Using Non Serializable objects in Activiti BPMN - serialization

I want to use Activiti BPMN process for some database update task. My process is as follows.
Start Event-> Service Task 1 -> Service Task 2 -> Service Task 3 -> End Event
In the service implementation class of Service task 1 : I created a java.sql.Connection for MySQL database. I need to pass the same Connection object to the Service Task 2 and Service Task 3. Basically those two classes will do some insertions for the database using the same Connection object.
I tried as follows (dbConn is the the Class which contains java.sql.Connection type dbConnection)
execution.setVariable("DBConn",dbConn);
But it gives an exception since the connection object is not serializable.
"org.activiti.engine.ActivitiException: Couldn't serialize value"
So what is the best way to pass such non serializable variables between service tasks of a process? Or is there any way to define such common objects to multiple Service Tasks in one place and use them within service Tasks ( Something like global variables for the process)

You can use Thradlocal in Java to pass connection object to different service tasks.
For example use a Base class like below and extend each service task from that. Then you can set the dbConnection and use whenever required by using get method.
public class BaseServiceTask
{
public static final ThreadLocal<Connection> localConnectionContext = new ThreadLocal<Connection>();
public static void initDBConnector(Connection dbConn)
{
localConnectionContext.set(dbConn);
}
public static Connection getDBConnector()
{
return localConnectionContext.get();
}
}
Notice :
This approach assumes all service tasks are executed in the same thread, which is the case for this particular question, but once you include some user task / timer (or any asynchronous logic) this is not a viable solution anymore !

First, you should be aware that there is absolutely no way to serialize a connection instance once it got created according to this.
The reason is that a connection uses a network resource (such as a TCP/IP socket) which uses the network stack on the machine, and eventually the machine's hardware.
Which leaves you only this alternative:
Setup a bean that will store the connection instances for you, let's call it myConnectionRegistry, this bean should be scoped as singleton and injected in all your java delegates (Service task implementations)
In the first task, you create the connection and then register it into myConnectionRegistry with something like this connectionRegistry.register(conn, wfId) which would add the connection instance to a private map ....
In the subsequent tasks, you retrieve your task from that same bean using a method that fetches the connection object from the private map, and throwing an exception if no connection object was registered in the map
Have a boundary event that gets fired on that exception and do whatever is necessary to insure data integrity (the use case i described in my comment for instance)
In the last Service task, un-register your connection (you should also close it) in order to prevent memory leaks
Make sure to take into account the db pool ... etc while designing your solution!
Cheers!

Related

Best strategy for creating a child container (or isolated scope) with Microsoft.Extensions.DependencyInjection

In my AspNetCore application, I process messages that arrive from a queue. In order to process a message, I need to resolve some services. Some of those services have a dependency on ITenantId, which I bind using information from the received message. To solve this, the processing of a messages starts with the creation of a child container, which I then use to create an IServiceScope from which I resolve all the needed dependencies.
The messages can be processed in parallel, so the scopes must be isolated from each other.
I can see to ways of creating the child container, but I'm not sure which is best in terms of performance, memory chrurn etc:
Option A: Each time a message arrives, clone the IServiceCollection into a new ServiceCollection, and rebind ITenantId on the cloned instance.
Option B: When the program starts, create an immutable copy of the IServiceCollection (using ImmutableList<ServiceDescriptor> or ImmutableArray<ServiceDescriptor>). Each time a message arrives, replace ITenantId (resulting in a new instance of ImmutableList<ServiceDescriptor>) and call CreateScope() on the new immutable instance.
The thing I don't like about option A is that the whole collection of services needs to be cloned every time a message arrives. I'm not sure if the immutable collections in option B handles this in a smarter way?
Both options cause the creation of a new container instance for each incoming messages. Although this allows each message to run in a completely isolated bubble, this has severe implications on performance and memory use of the application. Creating container instances is expensive and resolving a registered instance for the first time (per container) causes generation of expression trees, compilation of delegates, and JIT compiling them. This can even cause memory leaks.
Besides, it also means that any registered singleton class, will have a lifetime that equals that of any scoped classes. State can't be shared any longer.
So instead, I propose Option 3:
Use only one Container instance and don't call BuildProvider more than once
Create a ITenantId implementation that allows setting the Id after instantiation
Register that implementation as Scoped
At the start of every new IServiceScope, resolve that implementation and set its id.
This might look as follows:
// Code
class TenentIdImpl : ITenantId
{
public Guid Id { get; set; } // settable
}
// Startup:
services.AddScoped<TenentIdImpl>();
services.AddScoped<ITenantId>(c => c.GetRequiredService<TenantIdImpl>());
// In message pipeline
using (var scope = provider.CreateScope())
{
var tenant = scope.ServiceProvider.GetRequiredService<TenantIdImpl>();
tenant.Id = messageEnvelope.TenantId;
var handler =
scope.ServiceProvider.GetRequiredService<IMessageHandler<TMessage>>();
handler.Handle(messageEnvelope.Message);
}
This particular model, where you store state inside your object graph, which I explain in my blog, is called the Closure Composition Model.

Lifespan of FluentData IDbContext

I am using FluentData ORM for .NET in my data access layer. However, this question may apply to most ORMs.
What is the proper lifespan of the IDbContext? Should a new database context be instantiated every time a new query to the database is executed? I have two scenarios below, which is the proper usage? Note that I am querying MS SQL as well as MySQL servers. So, I would like to know if either have any implications as far as reaching the max connections. And yes, I am calling the database on multiple threads. Are both scenarios thread safe?
Scenario A:
My repository class implements IDisposable and holds a private variable of type IDbContext. This variable gets instantiated at the constructor and disposed when the class is disposed. There is only one instance of the context in the repository.
Scenario B:
My repository class has a private method that returns a new instance of IDbContext. Whenever another method makes a call to the database, they call this method to get a new database context.
From the official documentation:
When should you dispose?
DbContext must be disposed if you have enabled UseTransaction or
UseSharedConnection.
DbCommand must be disposed if you have enabled UseMultiResult (or MultiResultSql).
StoredProcedureBuilder must be disposed if you have enabled UseMultiResult.
In all the other cases dispose will be handled automatically by FluentData. This means that a database connection is opened just before a query is executed and closed just after the execution has been completed.
http://fluentdata.codeplex.com/documentation#Dispose
Hope it helps someone.

NHibernate Session gets disposed off on each wcf service method call

I am using Wcf in Sharp Architecture. I have configured my project following northwind sample using WcfSessionStorage etc. I have a method in the wcf service that gets a list of business object using Repository<>.GetAll(). I am testing the service method using WcfTestClient. When the method is called the very first time, everything works fine. However on the subsequent call, I get the following exception on the Repository<>.GetAll() method
[System.ObjectDisposedException]
Session is closed!
Object name: 'ISession'
It seems like the NHibernate session gets disposed after each call. I have got around this problem by decorating my service with the following attribute
[ServiceBehavior( InstanceContextMode = InstanceContextMode.PerCall )]
public class WcfService : IWcfService
{
}
However this means, an instance of the service will be created on each call that in turn will create a new nhibernate session etc. In my scenario there is no need of creating a new service instance per call and I think its an expensive process and not the right solution. I would like to know what is the best practice in my scenario and how to get this thing work with creating a new service instace per call.
Thanks
Nabeel
The easiest way is to create a new instance every time and it's not an expensive process, because creating a new object in .NET is like 0.00000000000000001 second (I read that on Ayande's blog or somewhere).
I use Autofac DI in my projects and I usually make ISession as container scoped (one per request). And then every class that uses (directly or indirectly) ISession has to be container scoped or lower (factory scoped == every class usage get's a new instance). If a class that uses ISession is higer scoped (session scoped == singleton) you'll run into problems that you currently have.
If your service is singleton service:
At first run the service is created, this service uses ISession, which should be container scoped, and it is on the first run.
The next request to service (service is now created) has still a reference to created ISession (which was closed on previous end request), so now it's closed.
I don't recomend using the same ISession that you'll open/close (it's not recomended in the NHibernate documentation), just use container scoped (I do and I don't have any performance issues), or you should create ISession manually in every method in your service like:
using(ISession s = ISessionFactory.OpenSession())
using(ITransaction t = .....)
....
But that isn't nice at all...
Please take a look at my answer to my own similar question: WCF/S#arpArch: underlying ISession is closed after the first call within a request.
#dmonlord is right that the creation of additional session instances within the same request is very cheap in this case.

Detect when client connected to wcf service

From a little bit of reading around, it is my understanding that the only way to detect that a client has connected to my service is through writing my own code. I am using a Singleton service. I would like to display a message every time a client connects to my service that client x with ip xxx has connected. There is no built-in event that is generated? Am I correct?
No, I don't think there's any support in WCF for your requirement.
Not sure what you want to achieve with this, either. Your service class (in your case, just a single instance) really doesn't have any business putting up messages (on screen, I presume) - that really not it's job. The service class is used to handle a request and deliver a response - nothing more.
The ServiceHost class might be more of a candidate for this feature - but again, it's job really is to host the service, spin up the WCF runtime etc. - and it's really not a UI component, either.
What you could possibly do is this
have an Admin UI (a Winforms, console, or WPF app) running on your server alongside your service, providing an admin service to call
define a fast connection between the two services (using e.g. netNamedPipe binding which is perfect for intra-application messaging)
when your "real" service gets a call, the first thing it does is send out a message to the admin UI which can then pick up that message and handle it
That way, you could cleanly separate your real service and it's job (to provide that service) and the Admin UI stuff you want to do and build a cleanly separated system.
I have actually implemented my own connect, disconnect and ping service methods which I manually call from my client once the channel has been created. By using them as a kind of header section in all of my ServiceContract interface definitions (and their implementations, of course), they form an makeshift "base service definition" that only requires a bit of cut-n-paste.
The string-based parameters of connect and disconnect will be used to send client info to the server and return server info and (perhaps a unique connection id) to the client. In addition a set of timing reference points may make its way in also.
Note how SessionMode is required and the individual OperationContract properties IsInitiating and IsTerminating are explicitly specified for each method, the end result being what I would call a "single-session" service in that it defines connect and disconnect as the sole session bookends.
Note also that the ping command will be used as the target of a timer-based "heartbeat" call that tests the service connection state and defeats ALL connection timeouts without a single config file :-)
Note also that I haven't determined my fault-handling structure yet which may very well add a method or more and/or require other kinds of changes.
[ServiceContract( SessionMode = SessionMode.Required )]
public interface IRePropDalSvr {
[OperationContract( IsInitiating=true, IsTerminating=false )]
string connect (string pClientInfo);
[OperationContract( IsInitiating=false, IsTerminating=true, IsOneWay=true )]
void disconnect (string pClientInfo);
// ------------------------------------------------------------------------------------------
[OperationContract( IsInitiating=false, IsTerminating=false )]
string ping (string pInp);
// ------------------------------------------------------------------------------------------
// REST OF ServiceContract DEFINITION GOES HERE
One caveat: while I am currently using this code and its implemention in my service classes, I have not verified the code yet.

WCF data persistence between sessions

We are developing a WCF based system. In the process we are trying to lock some data from being modified by more than one users. So we decided to have a data structure that will contain the necessary information for the locking logic to execute (by for example storing the ID of the locked objects)
The problem we are having is persisting that data between sessions. Is there anyway we can avoid executing expensive database calls?
I am not sure how can we do that in WCF since it can only persist data (in memory) during an open session.
Static members of the service implementing class are shared between sessions & calls.
One option would be to use static members as Jimmy McNulty said. I have a WCF service that opens network connections based on a user-specified IP address. My service is configured for PerCall service instance mode. In each session, I check a static data structure to see if a network connection is already opened for the specified IP address. Here's an example.
[ServiceContract]
public interface IMyService
{
[OperationContract]
void Start(IPAddress address);
}
[ServiceBehavior(InstanceContextMode=InstanceContextMode.PerCall)]
public class MyService : IMyService
{
private static readonly List<IPAddress> _addresses = new List<IPAddress>();
public void Start(IPAddress address)
{
lock(((ICollection)_addresses).SyncRoot)
{
if (!_addresses.Contains(address)
{
// Open the connection here and then store the address.
_addresses.Add(address);
}
}
}
}
As configured, each call to Start() happens within its own service instance, and each instance has access to the static collection. Since each service instance operates within a separate thread, access to the collection must be synchonized.
As with all synchronization done in multithreaded programming, be sure to minimize the amount of time spent in the lock. In the example shown, once the first caller grabs the lock, all other callers must wait until the lock is released. This works in my situation, but may not work in yours.
Another option would be to use the Single service instance mode as opposed to the PerCall service instance mode.
[ServiceBehavior(InstanceContextMode=InstanceContextMode.Single)]
public class MyService : IMyService
{ ... }
From everything I've read, though, the PerCall seems more flexible.
You can follow this link for differences between the two.
And don't forget that the class that implements your service is just that - a class. It works like all C# classes do. You can add a static constructor, properties, event handlers, implement additional interfaces, etc.
Perhaps a caching framework like velocity help you out.
Create a second class and set its InstanceContextMode to single and move all the expensive methods there, then in your original class use that methods.