[BEAM-3757] Shuffle read failed using python 2.2.0

• Type: Bug
• Status: Resolved
• Priority: P2
• Resolution: Fixed
• Affects Version/s: 2.2.0
• Fix Version/s: Not applicable
• Component/s: runner-dataflow
• Environment: gcp, macos
• Assignee: Unassigned
• Reporter: Jonathan Delfour
• Created: 27/Feb/18 21:02
• Updated: 16/May/20 13:56
• Resolved: 15/Mar/18 23:07

Description

Hi,

First issue is that the beam 2.3.0 python SDK is apparently not working on GCP. It gets stuck: 

Workflow failed. Causes: (bf832d44290fbf41): The Dataflow appears to be stuck. You can get help with Cloud Dataflow at https://cloud.google.com/dataflow/support. 

I tried two times.

Reverting back to 2.2.0: it usually works but today, after > 1 hour of processing, and 30 workers used, I get a failure with these in the logs:

Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/dataflow_worker/batchworker.py", line 582, in do_work
    work_executor.execute()
  File "/usr/local/lib/python2.7/dist-packages/dataflow_worker/executor.py", line 167, in execute
    op.start()
  File "dataflow_worker/shuffle_operations.py", line 49, in dataflow_worker.shuffle_operations.GroupedShuffleReadOperation.start
    def start(self):
  File "dataflow_worker/shuffle_operations.py", line 50, in dataflow_worker.shuffle_operations.GroupedShuffleReadOperation.start
    with self.scoped_start_state:
  File "dataflow_worker/shuffle_operations.py", line 65, in dataflow_worker.shuffle_operations.GroupedShuffleReadOperation.start
    with self.shuffle_source.reader() as reader:
  File "dataflow_worker/shuffle_operations.py", line 67, in dataflow_worker.shuffle_operations.GroupedShuffleReadOperation.start
    for key_values in reader:
  File "/usr/local/lib/python2.7/dist-packages/dataflow_worker/shuffle.py", line 406, in __iter__
    for entry in entries_iterator:
  File "/usr/local/lib/python2.7/dist-packages/dataflow_worker/shuffle.py", line 248, in next
    return next(self.iterator)
  File "/usr/local/lib/python2.7/dist-packages/dataflow_worker/shuffle.py", line 206, in __iter__
    chunk, next_position = self.reader.Read(start_position, end_position)
  File "third_party/windmill/shuffle/python/shuffle_client.pyx", line 138, in shuffle_client.PyShuffleReader.Read
IOError: Shuffle read failed: INTERNAL: Received RST_STREAM with error code 2  talking to my-dataflow-02271107-756f-harness-2p65:12346

i also get some information message:

Refusing to split <dataflow_worker.shuffle.GroupedShuffleRangeTracker object at 0x7f03a00fe790> at '\x00\x00\x00\t\x1d\x14\x87\xa3\x00\x01': unstarted

For the flow, I am extracting data from BQ, cleaning using pandas, exporting as a csv file, gzipping and uploading the compressed file to a bucket using decompressive transcoding (csv export, gzip compression and upload are in the same 'worker' as they are done in the same beam.DoFn).

PS: i can't find a reasonable way to export the logs from GCP but i can privately send the log file i have of the run on my machine (the log of the pipeline)