EntityFramework migrations become useless after switching to new Context.
DbMigrator is using list of Pending Migrations from first database instance, which makes means no migrations are applied to other databases, which then leads to errors during Seed();
C# .NET 4.5 MVC project with EF 6
MS SQL Server 2014, multiple instances of same database model.
CodeFirst approach with migrations.
DbContext initializer is set to null.
On Application Start we have custom Db Initialization to create and update databases. CreateDatabaseIfNotExists is working as intended, new databases have all migrations applied. However both MigrateDatabaseToLatestVersion initializer and our custom one are failing to update databases other than first one on list.
foreach (var connectionString in connectionStrings)
{
using (var context = new ApplicationDbContext(connectionString))
{
//Create database
var created = context.Database.CreateIfNotExists();
var conf = new Workshop.Migrations.Configuration();
var migrator = new DbMigrator(conf);
migrator.Update();
//initial values
conf.RunSeed(context);
}
}
context.Database.CreateIfNotExists(); works correctly.
migrator.GetLocalMigrations() is always returning correct values.
migrator.GetPendingMigrations() after first database is returning
empty list.
migrator.GetDatabaseMigrations() is mirror of pending migrations,
after first database it contains full list event for empty databases.
Fetching data (context.xxx.ToList()) from Db instance confirms connection is up and working, and links to correct instance.
Forcing update to most recent migration with migrator.Update("migration_name"); changes nothing. From what I gather by reading EF source code, it checks pending migration list on its own, which gives it faulty results.
There seems to be some caching going in under the hood, but it eludes me how to reset it.
Is there a way to perform migrations on multiple databases or is it yet another "bug by design" in EF?
Edit:
Real problem is DbMigrator creating new Context for its own use. It does it via default parameterless constructor, which in my case had fallback to default (first) connection string in web.Config.
I do not see good solution for this problem but primitive workaround in my case is to temporarily edit default connection string:
var originalConStr = WebConfigurationManager.ConnectionStrings["ApplicationDbContext"].ConnectionString;
var setting = WebConfigurationManager.ConnectionStrings["ApplicationDbContext"];
var fi = typeof(ConfigurationElement).GetField("_bReadOnly", BindingFlags.Instance | BindingFlags.NonPublic);
//disable readonly flag on field
fi.SetValue(setting, false);
setting.ConnectionString = temporaryConnectionString; //now it works
//DO STUFF
setting.ConnectionString = originalConStr; //revert changes
Cheat from: How do I set a connection string config programatically in .net?
I still hope someone will find real solution so for now I will refrain with self-answer.
You need to correctly set DbMigrationsConfiguration.TargetDatabase property, otherwise the migrator will use the default connection info.
So in theory you can do something like this
conf.TargetDatabase = new System.Data.Entity.Infrastructure.DbConnectionInfo(...);
Unfortunately the only 2 public constructors of the DbConnectionInfo are
public DbConnectionInfo(string connectionName)
connectionName: The name of the connection string in the application configuration.
and
public DbConnectionInfo(string connectionString, string providerInvariantName)
connectionString: The connection string to use for the connection.
providerInvariantName: The name of the provider to use for the connection. Use 'System.Data.SqlClient' for SQL Server.
I see you have the connection string, but have no idea how you can get the providerInvariantName.
UPDATE: I didn't find a good "official" way of taking the needed information, so I've ended using a hack with accessing internal members via reflection, but still IMO it's a quite more safer than what you have used:
var internalContext = context.GetType().GetProperty("InternalContext", BindingFlags.Instance | BindingFlags.NonPublic).GetValue(context);
var providerName = (string)internalContext.GetType().GetProperty("ProviderName").GetValue(internalContext);
var conf = new Workshop.Migrations.Configuration();
conf.TargetDatabase = new System.Data.Entity.Infrastructure.DbConnectionInfo(connectionString, providerName);
Related
I'm trying to test the newly supported transactions in Mongo DB with a simple example I wrote.
I'm using Mongo DB version 4.0.5 with driver version 2.8.1.
It's only a primary instance with no shards/replicas.
I must be missing something basic in the following code.
I create a Mongo client, session & database, then start a transaction, add a document and abort the transaction. After this code, I expect nothing to change in the database, but the document is added. When debugging I can also see the document right after the InsertOne() by using Robo 3T (Mongo client GUI).
Any idea what am I missing?
var client = new MongoClient("mongodb://localhost:27017");
var session = client.StartSession();
var database = session.Client.GetDatabase("myDatabase", new MongoDatabaseSettings
{
GuidRepresentation = GuidRepresentation.Standard,
ReadPreference = ReadPreference.Primary,
WriteConcern = new WriteConcern(1,
new MongoDB.Driver.Optional<TimeSpan?>(TimeSpan.FromSeconds(30))),
});
var entities = database.GetCollection<MyEntity>("test");
session.StartTransaction();
// After this line I can already see the document in the db collection using Mongo client GUI (Robo 3T), although I expect not to see it until committing
entities.InsertOne(new MyEntity { Name = "Entity" });
// This does not have any effect
session.AbortTransaction();
Edit:
It's possible to run MongoDB as a 1-node replica set, although I'm not sure what's the difference between a standalone and a 1-node replica set.
See my post below.
In any case, to use the started transaction the insertion code must receive the session as a parameter:
entities.InsertOne(session, new MyEntity { Name = "Entity" });
With these 2 change now the transaction works.
This is inherently a property of MongoDB itself. (More here and here)
Transactions are only available in a replica set setup
Why isnt it available for standalone instances?
With subdocuments and arrays, document databases (MongoDB) allow related data to be unified hierarchically inside a single data structure. The document can be updated with an atomic operation, giving it the same data integrity guarantees as a multi-table transaction in a relational database.
I found a solution, although not sure what the consequences are, maybe someone can point it out:
It seems it's possible to use Mongo DB as a 1-node replica set (instead of a standalone) by simply adding the following in the mongod.cfg file:
replication:
replSetName: rs1
Also, thanks to the following link the code should use the correct overload of InsertOne() which receives the session as the first parameter (see the edit on the original post):
multiple document transaction not working in c# using mongodb 4.08 community server
I'm kind of new with databases and SQL and I'm struggling trying to understand how SQL Change Tracking and Microsoft Sync Framework work together.
I couldn't find some clear examples about how to sync databases with Microsoft Sync Framework but hopefully I found this site, modified the code and got syncing working on my two databases, here is the code I got:
// Server connection
using (SqlConnection serverConn = new SqlConnection(serverConnectionString))
{
if (serverConn.State == ConnectionState.Closed)
serverConn.Open();
// Client connection
using (SqlConnection clientConn = new SqlConnection(clientConnectionString))
{
if (clientConn.State == ConnectionState.Closed)
clientConn.Open();
const string scopeName = "DifferentPKScope";
// Provision Server
var serverProvision = new SqlSyncScopeProvisioning(serverConn);
if (!serverProvision.ScopeExists(scopeName))
{
var serverScopeDesc = new DbSyncScopeDescription(scopeName);
var serverTableDesc = SqlSyncDescriptionBuilder.GetDescriptionForTable(table, serverConn);
// Add the table to the descriptor
serverScopeDesc.Tables.Add(serverTableDesc);
serverProvision.PopulateFromScopeDescription(serverScopeDesc);
serverProvision.Apply();
}
// Provision Client
var clientProvision = new SqlSyncScopeProvisioning(clientConn);
if (!clientProvision.ScopeExists(scopeName))
{
var clientScopeDesc = new DbSyncScopeDescription(scopeName);
var clientTableDesc = SqlSyncDescriptionBuilder.GetDescriptionForTable(table, clientConn);
// Add the table to the descriptor
clientScopeDesc.Tables.Add(clientTableDesc);
clientProvision.PopulateFromScopeDescription(clientScopeDesc);
clientProvision.SetCreateTrackingTableDefault(DbSyncCreationOption.CreateOrUseExisting);
clientProvision.Apply();
}
// Create the sync orchestrator
var syncOrchestrator = new SyncOrchestrator();
// Setup providers
var localProvider = new SqlSyncProvider(scopeName, clientConn);
var remoteProvider = new SqlSyncProvider(scopeName, serverConn);
syncOrchestrator.LocalProvider = localProvider;
syncOrchestrator.RemoteProvider = remoteProvider;
// Set the direction of sync session
syncOrchestrator.Direction = direction;
// Execute the synchronization process
return syncOrchestrator.Synchronize();
}
}
So on this way any changes are synchronized between my two databases. But I wanted a way for my C# app to automatically synchronize both databases when something changes so I found something called Change Tracking here. I downloaded the example code that provides a SynchronizationHelper that also creates tables in my databases called "{TableName}_tracking". This is another table that tracks the changes and indeed it does, whenever I change something in my database the _tracking is updated with the elements I changed, added or removed. Change Tracking doesn't automatically synchronize my databases, it just keeps track of the changes in them, what's the purpose of this?
With the first code, synchronization works but no _tracking table is created, does it just synchronize everything in the table no matter what changed? If that's the case, for big databases I should be using Change Tracking?
Maybe this is something trivial but I have been googling and testing a lot of code but I can't find a clear answer.
When you install Sync Framework, it comes with a help file that includes several walkthroughs of synchronizing databases. the first link you referred to and the second uses the same sync provider and they both have tracking tables. Sync Framework supports using the built-in SQL Change Tracking feature or using a custom-one that Sync Framework creates by itself (the _tracking).
Sync Framework sits outside of your database and you need to invoke it in order to fire the synchronization. Change Tracking is what it says it is- tracking changes.
if you want your databases to do the sync, you might want to check SQL Replication instead.
I'm trying to create a class to perform work on the database and have the need (or preference) to use a combination of DbContext and good old fashioned ADO. Why, well EF is great for simplifying a great deal of code but ADO still has many uses for more complex methods that EF cannot yet handle.
This link on MSDN states that I can pass an existing SqlConnection to my context as follows:
using (var conn = new SqlConnection("..."))
{
conn.Open();
using (var context = new SampleContext(conn, contextOwnsConnection: false))
{
// Do Something
}
}
Now I'm using Database-First so this constructor doesn't appear as standard. I therefore created a new Partial Class file and created the appropriate constructor as follows:
public partial class MyEntities : DbContext
{
public MyEntities(System.Data.Common.DbConnection conn, bool contextOwnsConnection = false)
: base(existingConnection: conn, contextOwnsConnection: contextOwnsConnection)
{
}
}
However, when I run the code the moment it hits a call to the new DbContext constructor, I get the following UnintentionalCodeFirstException() error thrown by OnModelCreating in my EDMX file:
"Code generated using the T4 templates for Database First and Model First development may not work correctly if used in Code First mode. To continue using Database First or Model First ensure that the Entity Framework connection string is specified in the config file of executing application. To use these classes, that were generated from Database First or Model First, with Code First add any additional configuration using attributes or the DbModelBuilder API and then remove the code that throws this exception."
Am I missing something obvious here, or can it just not be done with Database-First?
Clearly, I could just use two connections, one for my SqlConnection object, and another for my DbContext object but if I can, naturally I'd prefer to use a single connection if possible.
Any and all help greatly appreciated. For full disclosure, I'm using SQL-Server 2012, .NET 4.5.1, C# and EF6.0.2.
Connection strings used by the designer are not regular connection strings. Rather they are EntityConnection strings. The difference is that entity connection strings contain additional information about where to find metadata describing the model which is in form of the edmx at design time - read more here. Code First uses just regular connection strings since it builds the model on the fly based on the code. So, the UnintentionalCodeFirstException is preventing the user from using CodeFirst functionality with edmx models because the model is specified in the edmx and not in the code and if it was allowed you would effectively end up using two different models (one from edmx and one built from the code) which very likely won't be in sync which would result in weird behavior or even could lead to data corruption and crashes/exceptions.
Since the EntityConnection is derived from DbConnection and just wraps regular connection you can use it in places where you would use the provider connection. Alternatively you can access the wrapped provider connection using the StoreConnection provider on the EntityConnection.
I know this is an old thread, but later versions of Entity Framework actually can handle a shared connection, so I offer this as an alternative answer.
You can initialize an instance of the entity container with a shared connection. Use the EntityConnection(MetadataWorkspace workspace, DbConnection connection, bool entityConnectionOwnsStoreConnection) overload and specify false for the entityConnectionOwnsStoreConnection parameter. Then pass it into your context constructor as the existing connection. The EntityConnection will then prevent the connection from being automatically closed and disposed with the context.
Example:
using (var conn = new SqlConnection("..."))
{
conn.Open();
// Execute some ADO queries.
var md = new MetadataWorkspace(new[]{"res://*/SampleModel.csdl","res://*/SampleModel.ssdl","res://*/SampleModel.msl"}, new[]{System.Reflection.Assembly.GetExecutingAssembly()});
// Create the EntityConnection so the existing connection is not disposed.
var ec = new EntityConnection(md, conn, false);
using (var context = new SampleContext(conn, contextOwnsConnection: false))
{
// Do something using the entity context.
}
// Entity context is disposed but connection remains open.
// Do more ADO stuff.
}
Yes, it was a pain to figure out this stuff by examining System.Data and Entity Framework source code.
This pattern may be used within a TransactionScope to prevent escalation to a distributed transaction by virtue of using the same database connection.
I am using the devart component dotconnect for postgresql. I have created the site using linq to entities, however, I would like each user to have a seperate database. This means that I need to change the connection string for each person that has logged in. I understand the main part of how to generate a new connection string etc, however, when i pass that as a paramater to the object context object it comes back with the error
"user id keyword not supported, "
if i create a class that generates an entity connection the error message changes to:
"Unable to load the specified metadata resource."
Cannot work out what I have done wrong in these instances.
ok so, as usual, when i posted this question, about 3 minutes later i found the problem. The entity connection string, for general purposes should have a cool little
res://*/
this makes the metadata work. This solves the problem of metadata resource and this works. So to help others who may, like me, have spent development time doing this, i created a class, with a method like so.
public static string getConnString(string database)
{
string connectionstring = "User Id=USER ID HERE;Password=PASSWORD HERE;Host=server;Database="+database+";Persist Security Info=True;Schema=public";
EntityConnectionStringBuilder newconnstring = new EntityConnectionStringBuilder();
newconnstring.Metadata = #"res://*/";
newconnstring.Provider = "Devart.Data.PostgreSql";
newconnstring.ProviderConnectionString = connectionstring;
return newconnstring.ToString();
}
then create a constructor like so
dataEntities data = new dataEntities(databaseConnection.getConnString(INSERTDBNAMEHERE);
Then we can reference that in the same way as a usual linq statement. Simples!!
I use EntityFramework in a project to connect to a Mysql database. The provider is Devart Dot.connect.
This application needs to connect to a database using connexion parameters given by the user at runtime. This includes of course the Mysql Database name.
I used the EntityConnectionStringBuiler and EntityConnection classes to build-up and store the custom connexion parameters.
The problem is that even with such given parameters, the application always connect to the database named when designing the EntityModel using the visual studio wizard.
What is very strange is that when debuging and checking the status of the ObjectContext, the custom connexion parameters are correctly used...
It makes me mad !!!!!
Any clue ?
After spending one day on this issue, I finally came to understand that the problem was coming from the model.edmx file.
In this file, you have one line per EntitySet.
On each EntitySet element there is an attribute called schema. In case of SQL Server this attribute is set to the related table schema :
EntitySet Name="annee_civile"
EntityType="openemisModel.Store.annee_civile"
store:Type="Tables" Schema="mydatabase" />
If you provide the name of the Schema when constructiong you own EntityConnection, it seem that there is a conflict and that finally, the Schema defined in the edmx file will be used even if you specified another one in the connection parameters.
The solution is simply to remove the name of the schema in the edmx file.
THIS WORKS FOR MYSQL, probably not when connecting to a SQL server.
EntitySet Name="annee_civile"
EntityType="openemisModel.Store.annee_civile"
store:Type="Tables" Schema="" />
The EntityConnectionStringBuilder :
string providedString = "User Id=xxxx;Password=xxx;Host=xxxx;Database=anydatabasename";
EntityConnectionStringBuilder entityConnBuilder = new EntityConnectionStringBuilder();
entityConnBuilder.Provider = "Devart.Data.MySql";
entityConnBuilder.Metadata = #"res:///OpenEmisModel.csdl|res:///OpenEmisModel.ssdl|res://*/OpenEmisModel.msl";
entityConnBuilder.ProviderConnectionString = providedString;
The EntityConnection and the object context using it:
EntityConnection entityConnexionEmis = new EntityConnection(entityConnBuilder.ConnectionString);
objectcontextEntities testingContext = new objectcontextEntities(entityConnexionEmis);
The software is now able to connect to any database name.
Hope this helps.
Reference the dll Devart.Data.MySql.Entity.EF6.dll in the project.
Somewhere when your application is starting up and before database operations take place, add the following:
var config = MySqlEntityProviderConfig.Instance;
config.Workarounds.IgnoreSchemaName = true;
You will need to reference:
using Devart.Data.MySql.Entity.Configuration;