apache / apache/pulsar-client-cpp
[Bug] Keep-alive timeout when send/receive large message
- Lenguaje dominante
- C++
- Estrellas
- 71
- Forks
- 90
- Merge medio
- 2 h 33 min
- PR fusionados (30 d)
- 3
Descripción
### Search before asking
- [X] I searched in the [issues](https://github.com/apache/pulsar-client-cpp/issues) and found nothing similar.
### Version
client 3.2
server 3.2.2
### Minimal reproduce step
Use a producer to send one 300M message to the broker using synchronous mode with 5M bandwidth, just make sure the message to be sent in almost one minute.
BTW, it works fine for small message, for example, messages less than 100M can be successfully sent or consumed within 5M bandwidth
### What did you expect to see?
Send successfully
### What did you see instead?
2024-04-28 16:58:11.157 WARN [40520] D:\a\pulsar-client-cpp\pulsar-client-cpp\lib\ClientConnection:1158 | [10.214.4.3:49457 -> 8.218.192.194:6650] Forcing connection to close after keep-alive timeout
2024-04-28 16:58:11.157 ERROR [40520] D:\a\pulsar-client-cpp\pulsar-client-cpp\lib\ClientConnection:1234 | [10.214.4.3:49457 -> 8.218.192.194:6650] Connection closed with ConnectError
### Anything else?
When I use asynchronous mode in the producer side, this issue is mitigated, while the consumer stuck into it. I find the keepalive interval is hardcoded in ClientConnection.cc and the value is 30 seconds, and I also find it executed by a deadline timer which run in a separated thread, thus I am not sure if this has something to do with the large message sending or receiving
### Are you willing to submit a PR?
- [ ] I'm willing to submit a PR!
Guía de contribución
Línea de trabajo
Empieza en ClientConnection.cc, especialmente en el intervalo de keep-alive de 30 segundos codificado directamente y su temporizador de deadline. Reproduce el problema con client 3.2 y server 3.2.2 usando un mensaje síncrono de 300M con un ancho de banda de 5M; después compara el comportamiento de los productores síncronos y asíncronos y el comportamiento de los consumidores. Se considera terminado cuando el envío o la recepción de gran tamaño se completa sin que el timeout de keep-alive cierre la conexión.
Escrito por el modelo de indexación a partir del texto del issue.
Evaluación
- Stack tecnológico
- cpp
- Área
- distributed-systems, networking
- Tipo de issue
- Error
- Dificultad
- 4/5
- Tiempo estimado
- 3-5 días
- Estado de actividad
- Estancado
- Claridad
- Bastante claro
- Aptitud para principiantes
- 35/100