Zac Bentley
Zac Bentley
To expand on the design issues: fork safety _in pure Python_ is decently easy to achieve, thanks to the GIL. If a Python program is using multiple Python `threading.Thread`s, and...
@merlimat `pthread_atfork` has a before-fork hook that could be used for this. However, [this stacko](https://stackoverflow.com/questions/21529540) appears to indicate that `notify_fork` may not be sufficient to resolve all issues. I'm not...
A workaround that seemed effective at the application layer was to intentionally leak Client objects into a global in the child processes (an atfork child handler that appends them to...
@BewareMyPower > It's better to show your Python client version and paste your code. See https://github.com/apache/pulsar-client-python/issues/190; that contains Python/client/etc. version. The code I'm using is: ``` from pulsar import Client...
Could you clarify the difference between a setting of 0 and a setting of 1?
@BewareMyPower thanks, can that be added to the client documentation for receiver-queue related flags?
Update (and updated issue title): this only happens with *partitioned* topics. While [I'm not quite sure what receiver queue size does](https://github.com/apache/pulsar/issues/15702) on partitioned topics, I'm pretty sure it shouldn't start...
While I don't know much about this area of the code, that sounds fine to me. Other solutions that may make sense: - Document that a zero queue size with...
This feature would be very useful for numerous use cases of ours.
Could someone please point me at what code I should look at if I wanted to put up a PR to fix this? It's quite a hassle for environments with...