[
https://issues.apache.org/jira/browse/THRIFT-4858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16934230#comment-16934230
]
Qinghui Xu commented on THRIFT-4858:
------------------------------------
[~mczajkow] To summarize, the error messages you see are not something
critical. Those errors happens when a worker thread (under the hood, it's
TProcessor which is doing the work) tries to read from a closed InputStream,
and it is most likely because the underlying connection is closed by its peer
(client). This is also what [~osayankin] witnessed.
This error is late caught by the worker thread and logged as an ERROR message,
so nothing harmful in the end. It's just the error message is meaningless and
makes noises for people maintaining the service.
I propose a PR to address this by giving an explicit message to the error to
avoid confusion.
> Java TThreadPoolServer: confusing error message on closed socket
> ----------------------------------------------------------------
>
> Key: THRIFT-4858
> URL: https://issues.apache.org/jira/browse/THRIFT-4858
> 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 TBinaryProtocol.Factory());}}
> {{ 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 wrong.
>
> 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 properly:
> {{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
> org.apache.thrift.transport.TIOStreamTransport.read(TIOStreamTransport.java:132
> at org.apache.thrift.transport.TTransport.readAll(TTransport.java:86)
> at
> org.apache.thrift.transport.TFramedTransport.readFrame(TFramedTransport.java:132)
> at
> org.apache.thrift.transport.TFramedTransport.read(TFramedTransport.java:100)
> at org.apache.thrift.transport.TTransport.readAll(TTransport.java:86)
> at
> org.apache.thrift.protocol.TBinaryProtocol.readAll(TBinaryProtocol.java:425)
> at
> org.apache.thrift.protocol.TBinaryProtocol.readI32(TBinaryProtocol.java:321)
> at
> org.apache.thrift.protocol.TBinaryProtocol.readMessageBegin(TBinaryProtocol.java:225)
> at org.apache.thrift.TBaseProcessor.process(TBaseProcessor.java:33)
> at
> org.apache.thrift.server.TThreadPoolServer$WorkerProcess.run(TThreadPoolServer.java:318)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> 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 line:
> {{ProcessFunction fn = processMap.get(msg.name);}}
> 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 0.9.3.1. 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
(v8.3.4#803005)
