Concurrency

Entity Framework Core: SQL Server Concurrency Checks

Over the last couple of weeks, I have covered concurrency checks in SQLite and Postgres. This week I’m expanding my Entity Framework sample repo to include a SQL Server based example project in order to demo concurrency checks. for SQL Server. The repo with the SQL Server project added, but before the concurrency checks can be found here.

Note that all the changes in this post will be made under the SqlServer folder.

Context Changes and Data Migration

SQL Server has a built-in column type of rowversion which is automatically generated with a unique binary number when a row is inserted or updated. In order to utilize the rowversion column type via Entity Framework, we need to add a new column to our model with the Timestamp attribute like the following from the sample code in the Contact model.

[Timestamp] 
public byte[] Timestamp { get; set; }

Next, from a command prompt in the same directory as your project file using the following .NET CLI command to generate a migration for the above change.

dotnet ef migrations add ContactTimestamp --context ContactsDbContext

Then, use the following command to apply the migration to your database.

dotnet ef database update --context ContactsDbContext

Testing it out

For a quick test add a ConcurrencyTest razor page under the Contacts directory. This function is going to ensure a specific contact exists, then pull the contact from two different DBContexts, make a mutation on the resulting contact objects, then attempt to save. The first save will work and the second should fail. Please note that this function isn’t an example of how things should be done just a quick and dirty way to prove that the concurrency check is happening.

public void OnGet()
{
    var context1 = new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                                         .UseSqlServer("yourConnectionString")
                                         .Options);
    var contactFromContext1 = context1.Contacts
                              .FirstOrDefault(c => c.Name == "Test");

    if (contactFromContext1 == null)
    {
        contactFromContext1 = new Contact
                              {
                                  Name = "Test"
                              };

        context1.Add(contactFromContext1);
        context1.SaveChanges();
    }

    var context2 = new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                                         .UseSqlServer("yourConnectionString")
                                         .Options);
    var contactFromContext2 = context2.Contacts
                              .FirstOrDefault(c => c.Name == "Test");

    contactFromContext1.Address = DateTime.Now.ToString();
    contactFromContext2.Address = DateTime.UtcNow.ToString();

    context1.SaveChanges();
    context2.SaveChanges();
}

Run the application and hit the ConcurrenctTest route which is https://localhost:44324/Contacts/ConcurrencyTest for my test. The following is the resulting exception.

An unhandled exception occurred while processing the request.

DbUpdateConcurrencyException: Database operation expected to affect 1 row(s) but actually affected 0 row(s). Data may have been modified or deleted since entities were loaded. See http://go.microsoft.com/fwlink/?LinkId=527962 for information on understanding and handling optimistic concurrency exceptions.

Wrapping Up

Of the three providers, Postgres seems to have the simplest implementation, not that concurrency is going to be how you choose your database provider. I don’t plan on adding any more providers at the moment, but if you are using with that hasn’t be covered leave a comment and I will look at adding another example.

The code in its final state can be found here.

Entity Framework Core: SQL Server Concurrency Checks Read More »

Entity Framework Core: Postgres Concurrency Checks

Last week’s post on SQLite Concurrency Checks used this repo which contains examples of using Entity Framework Core with SQLite and Postgres. This post is going to tackle concurrency checks using Postgres to keep the projects in the repo with the same level of functionality. You can grab the sample code before any change here. This whole post will only be touch files found in the Postgres folder/project.

Context Changes and Data Migration

Unlike SQLite, Postgres has better built-in support for concurrency checks. If you read the official docs on Optimistic Concurrency and Concurrency Tokens you will find that all tables have an implicit/hidden system column call xmin which holds the ID of the latest updating transaction which means it gets changed automatically every time a row is changed.

The Postgres Entity Framework Core provide contains an extension that makes it very simple to use the xmin column as a concurrency token. In the ContactDbContext add the following to the OnModelCreating function to enable concurrency checking on the specified entity, in this case, a Contact.

modelBuilder.Entity<Contact>().ForNpgsqlUseXminAsConcurrencyToken();

Next, from a command prompt in the same directory as your project file using the following .NET CLI command to generate a migration for the above change. This migration is a bit strange since the column technically already exists, but the migration seemed to be needed.

dotnet ef migrations add Concurrency --context ContactsDbContext

Then, use the following command to apply the migration to your database.

dotnet ef database update --context ContactsDbContext

Testing it out

For a quick and dirty test, I added a ConcurrencyTest razor page under the Contacts directory. This function is going to ensure a specific contact exists, then pull the contact from two different DBContexts, make a mutation on the resulting contact objects, then attempt to save. The first save will work and the second should fail. Please note that this function isn’t an example of how things should be done just a quick and dirty way to prove that the concurrency check is happening.

public void OnGet()
{
    var context1 = 
        new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                              .UseNpgsql("yourConnectionString")
                              .Options);
    var contactFromContext1 = context1.Contacts
                                      .FirstOrDefault(c => c.Name == "Test");

    if (contactFromContext1 == null)
    {
        contactFromContext1 = new Contact
                              {
                                  Name = "Test"
                              };

        context1.Add(contactFromContext1);
        context1.SaveChanges();
    }

    var context2 = 
        new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                              .UseNpgsql("yourConnectionString")
                              .Options);
    var contactFromContext2 = context2.Contacts
                                      .FirstOrDefault(c => c.Name == "Test");

    contactFromContext1.Address = DateTime.Now.ToString();
    contactFromContext2.Address = DateTime.UtcNow.ToString();

    context1.SaveChanges();
    context2.SaveChanges();
}

Run the application and hit the ConcurrenctTest route which is https://localhost:44324/Contacts/ConcurrencyTest for my test. The following is the resulting exception.

An unhandled exception occurred while processing the request.

DbUpdateConcurrencyException: Database operation expected to affect 1 row(s) but actually affected 0 row(s). Data may have been modified or deleted since entities were loaded. See http://go.microsoft.com/fwlink/?LinkId=527962 for information on understanding and handling optimistic concurrency exceptions.

Wrapping Up

This process was much simpler using Postgres than SQLite. Not that the SQLite version was hard just not as simple of a path.

The code in its final state can be found here.

Entity Framework Core: Postgres Concurrency Checks Read More »

Entity Framework Core: SQLite Concurrency Checks

Most of the work I have done with SQLite has been on single-user systems, but recently I had to work on a SQLite project that was going to have a handful of concurrent users and a subset of the user activities would need to deal with concurrency issues. In the past, in a situation like this, I have been using SQL Server and use the rowversion or timestamp column type which places a unique value on the row on any updates or inserts.

There is a page in the official docs on Concurrency Tokens, but for me, it wasn’t super helpful. Thankfully after some searching, I came across the GitHub issue In ASP.Net Core 2.x with Entity Framework Core, Concurrency Control not working with SQLite which had a solid sample as one of the replies. This post is going to walk through an example implementation of that sample. The starting point of the code can be found in this GitHub repo.

Sample Background

The sample project being used is a simple web application to manage a contact list. The repo contains an implementation using Postgres and one using Sqlite. This whole post will only be touch files found in the Sqlite folder/project.

Model Changes and Data Migration

SQLite doesn’t have the concept of a timestamp column, but this solution is going to emulate one. To do this we are going to change the Contact model found in the Models folder. We are going to add a Timestamp property with a Timestamp data annotation. The following is the full model class with the new property at the bottom.

public class Contact
{
    public int Id { get; set; }
    public string Name { get; set; }
    public string Address { get; set; }
    public string City { get; set; }
    public string Subregion { get; set; }
    public string PostalCode { get; set; }
    public string Phone { get; set; }
    public string Email { get; set; }
    [Timestamp] public byte[] Timestamp { get; set; }
}

Next, let’s create a new migration with the change to the model. I’m using the .NET CLI so from a command prompt in the project directory run the following command.

dotnet ef migrations add ContactTimestamp --context ContactsDbContext

In the Migrations directory, open newly created migration. It should be named something like *_ContactTimestamp.cs.  In the Up function, we are going to add a couple of triggers to the new Timestamp column. These triggers are going to assign a random blob to the timestamp column when a row is inserted or updated which is how we are simulating the function of SQL Server’s Timestamp data type. The following is the full Up function with the added triggers.

protected override void Up(MigrationBuilder migrationBuilder)
{
    migrationBuilder.AddColumn<byte[]>(
        name: "Timestamp",
        table: "Contacts",
        rowVersion: true,
        nullable: true);

    migrationBuilder.Sql(
        @"
        CREATE TRIGGER SetContactTimestampOnUpdate
        AFTER UPDATE ON Contacts
        BEGIN
            UPDATE Contacts
            SET Timestamp = randomblob(8)
            WHERE rowid = NEW.rowid;
        END
    ");

    migrationBuilder.Sql(
        @"
        CREATE TRIGGER SetContactTimestampOnInsert
        AFTER INSERT ON Contacts
        BEGIN
            UPDATE Contacts
            SET Timestamp = randomblob(8)
            WHERE rowid = NEW.rowid;
        END
    ");
}

To apply the migration to the database you can use the following command.

dotnet ef database update --context ContactsDbContext

Testing it out

Now for a quick and dirty test, we are going to add a ConcurrencyTest function to the existing ContactsController. This function is going to ensure a specific contact exists, then pull the contact from two different DBContexts, make a mutation on the resulting contact objects, then attempt to save. The first save will work and the second should fail. Please note that this function isn’t an example of how things should be done just a quick and dirty way to prove that the concurrency check is happening.

[Route("ConcurrencyTest")]
public void ConcurrencyTest()
{
    var context1 = new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                                         .UseSqlite("Data Source=Database.db").Options);
    var context2 = new ContactsDbContext(new DbContextOptionsBuilder<ContactsDbContext>()
                                         .UseSqlite("Data Source=Database.db").Options);

    var contactFromContext1 = context1.Contacts.FirstOrDefault(c => c.Name == "Test");

    if (contactFromContext1 == null)
    {
        contactFromContext1 = new Contact
        {
            Name = "Test"
        };

        context1.Add(contactFromContext1);
        context1.SaveChanges();
    }

    var contactFromContext2 = context2.Contacts.FirstOrDefault(c => c.Name == "Test");

    contactFromContext1.Address = DateTime.Now.ToString();
    contactFromContext2.Address = DateTime.UtcNow.ToString();

    context1.SaveChanges();
    context2.SaveChanges();
}

Run the application and hit the ConcurrenctTest route which is http://localhost:1842/ConcurrencyTest for my test. The following is the resulting exception.

An unhandled exception occurred while processing the request.

DbUpdateConcurrencyException: Database operation expected to affect 1 row(s) but actually affected 0 row(s). Data may have been modified or deleted since entities were loaded. See http://go.microsoft.com/fwlink/?LinkId=527962 for information on understanding and handling optimistic concurrency exceptions.

Wrapping Up

While the information wasn’t the easiest in the world to locate, as you can see Entity Framework Core using SQLite has good support for concurrency control. The above is just one option for its implementation. I hope this saves you all so time.

The code in its final state can be found here.

Entity Framework Core: SQLite Concurrency Checks Read More »