Airflow异常：返回代码为1的DataFlow失败

Question

我试图通过气流脚本执行数据流罐。对于它，我使用DataFlowJavaoperator。在参数jar中，我传递本地系统中存在的可执行jar文件的路径。但是当我尝试运行这个工作我得到错误作为

{gcp_dataflow_hook.py:108} INFO - Start waiting for DataFlow process to complete.
[2017-09-12 16:59:38,225] {models.py:1417} ERROR - DataFlow failed with return code 1
Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/airflow/models.py", line 1374, in run
    result = task_copy.execute(context=context)
  File "/usr/lib/python2.7/site-packages/airflow/contrib/operators/dataflow_operator.py", line 116, in execute
    hook.start_java_dataflow(self.task_id, dataflow_options, self.jar)
  File "/usr/lib/python2.7/site-packages/airflow/contrib/hooks/gcp_dataflow_hook.py", line 146, in start_java_dataflow
    task_id, variables, dataflow, name, ["java", "-jar"])
  File "/usr/lib/python2.7/site-packages/airflow/contrib/hooks/gcp_dataflow_hook.py", line 138, in _start_dataflow
    _Dataflow(cmd).wait_for_done()
  File "/usr/lib/python2.7/site-packages/airflow/contrib/hooks/gcp_dataflow_hook.py", line 119, in wait_for_done
    self._proc.returncode))
Exception: DataFlow failed with return code 1`

我的脚本是：

from airflow.contrib.operators.dataflow_operator import DataFlowJavaOperator
from airflow.contrib.hooks.gcs_hook import GoogleCloudStorageHook
from airflow.models import BaseOperator
from airflow.utils.decorators import apply_defaults
from datetime import datetime, timedelta


default_args = {
'owner': 'airflow',
'start_date': datetime(2017, 03, 16),
'email': [<EmailID>],

'dataflow_default_options': {
        'project': '<ProjectId>',
       # 'zone': 'europe-west1-d', (i am not sure what should i pass here)
        'stagingLocation': 'gs://spark_3/staging/'
    }
 }

 dag = DAG('Dataflow',schedule_interval=timedelta(minutes=2), 
 default_args=default_args)

 dataflow1 = DataFlowJavaOperator(
 task_id='dataflow_example',
 jar ='/root/airflow_scripts/csvwriter.jar',
 gcp_conn_id  = 'GCP_smoke', 
 dag=dag)

我不知道我犯了什么错误，有人能帮我摆脱困境吗

注意：我正在创建这个jar，同时通过打包所有外部依赖项选择选项作为可运行的jar文件




Answer 1

问题出在我用的罐子上。在使用jar之前，请确保jar按预期执行。

示例：如果您的jar是dataflow_job1。jar，使用

java -jar dataflow_job_1.jar --parameters_if_any

jar成功运行后，继续在DataflowJavaOperator jar中使用jar。

此外，如果您遇到与编码器相关的错误，您可能必须使自己的编码器来执行代码。例如，我有一个问题与TableRow类，因为它没有一个默认的编码器，因此我必须弥补这一点：

表格行编码器：

public class TableRowCoder extends Coder<TableRow> {
private static final long serialVersionUID = 1L;
private static final Coder<TableRow> tableRow = TableRowJsonCoder.of();
@Override
public void encode(TableRow value, OutputStream outStream) throws CoderException, IOException {
    tableRow.encode(value, outStream);

}
@Override
public TableRow decode(InputStream inStream) throws CoderException, IOException {
    return new TableRow().set("F1", tableRow.decode(inStream));
}
@Override
public List<? extends Coder<?>> getCoderArguments() {
    // TODO Auto-generated method stub
    return null;
}
@Override
public void verifyDeterministic() throws org.apache.beam.sdk.coders.Coder.NonDeterministicException {


}
}

然后在你的代码中注册这个编码器

pipeline.getCoderRegistry().registerCoderForClass(TableRow.class, new TableRowCoder())

如果仍然存在错误（与编码器无关），请导航到：

*.jar\META-INF\services\FileSystemRegistrar 

并添加可能出现的任何依赖项。

例如，可能存在以下暂存错误：

Unable to find registrar for gs

我不得不添加以下行使其工作。

org.apache.beam.sdk.extensions.gcp.storage.GcsFileSystemRegistrar