Apache thrift: client timeout issues

Question

I have some Apache Thrift (v.0.6.1) test application with perl-server and php-client.

The behaviour I cannot explain: If we call server-method with invalid argument we see the error in server-output, but php-client stays waiting the response infinitely.

Here are the sources of server:

sub new {
    my $classname = shift;
    my $self      = {};

    return bless($self,$classname);
}

sub DateToTimestamp
{
    my ($self, $date) = @_;
    my $result = CommonAPI::DateToTimestamp($date);
    return $result;
}

eval {
  my $handler       = new RPCHandler;
  my $processor     = new RPCPerformanceTest::RPCPerformanceTestProcessor($handler);
  my $serversocket  = new Thrift::ServerSocket(9091);
  my $forkingserver = new Thrift::ForkingServer($processor, $serversocket);
  print "Starting the server...\n";
  $forkingserver->serve();
  print "done.\n";
}; if ($@) {
  if ($@ =~ m/TException/ and exists $@->{message}) {
    my $message = $@->{message};
    my $code    = $@->{code};
    my $out     = $code . ':' . $message;
    die $out;
  } else {
    die $@;
  }
}

and client:

try {

    $socket = new TSocket($server_host, $server_port);

    $transport = new TBufferedTransport($socket, 1024, 1024);
    $protocol = new TBinaryProtocol($transport);

    $client = new RPCPerformanceTestClient($protocol);
    $transport->open();

    $start = microtime(true);

    $result = $client->DateToTimestamp('071/26/2011 01:23:45');

    var_dump($result);

} catch (Exception $e) {
    echo 'Exception: <b>' . $e->getMessage() . '</b>';
}

Why is this happening? Is it my fault? Is it expected behavour?

Possibly a related (or the same) issue? https://stackoverflow.com/questions/14512489/thrift-tsimpleserver-becomes-unresponsive-after-several-successful-requests — joseAndresGomezTovar, Sep 16 '19 at 07:03

score 6 · Answer 1 · answered Dec 06 '12 at 16:35

6

The Thrift PHP library is a bit broken. You need to manually set the timeouts E.g.

  $socket = new TSocket('host', 9095);
  $socket->setSendTimeout(60000);
  $socket->setRecvTimeout(60000)

answered Dec 06 '12 at 16:35

Cosmin Lehene

558
4
11

it seems that adding timeout didn't work, error is the same! is there any other way to increase timout without changing code, like config file.. – Abdelali AHBIB Aug 13 '14 at 14:28

score 1 · Accepted Answer · answered May 30 '11 at 18:11

1

This happens often with protocols that do not supply message length: a client sends more data then the server expects and waits for the server to receive the data. The server receives some of the data, tries to parse it and fails. Now the server-side of the protocol is in errorneous state. If it continues to read the data, it may block. Most probably, the server-side has sent you some error response and is waiting at the same time for the client to receive the response, but that will never happen too.

This is my guess. The best strategy IMHO is to set a time-out for both client and server sockets.

answered May 30 '11 at 18:11

Dmitry Negoda

2,497
2
16
15

Timeout is set, it is set in `TSocket` transport by thrift developers team. But in some reason - it doesn't being triggered. Anyway, thanks for the response, the question is not actual for me now, so closing with your answer ;-) – zerkms May 30 '11 at 22:13
try to change it in your conf/hbase-site.xml : hbase.client.scanner.timeout.period 60000 more in this link : http://hbase.apache.org/book/config.files.html – Abdelali AHBIB Aug 25 '14 at 14:14

Apache thrift: client timeout issues

2 Answers2

Linked