thrift-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Qinghui Xu (Jira)" <>
Subject [jira] [Commented] (THRIFT-4858) Java TThreadPoolServer: confusing error message on closed socket
Date Fri, 20 Sep 2019 09:13:00 GMT


Qinghui Xu commented on THRIFT-4858:


I don't think it's related to THRIFT-3769 , neither will fix of THRIFT-4805 solve this problem.
In fact, the problem reported from THRIFT-4805 is some excessive error messages when the
server socket tries to accept incoming connections (especially because it's `TSaslServerTransport`
and it expects sasl messages). On the other hand, this ticket is about the worker threads
raising some meaningless / incomprehensible errors when the processor tries to read from a
closed socket.

So could you reopen it again while waiting for the fix?

> Java TThreadPoolServer: confusing error message on closed socket
> ----------------------------------------------------------------
>                 Key: THRIFT-4858
>                 URL:
>             Project: Thrift
>          Issue Type: Bug
>          Components: Java - Library
>    Affects Versions: 0.12.0
>         Environment: Here is some information on my environment and code that calls Thrift.
> Environment: Red Hat Linux 7.6, Java 1.8, Thrift 0.12.0. Both my client and server are
in Java. However, I have also seen this when a Python client calls the Java server.
> Here is what I do to set up the server, my service Iface name is "AlgorithmRepository"
> {{TServerTransport transport = new TServerSocket(10202);}}
> {{ TProcessor processor = new AlgorithmRepository.Processor<>(this);}}
> {{ Args args = new TThreadPoolServer.Args(transport).processor(processor).inputTransportFactory(new
TFramedTransport.Factory()).outputTransportFactory(new TFramedTransport.Factory()).protocolFactory(new
> {{ final TServer server = new TThreadPoolServer(args);}}
> {{ new Thread(server::serve).start();}}
> That last line corresponds to the Thread you see at the bottom of the exception.
> It is worth noting we did also set the Input and Output Protocol Factories too and it
had no affect on the result.
> We saw on stackoverflow that if you don't have the same Transport type, it will fail
to function. Here is client code that sets this up:
> {{TTransport transport = new TFramedTransport(new TSocket(host, port, socketTimeoutMillis));}}
> I'd be happy to share more client code if you felt it was something the client is doing
>            Reporter: Michael Czajkowski
>            Assignee: Aki Sukegawa
>            Priority: Major
>             Fix For: 0.13.0
>          Time Spent: 10m
>  Remaining Estimate: 0h
> I've spotted an error in using java libthrift that I'd like to call to your attention.
It is in version 0.12.0, but I see it in other versions. I will comment about the other versions
at the end, please read - it may be very *important* to you.
> I create a TThreadPoolServer, and when my client calls a method, the server spawns off
a worker thread and when it gets to this line (in TThreadPoolServer):
> {{if(stopped_ || !processor.process(inputProtocol, outputProtocol)) {}}
> {{break;}}
>  }
> It will call a TBaseProcessor's process method, which can only return either True or
throw an Exception. In my case, it gets to the very bottom and calls the ProcessFunction fn
> {{fn.process(msg.seqid, in, out, iface);}}
> Followed by returning true. However, since the TThreadPoolServer will not "break" out
of the while(true) loop it is in, it will attempt to call the TBaseProcessor's process method
again and this time around I get an exception when this line is called on TBaseProcessor's
process method:
> {{TMessage msg = in.readMessageBegin();}}
> Exception is:
> 17844 [pool-6-thread-5] ERROR org.apache.thrift.server.TThreadPoolServer - Thrift error
occurred during processing of message.
> {{org.apache.thrift.transport.TTransportException}}
>  at
>  at org.apache.thrift.transport.TTransport.readAll(
>  at org.apache.thrift.transport.TFramedTransport.readFrame(
>  at
>  at org.apache.thrift.transport.TTransport.readAll(
>  at org.apache.thrift.protocol.TBinaryProtocol.readAll(
>  at org.apache.thrift.protocol.TBinaryProtocol.readI32(
>  at org.apache.thrift.protocol.TBinaryProtocol.readMessageBegin(
>  at org.apache.thrift.TBaseProcessor.process(
>  at org.apache.thrift.server.TThreadPoolServer$
>  at java.util.concurrent.ThreadPoolExecutor.runWorker(
>  at java.util.concurrent.ThreadPoolExecutor$
>  at
> I surmise because my TProtocol in has already sent the bytes (I am using a TBinaryProtocol)
to the process function, there is nothing left to read and thus we never make it to the next
> {{ProcessFunction fn = processMap.get(;}}
> For giggles, I decided to remove the ! from the logic, and sure enough it gets by but
it appears to perhaps cause other problems when a client calls a method with zero parameters
(e.g. "ping()").
> *Other versions:* So far as I can tell this exact same code set up is in java libthrift
0.9.0, 0.9.1, and However, my Java code works just fine in those. I only discovered
this problem when I went to upgrade my code to Thrift 0.12.0. I honestly can't make sense
of why that would be OK in those older versions but, I've been blissfully unaware of this
problem until the upgrade. It is possible I need to now do something that I did not before.
That is because before, I used TThreadPoolServer with 0.9.x as I want to do now in 0.12.x.
I figured perhaps something else is happening in Thrift with my TThreadPoolServer. Finally,
I did see the problem in 0.12.1 and did not test it, though it appears changed in the master
branch - we tried it and it still has errors.

This message was sent by Atlassian Jira

View raw message