, the column needs to be dropped and recreated.
  2. We also know that we cannot drop the column without dropping the primary key constraint first. Our new migration becomes
migrationBuilder.DropPrimaryKey(
    name: "PK_Translations",
    table: "Translations");
  migrationBuilder.DropColumn(
    name: "Id",
    table: "Translations");
  migrationBuilder.AddColumn<Guid>(
    name: "Id",
    table: "Translations",
    type: "uniqueidentifier",
    nullable: false);

Remember to do the opposite in the Down override method in case you may want to reverse the migration

0
votes

I had a similar problem where I was changing the relational navigation component of a table's configuration from WithMany to WithRequiredDependent. Entity framework wanted to drop the index and recreate the column, even though nothing in the database should have changed.

To fix this, I rescaffolded the latest migration which allowed entity to absorb the change without any new migration being created. You can rescaffold the latest migration by reverting the migration from the target database, and re-running the Add-Migration script for the latest migration with the exact same migration name.