Cleanup logging

openeogeotrellis/backend.py

@@ -35,7 +35,7 @@
 
 from openeo.internal.process_graph_visitor import ProcessGraphVisitor
 from openeo.metadata import TemporalDimension, SpatialDimension, Band, BandDimension
-from openeo.util import dict_no_none, rfc3339, deep_get
+from openeo.util import dict_no_none, rfc3339, deep_get, repr_truncate
 from openeo_driver import backend
 from openeo_driver.ProcessGraphDeserializer import ConcreteProcessing, ENV_SAVE_RESULT
 from openeo_driver.backend import (ServiceMetadata, BatchJobMetadata, OidcProvider, ErrorSummary, LoadParameters,
@@ -817,25 +817,56 @@ def accept_process_graph(cls, process_graph):
             return SingleNodeUDFProcessGraphVisitor().accept_process_graph(process_graph)
         return GeotrellisTileProcessGraphVisitor().accept_process_graph(process_graph)
 
-    def summarize_exception(self, error: Exception) -> Union[ErrorSummary, Exception]:
-        if isinstance(error, Py4JJavaError):
+    def summarize_exception(self, error: Exception, width=2000) -> Union[ErrorSummary, Exception]:
+        return self.summarize_exception_static(error, width)
+
+    @staticmethod
+    def summarize_exception_static(error: Exception, width=2000) -> Union[ErrorSummary, Exception]:
+        if "Container killed on request. Exit code is 143" in str(error):
+            is_client_error = False  # Give user the benefit of doubt.
+            summary = "Your batch job failed because workers used too much Python memory. The same task was attempted multiple times. Consider increasing executor-memoryOverhead or contact the developers to investigate."
+
+        elif isinstance(error, Py4JJavaError):
             java_exception = error.java_exception
 
             while java_exception.getCause() is not None and java_exception != java_exception.getCause():
                 java_exception = java_exception.getCause()
 
             java_exception_class_name = java_exception.getClass().getName()
-            java_exception_message = java_exception.getMessage()
+            java_exception_message = repr_truncate(java_exception.getMessage(), width=width)
 
             no_data_found = (java_exception_class_name == 'java.lang.AssertionError'
                              and "Cannot stitch empty collection" in java_exception_message)
 
             is_client_error = java_exception_class_name == 'java.lang.IllegalArgumentException' or no_data_found
-            summary = "Cannot construct an image because the given boundaries resulted in an empty image collection" if no_data_found else java_exception_message
+            if no_data_found:
+                summary = "Cannot construct an image because the given boundaries resulted in an empty image collection"
+            elif "SparkException" in java_exception_class_name:
+                udf_stacktrace = GeoPySparkBackendImplementation.extract_udf_stacktrace(java_exception_message)
+                if udf_stacktrace:
+                    summary = f"UDF Exception during Spark execution: {udf_stacktrace}"
+                else:
+                    summary = f"Exception during Spark execution: {java_exception_message}"
+            else:
+                summary = java_exception_message
+        else:
+            is_client_error = False  # Give user the benefit of doubt.
+            summary = repr_truncate(error, width=width)
 
-            return ErrorSummary(error, is_client_error, summary)
+        return ErrorSummary(error, is_client_error, summary)
 
-        return error
+    @staticmethod
+    def extract_udf_stacktrace(full_stacktrace) -> Optional[str]:
+        """
+        Select all lines a bit under 'run_udf_code'.
+        This is what interests the user
+        """
+        regex = re.compile(r" in run_udf_code\n.*\n((.|\n)*)", re.MULTILINE)
+
+        match = regex.search(full_stacktrace)
+        if match:
+            return match.group(1).rstrip()
+        return None
 
     def changelog(self) -> Union[str, Path]:
         roots = []

openeogeotrellis/deploy/batch_job.py

@@ -354,23 +354,8 @@ def run_driver():
                 run_driver()
 
     except Exception as e:
-        if isinstance(e, Py4JJavaError):
-            java_exception = e.java_exception
-            if "SparkException" in java_exception.getClass().getName():
-                # TODO: Extract UDF specific stack to show user.
-                message = f"Your openEO batch job failed during Spark execution: {repr_truncate(str(java_exception.getMessage()), width=5000)}"
-                if java_exception.getCause() is not None:
-                    message = f"Your openEO batch job failed during Spark execution: {repr_truncate(str(java_exception.getCause().getMessage()), width=5000)}"
-            else:
-                message = f"Your openEO batch job failed: {repr_truncate(str(java_exception.getMessage()), width=1000)}"
-        else:
-            message = f"Your openEO batch job failed: {repr_truncate(e, width=1000)}"
-
-        if "Container killed on request. Exit code is 143" in str(e):
-            message = "Your batch job failed because workers used too much Python memory. The same task was attempted multiple times. Consider increasing executor-memoryOverhead or contact the developers to investigate."
-
-        if message is not None:
-            user_facing_logger.exception(message)
+        message = GeoPySparkBackendImplementation.summarize_exception(e)
+        user_facing_logger.exception("OpenEO batch job failed: " + message)
 
         raise
 

scripts/batch_job_log4j.properties

@@ -13,3 +13,6 @@ log4j.logger.org.apache.spark=ERROR, openeo, userlogs
 log4j.logger.org.apache.spark.storage.DiskBlockObjectWriter=OFF, openeo, userlogs
 log4j.logger.org.apache.zookeeper=WARN, openeo, userlogs
 log4j.logger.org.apache.curator=WARN, openeo, userlogs
+log4j.logger.org.apache.spark.executor.Executor=OFF
+# TaskSetManager will give a warning when a task fails the first time, and an error when it ran out of retries:
+log4j.logger.org.apache.spark.scheduler.TaskSetManager=WARN

tests/test_backend.py

@@ -5,7 +5,7 @@
 from openeo_driver.delayed_vector import DelayedVector
 from openeo_driver.utils import EvalEnv
 
-from openeogeotrellis.backend import GpsBatchJobs, GpsProcessing
+from openeogeotrellis.backend import GpsBatchJobs, GpsProcessing, GeoPySparkBackendImplementation
 
 def test_extract_application_id():
     yarn_log = """
@@ -217,3 +217,91 @@ def test_extra_validation_layer_too_large_geometrycollection(backend_implementat
     errors = list(processing.extra_validation(pg, env, None, env_source_constraints))
     assert len(errors) == 1
     assert errors[0]['code'] == "LayerTooLarge"
+
+
+def test_extract_udf_stacktrace_1():
+    summarized = GeoPySparkBackendImplementation.extract_udf_stacktrace("""
+    Traceback (most recent call last):
+ File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/worker.py", line 619, in main
+ process()
+ File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/worker.py", line 611, in process
+ serializer.dump_stream(out_iter, outfile)
+ File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/serializers.py", line 132, in dump_stream
+ for obj in iterator:
+ File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/util.py", line 74, in wrapper
+ return f(*args, **kwargs)
+ File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/utils.py", line 52, in memory_logging_wrapper
+ return function(*args, **kwargs)
+ File "/opt/venv/lib64/python3.8/site-packages/epsel.py", line 44, in wrapper
+ return _FUNCTION_POINTERS[key](*args, **kwargs)
+ File "/opt/venv/lib64/python3.8/site-packages/epsel.py", line 37, in first_time
+ return f(*args, **kwargs)
+ File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/geopysparkdatacube.py", line 701, in tile_function
+ result_data = run_udf_code(code=udf_code, data=data)
+ File "/opt/venv/lib64/python3.8/site-packages/openeo/udf/run_code.py", line 180, in run_udf_code
+ func(data)
+ File "<string>", line 8, in transform
+ File "<string>", line 7, in function_in_transform
+ File "<string>", line 4, in function_in_root
+Exception: This error message should be visible to user
+""")
+    assert summarized == """ File "<string>", line 8, in transform
+ File "<string>", line 7, in function_in_transform
+ File "<string>", line 4, in function_in_root
+Exception: This error message should be visible to user"""
+
+
+def test_extract_udf_stacktrace_2():
+    summarized = GeoPySparkBackendImplementation.extract_udf_stacktrace("""Traceback (most recent call last):
+  File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/worker.py", line 619, in main
+    process()
+  File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/worker.py", line 611, in process
+    serializer.dump_stream(out_iter, outfile)
+  File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/serializers.py", line 132, in dump_stream
+    for obj in iterator:
+  File "/opt/spark3_2_0/python/lib/pyspark.zip/pyspark/util.py", line 74, in wrapper
+    return f(*args, **kwargs)
+  File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/utils.py", line 49, in memory_logging_wrapper
+    return function(*args, **kwargs)
+  File "/opt/venv/lib64/python3.8/site-packages/epsel.py", line 44, in wrapper
+    return _FUNCTION_POINTERS[key](*args, **kwargs)
+  File "/opt/venv/lib64/python3.8/site-packages/epsel.py", line 37, in first_time
+    return f(*args, **kwargs)
+  File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/geopysparkdatacube.py", line 519, in tile_function
+    result_data = run_udf_code(code=udf_code, data=data)
+  File "/opt/venv/lib64/python3.8/site-packages/openeo/udf/run_code.py", line 175, in run_udf_code
+    result_cube = func(data.get_datacube_list()[0], data.user_context)
+  File "<string>", line 156, in apply_datacube
+TypeError: inspect() got multiple values for argument 'data'
+""")
+    assert summarized == """  File "<string>", line 156, in apply_datacube
+TypeError: inspect() got multiple values for argument 'data'"""
+
+
+def test_extract_udf_stacktrace_no_udf():
+    summarized = GeoPySparkBackendImplementation.extract_udf_stacktrace("""Traceback (most recent call last):
+  File "/usr/local/spark/python/lib/pyspark.zip/pyspark/worker.py", line 619, in main
+    process()
+  File "/usr/local/spark/python/lib/pyspark.zip/pyspark/worker.py", line 611, in process
+    serializer.dump_stream(out_iter, outfile)
+  File "/usr/local/spark/python/lib/pyspark.zip/pyspark/serializers.py", line 132, in dump_stream
+    for obj in iterator:
+  File "/usr/local/spark/python/lib/pyspark.zip/pyspark/util.py", line 74, in wrapper
+    return f(*args, **kwargs)
+  File "/opt/openeo/lib/python3.8/site-packages/epsel.py", line 44, in wrapper
+    return _FUNCTION_POINTERS[key](*args, **kwargs)
+  File "/opt/openeo/lib/python3.8/site-packages/epsel.py", line 37, in first_time
+    return f(*args, **kwargs)
+  File "/opt/openeo/lib/python3.8/site-packages/openeo/util.py", line 362, in wrapper
+    return f(*args, **kwargs)
+  File "/opt/openeo/lib/python3.8/site-packages/openeogeotrellis/collections/s1backscatter_orfeo.py", line 794, in process_product
+    dem_dir_context = S1BackscatterOrfeo._get_dem_dir_context(
+  File "/opt/openeo/lib64/python3.8/site-packages/openeogeotrellis/collections/s1backscatter_orfeo.py", line 258, in _get_dem_dir_context
+    dem_dir_context = S1BackscatterOrfeo._creodias_dem_subset_srtm_hgt_unzip(
+  File "/opt/openeo/lib64/python3.8/site-packages/openeogeotrellis/collections/s1backscatter_orfeo.py", line 664, in _creodias_dem_subset_srtm_hgt_unzip
+    with zipfile.ZipFile(zip_filename, 'r') as z:
+  File "/usr/lib64/python3.8/zipfile.py", line 1251, in __init__
+    self.fp = io.open(file, filemode)
+FileNotFoundError: [Errno 2] No such file or directory: '/eodata/auxdata/SRTMGL1/dem/N64E024.SRTMGL1.hgt.zip'
+""")
+    assert summarized is None

tests/test_error.py

@@ -0,0 +1,56 @@
+from geopyspark import TiledRasterLayer, Extent
+from openeo_driver.utils import EvalEnv
+from shapely.geometry import MultiPolygon
+
+from openeogeotrellis.backend import GeoPySparkBackendImplementation
+from openeogeotrellis.geopysparkdatacube import GeopysparkDataCube
+
+
+# Note: Ensure that the python environment has all the required modules installed.
+# Numpy should be installed before Jep for off-heap memory tiles to work!
+#
+# Note: In order to run these tests you need to set several environment variables.
+# If you use the virtual environment venv (with JEP and Numpy installed):
+# 1. LD_LIBRARY_PATH = .../venv/lib/python3.6/site-packages/jep
+#   This will look for the shared library 'jep.so'. This is the compiled C code that binds Java and Python objects.
+
+def test_chunk_polygon_exception(imagecollection_with_two_bands_and_three_dates):
+    udf_code = """
+import xarray
+from openeo.udf import XarrayDataCube
+
+def function_in_root():
+    raise Exception("This error message should be visible to user")
+
+def apply_datacube(cube: XarrayDataCube, context: dict) -> XarrayDataCube:
+    function_in_root()
+    array = cube.get_array()
+    return XarrayDataCube(array)
+"""
+    udf_add_to_bands = {
+        "udf_process": {
+            "arguments": {
+                "data": {
+                    "from_argument": "dimension_data"
+                },
+                "udf": udf_code
+            },
+            "process_id": "run_udf",
+            "result": True
+        },
+    }
+    env = EvalEnv()
+
+    polygon1 = Extent(0.0, 0.0, 4.0, 4.0).to_polygon
+    chunks = MultiPolygon([polygon1])
+    cube: GeopysparkDataCube = imagecollection_with_two_bands_and_three_dates
+    try:
+        result_cube: GeopysparkDataCube = cube.chunk_polygon(udf_add_to_bands, chunks=chunks, mask_value=None, env=env)
+        result_layer: TiledRasterLayer = result_cube.pyramid.levels[0]
+        result_layer.to_numpy_rdd().collect()
+    except Exception as e:
+        error_summary = GeoPySparkBackendImplementation.summarize_exception_static(e)
+        print(error_summary.summary)
+        assert "This error message should be visible to user" in error_summary.summary
+    else:
+        raise Exception("There should have been an exception raised in the try clause.")