[FLINK-28114] The path of the Python client interpreter could not point to an archive file in distributed file system

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.15.1, 1.16.0
• Component/s: API / Python
• Assignee: Dian Fu
• Reporter: Dian Fu
• Created: 17/Jun/22 14:57
• Updated: 20/Jun/22 01:08
• Resolved: 20/Jun/22 01:08

Description

See https://github.com/apache/flink/blob/master/flink-python/src/main/java/org/apache/flink/client/python/PythonEnvUtils.java#L178 for more details about this limitation.

Users could execute PyFlink jobs in YARN application mode as following:

./bin/flink run-application -t yarn-application \
      -Djobmanager.memory.process.size=1024m \
      -Dtaskmanager.memory.process.size=1024m \
      -Dyarn.application.name=<ApplicationName> \
      -Dyarn.ship-files=/path/to/shipfiles \
      -pyarch shipfiles/venv.zip \
      -pyclientexec venv.zip/venv/bin/python3 \
      -pyexec venv.zip/venv/bin/python3 \
      -py shipfiles/word_count.py

In the above case, venv.zip will be distributed to the TMs via Flink blob server. However, blob server doesn't support files with size exceeding of 2GB. See https://github.com/apache/flink/blob/ea52732dc48a4f1c5be0925890cd8aa1ea2a11ed/flink-runtime/src/main/java/org/apache/flink/runtime/blob/BlobServerConnection.java#L223 for more details. This is very serious problem as Python users usually tend to install a lot Python libraries inside the venv.zip and some Python libraries are very large.