[Java][FlightRPC] Java arrow flight server stuck in reader.getDescriptor()
- Linguagem predominante
- Java
- Estrelas
- 94
- Forks
- 152
- Merge médio
- 3d 16h
- PRs com merge (30d)
- 11
Descrição
### Describe the usage question you have. Please include as many useful details as possible.
Platform: Ubuntu 20.04.6 LTS
Arrow version: 15.
Client: C++
Server: Java
I have a Java arrow flight server that uses doExchange to read the context, my server code is as follows, I have identified in multiple thread environment, in the tests many calls would stuck at the line ```FlightDescriptor descriptor = reader.getDescriptor();``` in the second time(Which means the first call in the same thread usually is not blocked), and which makes the following requests also stucks.
``` public void doExchange(CallContext context, FlightStream reader, ServerStreamListener writer) {
log.info("Start to call doExchange.........");
// Trying to
try (BufferAllocator allocator = allocatorPool.submit(
() -> this.allocator.newChildAllocator("exchange", 0, Long.MAX_VALUE)).get()) {
FlightDescriptor descriptor = reader.getDescriptor();
List path = descriptor.getPath();
String type = path.get(0);
String funcSignature = path.get(1);
...
```
The client code is as follows
```
std::vector> UDFClient::Call(
const std::vector& paths,
std::shared_ptr& batch) const {
// Create a FlightDescriptor using a path
FlightDescriptor descriptor = FlightDescriptor::Path(paths);
auto exchange_result = Client_->DoExchange(descriptor);
if (!exchange_result.ok()) {
throw std::runtime_error("Do exchange descriptor: "
+ exchange_result.status().ToString());
}
auto exchange = std::move(exchange_result.ValueUnsafe());
...
```
I understand ```reader.getDescriptor(); ``` is a blocking call that will use future to wait the descriptor sent. Since doExchange in the server has been called, I assume the client ```auto exchange_result = Client_->DoExchange(descriptor);``` has been received. I can't figure what could make ```reader.getDescriptor();``` stuck, I am reporting this to the community that want to know is this a bug or there is something I did wrong. Thank you
This happened in the multiple threads environment. We also check the lock with jstack that shows ```SettableFuture descriptor``` is the one not set that caused the lock.
```
"pool-1-thread-16" #31 prio=5 os_prio=0 tid=0x00007f6d34009000 nid=0x3a19 waiting on condition [0x00007f6d253ea000]
java.lang.Thread.State: WAITING (parking)
at sun.misc.Unsafe.park(Native Method)
- parking to wait for <0x000000076d596180> (a com.google.common.util.concurrent.SettableFuture)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:175)
at com.google.common.util.concurrent.AbstractFuture.get(AbstractFuture.java:563)
at com.google.common.util.concurrent.AbstractFuture$TrustedFuture.get(AbstractFuture.java:110)
at org.apache.arrow.flight.FlightStream.getDescriptor(FlightStream.java:158)
at com.bytedance.dp.udf.UDFProducer.doExchange(UDFProducer.java:266)
at org.apache.arrow.flight.FlightService.lambda$doExchangeCustom$2(FlightService.java:382)
at org.apache.arrow.flight.FlightService$$Lambda$65/298187580.run(Unknown Source)
at io.grpc.Context$1.run(Context.java:566)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
```
### Component(s)
C++, FlightRPC, Java
Guia de contribuição
Direção de pesquisa
Comece por FlightStream.java:158, onde getDescriptor() aguarda, e trace o caminho Java de doExchange por meio de FlightService e UDFProducer.java:266. Compare esse fluxo com a chamada C++ Client_->DoExchange(descriptor) e com as evidências de jstack fornecidas em uma configuração multithread. O trabalho estará concluído quando for identificado por que o descriptor future permanece não definido e for documentada uma causa reproduzível ou uma resolução confirmada.
Escrita pelo modelo de indexação a partir do texto da issue.
Avaliação
- Stack de tecnologia
- cpp, java
- Domínio
- api
- Tipo de issue
- Bug
- Dificuldade
- 4/5
- Tempo estimado
- 3-5 dias
- Status de atividade
- Estagnada
- Clareza
- Precisa de esclarecimento
- Facilidade para iniciantes
- 35/100