DroidSolutions.Oss.JobService.Postgres 3.6.0

DroidSolutions Job Service

Library to manage recurring jobs.

The DroidSolutions Job service offers tools that help managing recurring jobs. It is split in multiple packages that are explained below. These contains NuGet packages for .NET 7 as well as an NPM package.

Examples of jobs are:

• check every hour for new entries in a database to process
• download current information from an API every day
• run a job to clean up no longer needed data
• delay a task coming from UI to execute at a later time

Why not Cron?

For recurring jobs you could also use a cronjob and there is nothing wrong with that. In general this library allows to have a similar experience like with a cron job with the following differences:

• Jobs are not executed exactly at a given time. Instead they have a due date and whenever a runner is checking for a job it will execute the job with the oldest due date that's already passed. Depending on the check interval you configure there might be some time between due date and actual execution. In fact, jobs do not need to be recurring at all, you can use this library to manage one time jobs.
• Job execution intervals are not always equal. When a job is executed and it is configured to add another job, the next job is added after the previous job is done executing. The next due date will be calculated from that point on, so time between jobs depend on the execution time and the configured interval.
• Jobs are persistent, you have a storage for all jobs (presumably a database table) that you can use to display job executions or view results of the jobs.
• Job execution can be integrated in your application. While a cronjob probably is an own piece of software or a script, the job worker can be integrated in your application and share code with it. This is useful when the job needs logic that other parts of your application also need. You also can easily dynamically add a job from your application logic if it is needed.

Packages

This section describes which packages are availabe for which eco systems.

DroidSolutions.Oss.JobService

This NuGet package contains interfaces for a job and a job repository as well as a base worker service to work with them.

DroidSolutions.Oss.JobService.EFCore

This NuGet package contains a concrete implementation of the IJob and IJobRepository using Entity Framework Core. It also offers an interface that a DbContext can implement that the repository implementation needs.

DroidSolutions.Oss.JobService.Postgres

This NuGet package contains a concrete implementation of the IJobRepository with Postgres specific JSON querying using Npgsql.

@droidsolutions-oss/job-service

This NPM package contains TypeScript interfaces that are generated from the .NET interfaces of the DroidSolutions.Oss.JobService package and an abstract JobWorkerBase class. These can be used for a NodeJS implementation of the job repository and the worker service.

Interfaces and Classes

The main interfaces and classes of the packages are described in this section.

IJobBase

The basis of this library are jobs. Jobs can be added, started, finished and processed.

This interface is available in C# as well as TypeScript.

A job is meant to be stored somewhere (like a table in a database) and holds information about the job. This includes dates of creation, last update and when the job is due. There is also a state and a type to distinguish jobs of different types as well as optional properties that relate to job progress.

This interface describes base jobs without any parameters or result. They can be used when they don't need additional information to run or store their result elsewhere.

IJob<TParams, TResult>

An extension of the base jobs with generic type arguments for parameters and results.

In addition to base jobs they store parameters and a result with them. There is an interface IJob<TParams, TResult> which serves as the basis for a concrete database entity.

This interface is available in C# as well as TypeScript.

Jobs are generic, where the type arguments are meant for parameters and the result that a job can have. Those can be serialized to and from JSON when interacting with the database.

There is a concrete implementations of IJob<TParams, TResult> for Entity Framework Core in the DroidSolutions.Oss.JobService.EFCore package.

Entity Framework Core specifics

The JobBase class is a concrete implementation of the IJobBase interface which has annotations that may be useful when using it as an entity for Entity Framework Core.

The Job<TParams, TResult> class is an extension that implementats the IJob<TParams, TResult> interface. The paramaters and results are currently directly mapped to jsonb column for PostgreSQL.

Internally the properties for parameter and result are mapped to fields in the database to allow interoperatibility with NodeJS or other projects. This means there are ParametersSerialized and ResultSerialized properties that contain the JSON string and are mapped to the database where the Paramters and Result properties contain the actual deserialized values for you to use.

When implementing your own IJobRepository<TParams, TResult> you are responsible for serializing and deserializing the values yourself.

IJobRepository<TParams, TResult>

There is a C# and a TypeScript interface for a repository that works with the job entity. It can be implemented to serve as a wrapper for database related actions. The generic type parameters are the same as for jobs, so a repository is repsonsible for exactly one type of jobs. The repository acts as the data layer of the application and wraps around all database interaction regarding jobs.

Like for the job there are concrete implementations for EF Core and TypeORM.

Entity Framework Core specifics

On the .NET side there is an interface IJobContext that can be implemented in your DbContext. This uses the JobBase class which has annotations that may be useful when using it as an entity for Entity Framework Core.

The JobRepositoryBase<TContext, TParams, TResult> is a full implementation of the IJobRepository<TParams, TResult> interface using the Job<TParams, TResult> entity where necessary. Parameters and results are optional, that's why the IJobContext only uses the BaseJob entity. The JobRepositoryBase<TContext, TParams, TResult> handles checks for parameter and result existence and uses JobBase where they don't apply. If theyare needed, JobBase entities are casted to Job<TParams, TResult>. For this to work, your DbContext must register the entity, for example like this:

public class TestContext : DbContext, IJobContext
{
  public DbSet<JobBase> Jobs { get; set; }

  protected override void OnModelCreating(ModelBuilder modelBuilder)
  {
    // Register entity with concrete params and result types
    modelBuilder.Entity<Job<TestParameter, TestResult>>();
  }
}

To allow interoperatibility with job workers that use NodeJS the parameters and result are serialized with special options using camelCase property names. This can be extended or overwritten via the protected GetSerializerOptions method. In the background the Job<TParams, TResult> class has ParametersSerialized and ResultSerialized properties that hold the converted json and act as the actual database columns. The repository handles this in all methods where it is necessary so you shouldn't have to do this yourself.

Adding jobs, starting jobs and setting job progress is done via transactions and table or row locks. This ensures no two parallel running worker can receive the same job or work on the same job.

The repository constructor expects an instance of the data context that implements IJobContext as well a an ILogger instance. If you have set up dependency injection in a standard ASP.NET Core app there should be no problem, just make sure you have added the db context that implements IJobContext to the depdency injection via services.AddDbContext().

Worker

The JobWorker<TParams, TResult> (C#) or JobWorkerBase (TypeScript) is an abstract base class for managing the jobs. It uses the IJobRepository<TParams, TResult> interface to work with the database. It checks for available jobs, calls a processing function for each and finishes the job (or resets it in case of a failure). It offers some extensibility methods where the implementer can add code specific for the type of job it wants to process.

Pass an instance of the JobRepository or any other class that implements IJobRepository<TParams, TResult> to the constructor, along with a settings object. You can use the IJobWorkerSettings interface in NodeJS to know which properties are expected.

There are two lifecycle hooks before and after each run that can optionally be implemented to add custom logic before and after each job run.

ASP.NET Core specifics

Add a class for your worker that extends the JobWorker<TParams, TResult> class and implements the abstract methods.

The extended worker class can be added as a hosted service. Be sure to configure the dependency injection to be able to provide an OptionsMonitor with the JobWorkerSettings to the worker constructor. For example you could use something like this services.Configure<JobWorkerSettings>(Configuration.GetSection("WorkerSettings"));. You can also extend the settings class if you need other settings for your implemented worker.

The PreJobRunHook is called when a new run is executed. At this time the worker has not checked if a job is available and therefore no job exists yet. This hook can be used to set up a logger, correlation id or Sentry transactions, though it is not guarenteed that a job is available.

Note The JobWorkerBase uses the NanoId package. Up until version 2.1.2 of this package NanoId 2.1.0 was used. In 3.0.0 NanoId changed the general namespace which can lead to incompatibilities. Since version 3.0.0 of the DroidSolutions.Oss.JobService package NanoId 3.0.0 is used. If you don't use the NanoId package yourself, this should not be a problem, but if you use a different version of NanoId there can be problems.

NodeJS specifics

The TypeScript side has an abstract JobWorkerBase class that can be used as the basis of a worker implementation. For this you'll need a concrete implementation of the IJobRepository interface and pass it along with settings to the constructor. You'll then have to implement the abstract methods and you'll should be good to go.

Usage

The following is a detailed guide on how to use this library for either ASP.NET Core (C#) or NodeJS (TypeScript).

ASP.NET Core

The concrete implementation depends on the kind of database you want to use. There are already implementations for Entity Framework Core and PostgreSQL for which the following guide is specific to.

1. Add the DroidSolutions.Oss.JobService, DroidSolutions.Oss.JobService.EFCore and DroidSolutions.Oss.JobService.Postgres packages as well as references to Microsoft.EntityFrameworkCore and Npgsql.EntityFrameworkCore.PostgreSQL.

2. Create types for your job parameters and result. If you don't need one or both of them you can use object?.

3. Let your db context implement the IJobContext interface. That could look like this

   public class TestContext : DbContext, IJobContext
   {
     public TestContext([NotNull] DbContextOptions options)
       : base(options)
     {
     }
   
     public DbSet<JobBase> Jobs { get; set; }
   }
   

4. Register the job entities you need in the DbContext, if you don't need parameters and or results, use object:

   public class TestContext : DbContext, IJobContext
   {
      protected override void OnModelCreating(ModelBuilder modelBuilder)
      {
        // Register entity with object for jobs without params and result
        modelBuilder.Entity<Job<object, object>>();
        // or register entity with concrete params and result types
        modelBuilder.Entity<Job<TestParameter, TestResult>>();
      }
   }
   

5. If you want to share the database with services in other languages (such as NodeJS) or generally want to have the JobState enum as a text column you can use the JobStateToDescriptionConverter by adding it in the OnConfiguring method of the db context.

   using DroidSolutions.Oss.JobService.EFCore.Entity;
   using DroidSolutions.Oss.JobService.EFCore.Converter;
   // ...
     protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)
     {
       modelBuilder.Entity<Job<TestParameter, TestResult>>(entity =>
       {
         entity.Property(x => x.State).HasConversion(new JobStateToDescriptionConverter());
       });
     }
   

6. Register the PostgresJobRepository class in the dependency injection. For example

   services.AddScoped<IJobRepository<TestParameter, TestResult>, PostgresJobRepository<TestContext, TestParameter, TestResult>>();
   

   Don't forget to also register the db context.

7. Add your own worker settings class extending from JobWorkerSettings. This contains settings that control how your worker behaves.

   public class DeleteJobSettings : JobWorkerSettings
   {
     public DeleteJobSettings()
       : base("visitor:cleanup")
     {
     }
   
     public int MySpecialSetting { get; set; }
   }
   

8. Add a class that extends and implements JobWorkerBase<TParams, TResult>. Inject an IOptionsMonitor<DeleteJobSettings> (or the default JobWorkerSettings), an ILoggerFactory and an instance of the IServiceProvider. Pass all three of those to the base contructor and implement the abstract methods. The result could look like this:

   public class DeleteVisitorWorker : JobWorkerBase<TestParameter, TestResult>
   {
     public DeleteVisitorWorker(
       IOptionsMonitor<DeleteJobSettings> settings,
       IServiceProvider serviceProvider,
       ILoggerFactory loggerFactory)
       : base(settings, serviceProvider, loggerFactory.CreateLogger<JobWorkerBase<TestParameter, TestResult>>())
     {
     }
   
     protected override string GetRunnerName()
     {
       string version = typeof(Program).Assembly
         .GetCustomAttribute<AssemblyInformationalVersionAttribute>()
         .InformationalVersion
   
       return $"my-app-name-{version}";
     }
   
     // optional
     protected override TestParameter? GetInitialJobParameters()
     {
       return new TestParameter { SomeProp = "MyValue" };
     }
   
     // optional
     protected override void PreJobRunHook()
     {
       // optional logic before a job run starts
     }
   
     // optional
     protected override void PostJobRunHook()
     {
       // optional logic after a job is complete
     }
   
     /// <inheritdoc/>
     protected override async Task<TestResult> ProcessJobAsync(
       IJob<TestParams, TestResult> job,
       IServiceScope serviceScope,
       CancellationToken cancellationToken)
     {
       var myService = serviceScope.ServiceProvider.GetService<MyService>();
   
       var result = await myService.DoSomething(job.Parameters, cancellationToken);
   
       return new TestResult
       {
         MyProp = result.Something,
       };
     }
   }
   

   • The GetRunnerName method should return a string that is unique for this worker class. The base implementation will add a random string to it so it is distinguishable if your application is running in more than one instance.
   • GetInitialJobParameters is called when you set AddInitialJob in your settings to true. This way the worker will call this method to get the parameters of the first job.
   • The PreJobRunHook is called once before the worker checks if a job is available and can be used for custom logic. For example you could generate a correlation id and set it to the log context for following logs.
   • The PostJobRunHook is called once after a job run and can be used for custom cleanup logic. For example you could remove a previously set correlation id from the log context. The post hook is also called when no job was actually executed.
   • The ProcessJobAsync method is where you put your logic to actually process the job. it is only called when a job to execute exists and contains the job, a service scope for this execution and a CancellationToken. If your job has a result you should return it from this method.

9. Add your worker as a hosted service.

   services.AddHostedService<DeleteVisitorWorker>();
   

10. Generate Migration to add or update the Job table.

NodeJS

The worker service can be used in JavaScript/TypeScript projects.

1. Install the package

   npm install @droidsolutions-oss/job-service
   

2. Create an implementation of the IJobRepository<TParams, TResult> interface. If you are using TypeORM you can use the @droidsolutions-oss/job-service-typeorm package which already comes with an implemantation of the interface. Otherwise you can implement it with the database layer of your choice. In TypeScript it would look like this (note Prisma is used in this example):

   import { IJob, IJobRepository, JobState } from "@droidsolutions-oss/job-service";
   import { Prisma, PrismaClient } from "@prisma/client";
   
   // Note: if you only need one type of job you can also use the type directly instead of a generic here
   export class JobRepository<TParams, TResult> implements IJobRepository<TParams, TResult> {
     constructor(private readonly client: PrismaClient) {}
   
     public async addJobAsync(
       type: string,
       dueDate?: Date,
       parameters?: TParams,
       cancellationToken?: AbortSignal,
     ): Promise<IJob<TParams, TResult>> {
       cancellationToken?.throwIfAborted();
   
       // create a new job
       const job = await this.client.job.create({
         data: {
           dueDate: dueDate ?? new Date(),
           parameters: parameters as Prisma.JsonObject,
           state: JobState.Requested,
           type,
         },
       });
   
       // save the job to the database, reload it and then return it with the new id
   
       return job;
     }
   
     // Implement the other methods from the interface
   }
   

3. Optional if you need custom settings create your own settings interface that extends from IJobWorkerSettings.

   import { IJobWorkerSettings } from "@droidsolutions-oss/job-service";
   
   export interface MyWorkerSettings extends IJobWorkerSettings {
     checkInput: boolean;
   }
   

4. Create your worker, extending from JobWorkerBase<TParams, TResult>.

   import { IJob, JobWorkerBase, LoggerFactory } from "@droidsolutions-oss/job-service";
   import { JobRepository } from "../repository/JobRepository";
   import { MyWorkerSettings } from "../dto/MyWorkerSettings";
   
   export interface ExampleResult {
     errors?: string[];
   }
   
   export class ExampleWorker extends JobWorkerBase<string[], ExampleResult> {
     constructor(
       settings: MyWorkerSettings,
       jobRepo: JobRepository<string[], ExampleResult>,
       loggerFactory?: LoggerFactory,
       private readonly appVersion: string,
     ) {
       super(settings, jobRepo, loggerFactory);
     }
   
     public getRunnerName(): string {
      // Use host name and the current application version, JobWorkerBase will append a random string to it
       return `${process.env.HOSTNAME}-v${this.appVersion}`;
     }
   
     public getInitialJobParameters(): string[] | undefined {
       return undefined;
     }
   
     public preJobRunHook(): void {
       // Optional: run any things you want before each job run.
       // This will be executed before checking if there is a job to execute
     }
   
     public postJobRunHook(): void {
       // Optional: run any things for cleanup
       // This will be executed after each job, even if there was no job to execute
     }
   
     public async processJobAsync(job: IJob<string[], ExampleResult>, cancellationToken: AbortSignal): Promise<ExampleResult> {
       // You can safely throw here, it will be catched by JobWorkerBase
       cancellationToken.throwIfAborted();
       const result: ExampleResult = {
         errors: [];
       }
   
       for (const input of job.parameters)
       {
         if (cancellationToken.aborted) {
           break;
         }
   
         try {
           // Do you job handling here
         } catch (err) {
           result.errors.push[`Error handling input ${input}: err.message`];
         }
       }
   
       return result;
     }
   }
   

5. Let your application initialize the worker.

   import { Prisma, PrismaClient } from "@prisma/client";
   import { MyWorkerSettings } from "./dto/MyWorkerSettings";
   import { JobRepository } from "./repository/JobRepository";
   import { ExampleResult, ExampleWorker } from "./worker/ExampleWorker";
   
   const prismaClient = new PrismaClient();
   const exampleRepo = new JobRepository<string[], ExampleResult>>(prismaClient);
   const settings: MyWorkerSettings = {
     addInitialJob: true, // Create a job if none exists
     addNextJobAfter: { hours: 1 }, // Repeat job every hour
     jobType: "my-example-job",
     checkInput: true
   };
   const abortController = new AbortController();
   
   const worker = new ExampleWorker(settings, exampleRepo, undefined); // Optional provide a logger factory
   const executePromise = worker.executeAsync(abortController.signal);
   
   // When you want to stop the worker (e.g. shutting down the app) use the controller to abort
   abortController.abort(new Error("App is shutting down"));
   await executePromise; // Promise will resolve once the worker finished processing the current job or reset it if cancellationToken.throwIfAborted(); was used
   

Worker

The worker service is a kind of background service that regularily checks if a job should be executed and executes the job with the earliest due date. It can also add a new job after execution thus creating end endless reoccuring job.

The worker is controlled via settings. You can create your own settings class that must extend from JobWorkerSettings and add your own properties that you need for job processing. But you can also just just the provided settings if you don't need your own. Those settings are explained below.

JobWorkerSettings

JobWorkerSettings control how to worker behaves.

JobType

(Required)

The most important one is the type. This string controls which type of job the worker processes. Only jobs with the exact same type are fetched and executed by the worker.

You can provide the job type in the constructor when instantiating your settings instance or set the JobType property.

InitialDelaySeconds

(Optional, default 30)

This settings controls the initial delay before the worker will start. This is used to prevent the first job run before the rest of the application is finished starting, due to the way ASP.NET Core handles BackgroundServices. The worker will wait the given amount of seconds before looking for the first job.

JobPollingIntervalSeconds

(Optional, default 10)

This settings controls the time between job processings. Once the InitialDelaySeconds have passed the worker will begin looking for the job of the type specified in the JobType that has the oldest due date. After the job is processed or if no job exists the worker will wait the amount of seconds given in this setting until it checks for a job again.

AddNextJobAfter

(Optional, default null)

If a TimeSpan is given to this setting then the worker will add a new job after it finishes processing one and set the due date to the current date plus the time span. With this you can create an endless job that is executed every given time span.

Leave this empty if you want a one time job (or create jobs from somewhere else).

Note: You can specify TimeSpan values via appsettings.json by following Standard TimeSpan format strings.

The JobWorkerBase has a protected method AddNextJobIn that receives the settings instance and the job result. It can be overridden to customize if a new job should be added based on the result of a job. For example you can create a new job if something didn't work out and you want to create a new job to retry the action.

AddInitialJob

(Optional, default false)

If given the worker will look for a job of the given type with a due date in the past. If none is found the worker creates one and calls the GetInitialJobParameters method to get the parameters of it. The due date of the initial job will be the current date plus the AddNextJobAfter period of time.

DeleteJobsOlderThan

(Optional, default null)

If a TimeSpan is given to this setting then the worker will delete finished jobs after each job run. From the current timestamp the given timespan is substracted, every finished job of the type that is set via JobType whose UpdatedAt is older than the calculated time is removed from the database. You can use this to prevent an ever growing database, especially when you run jobs in short intervalls.

The worker will delete jobs the first time it runs and than every 24 hours. The cleaning will be between PreJobRunHook and PostJobRunHook after the current job was executed.

Note: You can specify TimeSpan values via appsettings.json by following Standard TimeSpan format strings.

In NodeJS you can give an object with days, hours, minutes, seconds or a combination of those. For example, to remove jobs older than 6 months you could set this in you config.json:

{
  "someWorker": {
    "jobType": "some-job",
    "deleteJobsOlderThan": {
      "days": 180
    }
  }
}

Metrics

The job worker currently implements two metrics using the .NET metrics. The static Meter is protected, so your worker can use it to add its own metrics to it.

For example you can create a counter and then add to it:

public class MyWorker : JobWorkerBase<void, void>
{
  private static readonly Counter<int> MyCounter = WorkerMeter.CreateCounter<int>("my_counter");

  // ...
  protected override async Task<TestResult> ProcessJobAsync(
    IJob<void, void> job,
    IServiceScope serviceScope,
    CancellationToken cancellationToken)
  {
    // process job, then increment counter for some objects in your job
    MyCounter.Add(1);
  }
}

Health

Implementing proper health checking of your job worker is up to you, however there are some metrics provided by the JobWorkerBase that can help you determining if the worker is healthy.

The protected properties LastJobExecutionStart and LastJobExecutionStop (lastJobExecutionStart and lastJobExecutionStop in NodeJS) are set each time the main execution loop starts and ends. This can be used (along with JobPollingIntervalSeconds from worker settings) to calculate if the loop is still beeing executed.

You can also retrieve basic worker metrics so you know about last time a job actually finished and how long it took to process.