Skip to main content
Skip to main content

Create a Batch Job Strategy

In this document, you’ll learn how to create a batch job strategy in Medusa.

Info

If you’re interested to learn more about what Batch Jobs are and how they work, check out this documentation.

Overview​

Batch jobs can be used to perform long tasks in the background of your Medusa backend. Batch jobs are handled by batch job strategies. An example of a batch job strategy is the Import Products functionality.

This documentation helps you learn how to create a batch job strategy. The batch job strategy used in this example changes the status of all draft products to published.

Prerequisites​

Medusa Components​

It is assumed that you already have a Medusa backend installed and set up. If not, you can follow our quickstart guide to get started. The Medusa backend must also have an event bus module installed, which is available when using the default Medusa backend starter.

1. Create a File​

A batch job strategy is essentially a class defined in a TypeScript or JavaScript file. You should create this file in src/strategies.

Following the example used in this documentation, create the file src/strategies/publish.ts.

2. Create Class​

Batch job strategies must extend the abstract class AbstractBatchJobStrategy and implement its abstract methods.

Add the following content to the file you created:

src/strategies/publish.ts
import { 
  AbstractBatchJobStrategy, 
  BatchJobService, 
} from "@medusajs/medusa"
import { EntityManager } from "typeorm"

class PublishStrategy extends AbstractBatchJobStrategy {
  protected batchJobService_: BatchJobService
  processJob(batchJobId: string): Promise<void> {
    throw new Error("Method not implemented.")
  }
  buildTemplate(): Promise<string> {
    throw new Error("Method not implemented.")
  }
  protected manager_: EntityManager
  protected transactionManager_: EntityManager

}

export default PublishStrategy

3. Define Required Properties​

A batch job strategy class must have two static properties: the identifier and batchType properties. The identifier must be a unique string associated with your batch job strategy, and batchType must be the batch job's type.

You will use the batchType later when you interact with the Batch Job APIs.

Following the same example, add the following properties to the PublishStrategy class:

class PublishStrategy extends AbstractBatchJobStrategy {
  static identifier = "publish-products-strategy"
  static batchType = "publish-products"

  // ...
}

4. Define Methods​

(Optional) prepareBatchJobForProcessing​

Medusa runs this method before it creates the batch job to prepare the content of the batch job record in the database. It accepts two parameters: the batch job data sent in the body of the Create Batch Job request, and the request instance.

Implementing this method is optional. For example:

class PublishStrategy extends AbstractBatchJobStrategy {
  // ...
  async prepareBatchJobForProcessing(
    batchJob: CreateBatchJobInput, 
    req: Express.Request
  ): Promise<CreateBatchJobInput> {
    // make changes to the batch job's fields...
    return batchJob
  }
}

(Optional) preProcessBatchJob​

Medusa runs this method after it creates the batch job, but before it is confirmed and processed. You can use this method to perform any necessary action before the batch job is processed. You can also use this method to add information related to the expected result.

For example, this implementation of the preProcessBatchJob method calculates how many draft products it will published and adds it to the result attribute of the batch job:

class PublishStrategy extends AbstractBatchJobStrategy {
    // ...
  async preProcessBatchJob(batchJobId: string): Promise<void> {
    return await this.atomicPhase_(
      async (transactionManager) => {
        const batchJob = (await this.batchJobService_
          .withTransaction(transactionManager)
          .retrieve(batchJobId))
          
        const count = await this.productService_
          .withTransaction(transactionManager)
          .count({
            status: ProductStatus.DRAFT,
          })

        await this.batchJobService_
          .withTransaction(transactionManager)
          .update(batchJob, {
            result: {
              advancement_count: 0,
              count,
              stat_descriptors: [
                {
                  key: "product-publish-count",
                  name: "Number of products to publish",
                  message: 
                    `${count} product(s) will be published.`,
                },
              ],
            },
          })
    })
  }
}

The result attribute is an object that can hold many properties including:

  • count: used to indicate how many items (in this case, products) that the task will run on.
  • advancement_count: used to indicate the current number of items processed at a given moment. Since the batch job isn't processed yet, you set it to zero.
  • stat_descriptors: can be used to show human-readable messages.

processJob​

Medusa runs this method to process the batch job once it is confirmed.

For example, this implementation of the processJob method retrieves all draft products and changes their status to published:

class PublishStrategy extends AbstractBatchJobStrategy {
  // ...
  async processJob(batchJobId: string): Promise<void> {
    return await this.atomicPhase_(
      async (transactionManager) => {
        const productServiceTx = this.productService_
          .withTransaction(transactionManager)

        const productList = await productServiceTx
          .list({
            status: [ProductStatus.DRAFT],
          })

        productList.forEach(async (product: Product) => {
          await productServiceTx
            .update(product.id, {
              status: ProductStatus.PUBLISHED,
            })
        })
          
        await this.batchJobService_
          .withTransaction(transactionManager)
          .update(batchJobId, {
            result: {
              advancement_count: productList.length,
            },
          })
      }
    )
  }
}
Note

When a batch job is canceled, the processing of the batch job doesn’t automatically stop. You will have to manually check for changes in the status of the batch job. For example, you can retrieve the batch job and use the condition batchJob.status === BatchJobStatus.CANCELED to check if the batch job was canceled.

buildTemplate​

This method can be used in cases where you provide a template file to download, such as when implementing an import or export functionality.

If not necessary to your use case, you can simply return an empty string:

class PublishStrategy extends AbstractBatchJobStrategy {
  // ...
  async buildTemplate(): Promise<string> {
    return ""
  }
}

(Optional) shouldRetryOnProcessingError​

Medusa uses this method to decide whether it should retry the batch job if an error occurs during processing.

By default, the AbstractBatchJobStrategy class implements this method and returns false, indicating that if a batch job’s process fails it will not be retried.

If you would like to change that behavior, you can override this method to return a different value:

class PublishStrategy extends AbstractBatchJobStrategy {
  // ...
  protected async shouldRetryOnProcessingError(
    batchJob: BatchJob, 
    err: unknown
  ): Promise<boolean> {
    return true
  }
}

(Optional) handleProcessingError​

Medusa uses this method to handle errors that occur during processing. By default, it changes the status of the batch job to failed and sets the errors property of the batch job’s result attribute.

You can use this method as implemented in AbstractBatchJobStrategy at any point in your batch job process to set the batch job as failed.

You can also override this method in your batch job strategy and change how it works:

class PublishStrategy extends AbstractBatchJobStrategy {
  // ...
  protected async handleProcessingError<T>(
    batchJobId: string, 
    err: unknown, 
    result: T
  ): Promise<void> {
    // different implementation...
  }
}

5. Run Build Command​

After you create the batch job and before testing it out, you must run the build command in the directory of your Medusa backend:

npm run build

Test your Batch Job Strategy​

This section covers how to test and use your batch job strategy. Make sure to start your backend first:

npx medusa develop

You must also use an authenticated user to send batch job requests. You can refer to the authentication guide in the API reference for more details.

If you follow along with the JS Client code snippets, make sure to install and set it up first.

Create Batch Job​

The first step is to create a batch job using the Create Batch Job API Route. In the body of the request, you must set the type to the value of batchType in the batch job strategy you created.

For example, this creates a batch job of the type publish-products:

medusa.admin.batchJobs.create({
  type: "publish-products",
  context: { },
  dry_run: true,
})
.then(( batch_job ) => {
  console.log(batch_job.status)
})

You set the dry_run to true to disable automatic confirmation and running of the batch job. If you want the batch job to run automatically, you can remove this body parameter.

Make sure to replace <BACKEND_URL> with the backend URL where applicable.

(Optional) Retrieve Batch Job​

You can retrieve the batch job afterward to get its status and view details about the process in the result property:

medusa.admin.batchJobs.retrieve(batchJobId)
.then(( batch_job ) => {
  console.log(batch_job.status, batch_job.result)
})

Based on the batch job strategy implemented in this documentation, the result property could be something like this:

"result": {
    "count": 1,
    "stat_descriptors": [
        {
            "key": "product-publish-count",
            "name": "Number of products to publish",
            "message": "1 product(s) will be published."
        }
    ],
    "advancement_count": 0
},

Confirm Batch Job​

To process the batch job, send a request to confirm the batch job:

medusa.admin.batchJobs.confirm(batchJobId)
.then(( batch_job ) => {
  console.log(batch_job.status)
})

The batch job will start processing afterward. Based on the batch job strategy implemented in this documentation, draft products will be published.

You can retrieve the batch job at any given point to check its status.

Was this section helpful?
© 2024 Medusa, Inc. All rights reserved.