Skip to content
Snippets Groups Projects
Commit ba3c730e authored by Andrew Or's avatar Andrew Or
Browse files

[SPARK-3140] Clarify confusing PySpark exception message

We read the py4j port from the stdout of the `bin/spark-submit` subprocess. If there is interference in stdout (e.g. a random echo in `spark-submit`), we throw an exception with a warning message. We do not, however, distinguish between this case from the case where no stdout is produced at all.

I wasted a non-trivial amount of time being baffled by this exception in search of places where I print random whitespace (in vain, of course). A clearer exception message that distinguishes between these cases will prevent similar headaches that I have gone through.

Author: Andrew Or <andrewor14@gmail.com>

Closes #2067 from andrewor14/python-exception and squashes the following commits:

742f823 [Andrew Or] Further clarify warning messages
e96a7a0 [Andrew Or] Distinguish between unexpected output and no output at all
parent c9f74395
No related branches found
No related tags found
No related merge requests found
......@@ -54,12 +54,19 @@ def launch_gateway():
gateway_port = proc.stdout.readline()
gateway_port = int(gateway_port)
except ValueError:
# Grab the remaining lines of stdout
(stdout, _) = proc.communicate()
exit_code = proc.poll()
error_msg = "Launching GatewayServer failed"
error_msg += " with exit code %d! " % exit_code if exit_code else "! "
error_msg += "(Warning: unexpected output detected.)\n\n"
error_msg += gateway_port + stdout
error_msg += " with exit code %d!\n" % exit_code if exit_code else "!\n"
error_msg += "Warning: Expected GatewayServer to output a port, but found "
if gateway_port == "" and stdout == "":
error_msg += "no output.\n"
else:
error_msg += "the following:\n\n"
error_msg += "--------------------------------------------------------------\n"
error_msg += gateway_port + stdout
error_msg += "--------------------------------------------------------------\n"
raise Exception(error_msg)
# Create a thread to echo output from the GatewayServer, which is required
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment