BewareMyPower commented on issue #10721:
URL: https://github.com/apache/pulsar/issues/10721#issuecomment-850765134


   First I noticed there's a line in your log:
   
   ```
   2021-05-28 19:57:42.121 WARN  [140473932080896[] PartitionedProducerImpl:369 
| Failed to getPartitionMetadata: TimeOut
   ```
   
   It could only be caused by operation timeout so there's nothing wrong with 
`operation_timeout_seconds`.
   
   >  Since I'm running on Kubernetes, a random broker would have been selected 
for topic lookup. I suspect that my blocked clients were connected to an 
unavailable broker
   
   I'm afraid it's not the cause because if the broker is not available, the 
connection could not be established, right? No topic lookup request could be 
sent if the connection was not established. In this case the error logs should 
be
   
   ```
    ClientImpl:<line-no> | Error Checking/Getting Partition Metadata while 
creating producer on <topic-name>
   ```
   
   > In fairly short order, that code stopped emitting "Sent: " updates, and 
printed Reconnecting to producer, at which point it got stuck.
   
   So the `Reconnecting to producer` log only occurred once? Then your program 
must stuck at `create_producer` and it's not caused by topic lookup. When a 
producer to a partitioned topic with N partitions is created, N internal 
producers will be created. However, only after all producers are created would 
`create_producer` be completed normally or exceptionally.
   
   For example, the logs should have N following lines if all internal 
producers are created successfully.
   
   ```
   ProducerImpl:<line-no> | 
[persistent://public/default/<topic>-partition-<partition-no>, ] Created 
producer on broker
   ```
   
   I noticed your log is not completed, could you check how many internal 
producers were created after `Reconnecting to producer`? And is there any log 
like `Unable to create Producer for partition`?
   
   


-- 
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:
[email protected]


Reply via email to