Xgboost机器学习训练模型在将df转换为h2o数据帧时失败

Question

将df转换为h2o（数据块）数据帧时，Xgboost机器学习训练模型失败。

在执行时，函数spark_app（）失败：

    hf = H2OContext.getOrCreate(spark.spark_app())

但这并不总是发生。很多时候，它运行良好并且可以进行批处理工作，而我计划每天运行它。它每周至少失败一次。

错误日志：

File "*/taskModel.py", line 304, in _sc_xgboost
hc = H2OContext.getOrCreate(spark.spark_app())
File "/databricks/python/lib/python3.5/site-packages/pysparkling/context.py", 
line 170, in getOrCreate
h2o_connect_hook(h2o_context, verbose=verbose, **kwargs)
File "/databricks/python/lib/python3.5/site-packages/pysparkling/context.py", 
line 119, in __default_h2o_connect
return h2o.connect(ip=h2o_context._client_ip, port=h2o_context._client_port, 
**kwargs)
File "/databricks/python/lib/python3.5/site-packages/h2o/h2o.py", line 90, in 
connect
verbose=verbose)
File "/databricks/python/lib/python3.5/site- 
packages/h2o/backend/connection.py", line 323, in open
conn._cluster = conn._test_connection(retries, messages=_msgs)
File "/databricks/python/lib/python3.5/site- 
packages/h2o/backend/connection.py", line 598, in _test_connection
raise H2OServerError("Cluster reports unhealthy status")
h2o.exceptions.H2OServerError: Cluster reports unhealthy status