Is it possible to use some kind of 'critical section' so that it is safe to do something like the following within an action...
public ActionResult GenerateTasks()
{
string someDir = ....
if (!Directory.Exists(someDir))
{
Directory.CreateDirectory(someDir);
}
...
}
You can do this only by using a system-wide mutex. Process or app-domain locking primitives will fail to work under certain conditions (for instance when an application pool is recycled).
However, for the specific case here that's not necessary: Directory.CreateDirectory already does implement an existence check on its own, so that you shouldn't need to do anything in this regard.
I'm assuming by your question that the concurrent safety you're interested in is whether or not the directory is created between the Directory.Exists and the Directory.CreateDirectory on a different thread. (If you're concerned about Directory.CreateDirectory throwing an exception if the directory already exists, it won't.) If so, and this is the point in your code that will have the potential to do that, then you can simply use a lock object to make these set of operations safe across multiple threads:
private static object lockObject = new object();
public ActionResult GenerateTasks()
{
string someDir = ....
lock(lockObject)
{
if (!Directory.Exists(someDir))
{
Directory.CreateDirectory(someDir);
}
}
...
}
This does not however make any garauntees that the directory isn't being interacted with outside of your control, say, in another application process.
Related
I've encountered a dependency injection scenario which I cannot find a way through.
We currently have an Azure function.
We are using dependency injection via the FunctionsStartup attribute.
That all works fine, until I get asked to make it work for multiple environments.
The tester found it too onerous to deploy to 7 different environments, so I was asked to re-jig the function so that it runs (in a loop) for those environments.
That means 7 different IConfigurations and somehow having 7 separate compartmentalised IOC registrations of services.
I can't think of a way of doing that, without significantly re-structuring the way abstractions are being resolved. Even if you set up registrations in a loop and inject an IEnumerable of a service, when it goes to resolve a child dependency, it just pulls the last one registered, rather than the one which was meant to correlate with the current item being iterated.
So, something like this (using Autofac):
Registration
foreach (var configuration in configurations)
{
containerBuilder.Register<ICosmosDbService<AccountUsage>>(sp =>
{
var dBConfig = CosmosDBHelper.GetProjectDatabaseConfig(configuration.Value, Project.Jupiter);
return CosmosClientInitializer<AccountUsage>.Initialize(dBConfig);
}).As<ICosmosDbService<AccountUsage>>();
}
Usage
private readonly IEnumerable<IAccountUsageService> _accountUsageService;
public JobScheduler(IEnumerable<IAccountUsageService> accountUsageService)
{
_accountUsageService = accountUsageService;
}
[FunctionName("JobScheduler")]
public async Task Run([TimerTrigger("0 */2 * * * *")] TimerInfo myTimer, ILogger log)
{
log.LogInformation($"Job Scheduler Timer trigger function executed at: {DateTime.Now}");
try
{
foreach (var usageService in _accountUsageService)
{
var logs = await usageService.GetCurrentAccountUsage("gfkjdsasjfa");
// ...
}
}
I realise this kind of DI usage is not ideal (and does not even work).
Is there a way to structure an Azure Function such that it can execute for different configurations in a compartmentalised manner? Or is this really just fighting against the technology?
You've got a couple of ways to do this - either inject the right dependencies into the function constructor, or resolve them dynamically using a service-locater type approach with a named instance.
Let's consider the second approach and what it would mean for your implementation. As you demonstrated, you'd be looping through your instances and resolving the dependency you want to use, then invoking it
foreach (var usageService in _accountUsageService)
{
var logs = await usageService.GetCurrentAccountUsage("named-instance");
logs.DoSomething();
}
This is technically possible, but now you're doing batch processing - you're doing more than once piece of work that's been triggered by a single event (the timer object), which means you have to deal with a couple of extra problems. What should you do if there's a failure with one of the instances, and what to do if one of the instances is running slowly?
Ideally, you want functions to do the smallest bit of work they can, and complete quickly - You don't want failure or slowness with one particular instance impacting the other instances. By breaking it down to the smallest piece of work (think, one event trigger does one piece of work) then you can take advantage of the functions runtime for things like retries on failures, and threading and concurrency is now being done for you by the runtime.
You could then think about a couple of ways you could do this. a) multiple function signatures and a service resolver approach, e.g.
public class JobScheduler
{
public JobScheduler(IEnumerable<IAccountUsageService> accountUsageService)
{
_accountUsageService = accountUsageService;
}
[FunctionName("FirstInstance")]
public Task FirstInstance([TimerTrigger("%MetricPoller:Schedule%")] TimerInfo myTimer)
{
var logs = await _accountUsageService.GetNamedInstance("instance-a");
logs.DoSomething();
}
[FunctionName("SecondInstance")]
public Task SecondInstance([TimerTrigger("%MetricPoller:Schedule%")] TimerInfo myTimer)
{
var logs = _accountUsageService.GetNamedInstance("instance-b");
logs.DoSomething();
}
}
or b), multiple classes with the necessary dependencies injected
public class JobSchedulerFirstInstance
{
public JobSchedulerFirstInstance(ILogs logs)
{
_logs = logs;
}
[FunctionName("FirstInstance")]
public Task FirstInstance([TimerTrigger("%MetricPoller:Schedule%")] TimerInfo myTimer)
{
_logs.DoSomething();
}
}
I'd personally lean towards multiple classes approach, and register named instances with my container. A bit of extra wire up work needed, but you'll end up with lots of small classes that all look very similar that are basically jus t plumbing that the functions runtime executes.
IoC service should retrieve information from DB that
never changed during session. May be this is constant value.
is time expensive
Should I use in this case Pattern "lazy load" for property that encapsulates access to service method? For example:
public interface IMyDbConstantService {
string MyDbConstant {get;}
string GetMyDbConstant();
}
public class MyDbConstantService: IMyDbConstantService {
public string MyDbConstant
{
get
{
// Implementing lazy load pattern
return myDbConstant ?? (myDbConstant = this.GetMyDbConstant());
}
}
public string GetMyDbConstant() {
// time expensive operation
}
}
I assume myDbConstant is a field? You don't declare it anywhere. I would make it a private static field--you may also need to add locks around updates to it.
I would also remove the getter in the interface and implementation, and just put the lazy initialization directly in the GetMyDBConstant method.
As far as whether this is a good idea: you haven't provided enough information to say one way or another. Does it make more sense to load on on first use, or can you load it at application startup, possibly in a background thread?
What is the preferred way to remove a default pipeline contributor (OpenRasta 2.0.3)?
I haven't found a lot on that on the net, but one way seems to be writing a custom DependencyRegistrar, i.e. deriving from DefaultDependencyRegistrar and then e.g. overriding AddDefaultContributors(). Apart from that I doubt that it's the best way to remove just a single pipeline contributor, it seems to need additional per-host (ASP vs. InMemory) work, whereas I would consider messing with pipeline handlers to be a host-agnostic affair.
But even if I'd go this route, this guy here seems to have tried it without success: http://groups.google.com/group/openrasta/browse_thread/thread/d72b91e5994f402b
I tried similar things, but so far couldn't make my custom registrar replace the default.
So what's the simplest and best way to remove a default pipeline contributor, preferable in a host agnostic way? Is there a working example somewhere?
No, you just need to derive from the registrar and use the protected members that are available to imperatively remove the types you don't want auto-registered.
The registrar needs to be registered in your container before you provide it to OpenRasta, otherwise the type has been resolved already.
Answering myself with working code snippets as they might be helpful to others.
So it looks like removing default pipeline contributors cannot be done
in a host agnostic way (although I don't see why OpenRasta could not
be modified to allow for easy deletion of handlers in the future).
The 2 classes that need to be written are in fact independent of the
host(s) used:
public class MyDependencyRegistrar : DefaultDependencyRegistrar
{
protected override void AddDefaultContributors()
{
base.AddDefaultContributors();
PipelineContributorTypes.Remove(typeof(HandlerResolverContributor));
// If we remove the only contributor for the 'well-known'
// IHandlerSelection stage, like done above, we need to add
// another one implements IHandlerSelection, otherwise
// we'll run into errors (and what's the point of a pipeline
// without a handler selector anyway?). So let's do that here:
AddPipelineContributor<MyOwnHandlerResolverContributor>();
}
}
In order to make that Registrar available, we need to create an accessor
like the following, which then needs to be set in the various hosts:
public class MyDependencyResolverAccessor : IDependencyResolverAccessor
{
InternalDependencyResolver resolver;
public IDependencyResolver Resolver
{
get
{
if (resolver == null)
{
resolver = new InternalDependencyResolver();
resolver.AddDependency<IDependencyRegistrar, MyDependencyRegistrar>();
}
return resolver;
}
}
}
For Asp.Net, this seems to work for me:
public class Global : System.Web.HttpApplication
{
void Application_Start(object sender, EventArgs e)
{
OpenRastaModule.Host.DependencyResolverAccessor =
new MyDependencyResolverAccessor();
For InMemoryHost, which I use for integration testing and in-process access
of my handlers, I haven't found a way around copying the whole class
InMemoryHost and modifying it to my needs. In fact, we don't need
MyDependencyResolverAccessor in this case, as InMemoryHost implements
IDependencyResolverAccessor already. So here's how it could look like. Only the
last line was actually added to the existing code in InMemoryHost:
public class TwinMemoryHost : IHost, IDependencyResolverAccessor, IDisposable
{
readonly IConfigurationSource _configuration;
bool _isDisposed;
public TwinMemoryHost(IConfigurationSource configuration)
{
_configuration = configuration;
Resolver = new InternalDependencyResolver();
Resolver.AddDependency<IDependencyRegistrar, MyDependencyRegistrar>();
...
Recently I've switched to Ninject 2.0 release and started getting the following error:
Error occured: Error activating SomeController
More than one matching bindings are available.
Activation path:
1) Request for SomeController
Suggestions:
1) Ensure that you have defined a binding for SomeController only once.
However, I'm unable to find certain reproduction path. Sometimes it occurs, sometimes it does not.
I'm using NinjectHttpApplication for automatic controllers injection. Controllers are defined in separate assembly:
public class App : NinjectHttpApplication
{
protected override IKernel CreateKernel()
{
INinjectModule[] modules = new INinjectModule[] {
new MiscModule(),
new ProvidersModule(),
new RepositoryModule(),
new ServiceModule()
};
return new StandardKernel(modules);
}
protected override void OnApplicationStarted()
{
RegisterRoutes(RouteTable.Routes);
RegisterAllControllersIn("Sample.Mvc");
base.OnApplicationStarted();
}
/* ............. */
}
Maybe someone is familiar with this error.
Any advice?
I finally figured this issue out recently. Apparently, the NinjectHttpApplication.RegisterAllControllersIn() function doesn't do all of the proper bindings needed. It binds your concrete controller implementations to IController requests. For example, if you have a controller class called SampleMvcController, which inherits from System.Web.Mvc.Controller. It would do the following named binding during application start:
kernel.Bind<IController>().To(SampleMvcController).InTransientScope().Named("SampleMvc");
But when debugging the NinjectControllerFactory, I find that request are being made for the Ninject Kernel to return an object for the class "SampleMvcController", not for a concrete implementation of IController, using the named binding of "SampleMvc".
Because of this, when the first web request that involves the SampleMvcController is made, it creates a binding of SampleMvcController to itself. This is not thread safe though. So if you have several web requests being made at once, the bindings can potentially happen more than once, and now you are left with this error for having multiple bindings for the SampleMvcController.
You can verify this by quickly refreshing an MVC URL, right after causing your web application to restart.
The fix:
The simplest way to fix this issue is to create a new NinjectModule for your controller bindings, and to load this module during application start. Within this module, you self bind each of your defined controllers, like so:
class ControllerModule : StandardModule {
public override Load() {
Bind<SampleMvcController>().ToSelf();
Bind<AnotherMvcController>().ToSelf();
}
}
But if you don't mind changing the Ninject source code, you can modify the RegisterAllControllersIn() function to self bind each controller it comes across.
I have been dealing with this problem for months. I tried so many options but was unable to come to a solution. I knew that it was a threading problem because it would only occur when there was a heavy load on my site. Just recently a bug was reported and fixed in the ninject source code that solves this problem.
Here is a reference to the issue. It was fixed in build 2.1.0.70 of the Ninject source. The key change was in KernelBase.cs by removing the line
context.Plan = planner.GetPlan(service);
and replacing it with
lock (planner)
{
context.Plan = planner.GetPlan(service);
}
To use this new build with MVC you will need to get the latest build of Ninject then get the latest build of ninject.web.mvc. Build ninject.web.mvc with the new Ninject build.
I have been using this new build for about a week with a heavy load and no problems. That is the longest it has gone without a problem so I would consider this to be a solution.
Are you sure you really are creating a single completely new Kernel from scratch in your OnApplicationStarted every time it's invoked ? If you're not and you're actually creating it once but potentially running the registration bit twice. Remember that you're not guaranteed to only ever have one App class instantiated ever within a given AppDomain.
My answer was a bit more obvious.
I had declared the binding for one of my controllers more than once during refactor of my code.
I added this to my global.ascx.cs file:
public void RegisterAllControllersInFix(Assembly assembly)
{
RegisterAllControllersInFix(assembly, GetControllerName);
}
public void RegisterAllControllersInFix(Assembly assembly, Func<Type, string> namingConvention)
{
foreach (Type type in assembly.GetExportedTypes().Where(IsController))
Kernel.Bind(type).ToSelf();
}
private static bool IsController(Type type)
{
return typeof(IController).IsAssignableFrom(type) && type.IsPublic && !type.IsAbstract && !type.IsInterface;
}
private static string GetControllerName(Type type)
{
string name = type.Name.ToLowerInvariant();
if (name.EndsWith("controller"))
name = name.Substring(0, name.IndexOf("controller"));
return name;
}
Then called it from my OnApplicationStarted() method as follows:
RegisterAllControllersIn(Assembly.GetExecutingAssembly());
RegisterAllControllersInFix(Assembly.GetExecutingAssembly());
Difficult to know whether this fixed it though because it's so intermittent.
In the C# language, using StructureMap 2.5.4, targeting .NET Framework 3.5 libraries.
I've taken the step to support multiple Profiles in a structure map DI setup, using ServiceLocator model with Bootstrapper activation. First setup was loading default registry, using the scanner.
Now I like to determine runtime what Registry configuration I like to use. Scanning and loading multiple assemblies with registries.
Seems it's not working for the actual implementation (Getting the 202, default instance not found), but a stripped test version does work. The following setup.
Two assemblies containing Registries and implementations
Scanning them in running AppDomain, providing the shared Interface, and requesting Creation Of Instance, using the interfaces in constructor (which get dealt with thanx to the profile on Invokation)
Working code sample below (same structure for other setup, but with more complex stuff, that get's a 202):
What type of couses are possible for a 202, specifically naming the System.Uri type, not being handles by a default type?? (uri makes no sense)
// let structure map create instance of class tester, that provides the registered
// interfaces in the registries to the constructor of tester.
public class Tester<TPOCO>
{
private ITestMe<TPOCO> _tester;
public Tester(ITestMe<TPOCO> some)
{
_tester = some;
}
public string Exec()
{
return _tester.Execute();
}
}
public static class Main {
public void ExecuteDIFunction() {
ObjectFactory.GetInstance<Tester<string>>().Exec();
}
}
public class ImplementedTestMe<TSome> : ITestMe<TSome>
{
public string Execute()
{
return "Special Execution";
}
}
public class RegistryForSpecial : Registry
{
public RegistryForSpecial()
{
CreateProfile("Special",
gc =>
{
gc.For(typeof(ITestMe<>)).UseConcreteType(typeof(ImplementedTestMe<>));
});
}
}
Background articles on Profiles I used.
How to setup named instances using StructureMap profiles?
http://devlicio.us/blogs/derik_whittaker/archive/2009/01/07/setting-up-profiles-in-structuremap-2-5.aspx
http://structuremap.sourceforge.net/RegistryDSL.htm
EDIT:
It seemed the missing interface was actually the one being determined runtime. So here is the next challange (and solved):
I provided a default object whenever StructureMap needs to create the object. Like:
x.ForRequestedType<IConnectionContext>()
.TheDefault.Is.Object(new WebServiceConnection());
This way I got rid of the 202 error, because now a real instance could be used whever structure map needed the type.
Next was the override on runtime. That did not work out at first using the ObjectFactory.Configure method. Instead I used the ObjectFactory.Inject method to overide the default instance. Works like a charm.
ObjectFactory.Inject(typeof(IConnectionContext), context);
Loving the community effort.
Error code 202 means a default instance could not be built for the requested type. Your test code is apparently not equal to your real code that fails. If you are getting an error about Uri, you likely have a dependency that requires a Uri in its constructor. It may not be the class you are asking for - it may be one of that classes dependendencies - or one of the dependencies dependencies... somewhere down the line someone is asking StructureMap to resolve a Uri, which it cannot do, without some help from you.