Message queue batch processing refund order through queue

In the last tutorial, we saw the benefits of splitting a single time-consuming Queue task into multiple subtasks ：

But it also introduces new problems , If you use Redis As a queue driver , There will be multiple executions Redis Command push Queue task problem , And every time you do it Redis Command to establish a network connection ,IO Costly .

Batch push

So , We can go through Laravel Message queuing provides The batch Function to push multiple tasks at once to improve queue performance .

Take the refund business as an example , We can pass before  each  Function to traverse all participants , Then push... In turn in the closure  RefundAttendee  The queue task operation is adjusted to pass  map  The function returns all valid  RefundAttendee  Queue task instances are combined into one  jobs  aggregate , And then through  Bus::batch($jobs)->dispatch()  Batch push  $jobs  All tasks in the collection are executed to the message queue ：

In order to make  RefundAttendee  The task class supports all the functions of batch processing , Can be introduced in this class  Batchable Trait, So that the batch instance to which the current task belongs can be obtained later , And cancel batch processing ：

If you use a database as a queue driver , A new data table needs to be introduced  job_batches：

You can refer to  Laravel Queue documents Learn more about Queue task batching .

Cancel batch

You can get the corresponding batch instance after executing the batch push , And then take it. ID Persist to  conference  Model instance  refunds_batch  Field ：

If you want to cancel the batch operation in the middle , You can use the batch stored above ID Get the corresponding batch instance , Then execute... On this instance  cancel  Method to cancel it ：

however , It should be noted that , This will only set the batch records stored in the data table to  canceled, You cannot recall a task that has been pushed to a queue .

To plug this loophole , Before executing a queue task, you need to determine whether the batch it belongs to is valid , If the batch has been canceled , Then the task will not be performed ：

Processing batch failed

The default configuration is that if any task fails to execute in the batch, the whole batch will be canceled , This means that if batch processing 1000 Refunds , But it goes to 300 I made a mistake when I was , Then the rest will not be executed because the batch has been canceled .

To force batch processing to continue even if a task fails , This can be done by... At the time of distribution  allowFailures()  Method to set ：

thus , Even if a task fails , Batch processing will also continue , Then the failed task ID Store in  job_batches  surface .

You can retry a single batch of failed tasks with the following command ：

You can also retry all failed tasks in the specified batch at once ：