[GitHub] [pulsar] zbentley edited a comment on issue #10721: (2.7.1) Python client blocks forever in "create_producer" during a broker failure, even if "operation_timeout_seconds" is set

[GitBox <gi...@apache.org>]

zbentley edited a comment on issue #10721:
URL: https://github.com/apache/pulsar/issues/10721#issuecomment-852238692


   Yes, the program is stuck in the `create_producer` call. 
   
   There are no log lines indicating "Unable to create Producer"; I believe it is either getting stuck before that or not flushing its output (while I'm running with PYTHONUNBUFFERED=1, that may not apply to whatever autoflush behavior is configured inside the C++ code). 
   
   The fact that the producers become unstuck when the brokers are unpaused (unstopped) also indicates that there may be a missing connect timeout.
   
   Attached is an entire unedited log segment recorded during the SIGSTOP of a single broker. These logs were generated in a foreground PTY (no kubernetes swallowing/buffering output). The code used is exactly the code copied above with no changes. The logs below showed no activity for more than 5 minutes.
   
   In this log segment, the broker was killed at 15:48:46
   
   [logs_f.txt](https://github.com/apache/pulsar/files/6577581/logs_f.txt)
   
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org