Vincent Tran created IMPALA-11653:
-------------------------------------

             Summary: Identify and time out connections that are not from a 
supported Impala client more eagerly
                 Key: IMPALA-11653
                 URL: https://issues.apache.org/jira/browse/IMPALA-11653
             Project: IMPALA
          Issue Type: Improvement
            Reporter: Vincent Tran


When a tcp client opens a connection to an Impala client interface (hs2 or 
beeswax), the connection is accepted immediately after the 3-way handshake 
(SYN, SYN-ACK, ACK) and is queued forĀ *TAcceptQueueServer::SetupConnection()*.  
However, if the client sends nothing else, the ImpalaServer will block in 
apache::thrift::transport::TSocket::read() until the client sends a RST/FIN or 
until sasl_connect_tcp_timeout_ms elapses (which is by default, 5 minutes).

The connection setup thread stacktrace can be observed below during this period.

{noformat}
(gdb) bt
#0  0x00007f3b972ee20d in poll () from ./lib64/libc.so.6
#1  0x0000000002dcd5bc in apache::thrift::transport::TSocket::read(unsigned 
char*, unsigned int) ()
#2  0x0000000002dd1803 in unsigned int 
apache::thrift::transport::readAll<apache::thrift::transport::TSocket>(apache::thrift::transport::TSocket&,
 unsigned char*, unsigned int) ()
#3  0x0000000001330cc9 in readAll (len=5, buf=0x7f3277ea4f8b "", 
this=<optimized out>) at 
../../../toolchain/toolchain-packages-gcc7.5.0/thrift-0.9.3-p8/include/thrift/transport/TTransport.h:121
#4  apache::thrift::transport::TSaslTransport::receiveSaslMessage 
(this=this@entry=0x278a96b0, status=status@entry=0x7f3277ea500c, 
length=length@entry=0x7f3277ea5008) at TSaslTransport.cpp:259
#5  0x000000000132db14 in 
apache::thrift::transport::TSaslServerTransport::handleSaslStartMessage 
(this=0x278a96b0) at TSaslServerTransport.cpp:95
#6  0x0000000001330e33 in 
apache::thrift::transport::TSaslTransport::doSaslNegotiation (this=0x278a96b0) 
at TSaslTransport.cpp:81
#7  0x000000000132e723 in open (this=0x12e29750) at 
../../../toolchain/toolchain-packages-gcc7.5.0/thrift-0.9.3-p8/include/thrift/transport/TBufferTransports.h:218
#8  apache::thrift::transport::TSaslServerTransport::Factory::getTransport 
(this=0xf825a70, trans=...) at TSaslServerTransport.cpp:173
#9  0x00000000010cd49d in 
apache::thrift::server::TAcceptQueueServer::SetupConnection (this=0x174270c0, 
entry=...) at TAcceptQueueServer.cpp:233
#10 0x00000000010cef4d in operator() (tid=<optimized out>, item=..., 
__closure=<optimized out>) at TAcceptQueueServer.cpp:323
#11 
boost::detail::function::void_function_obj_invoker2<apache::thrift::server::TAcceptQueueServer::serve()::<lambda(int,
 const boost::shared_ptr<apache::thrift::server::TAcceptQueueEntry>&)>, void, 
int, const 
boost::shared_ptr<apache::thrift::server::TAcceptQueueEntry>&>::invoke(boost::detail::function::function_buffer
 &, int, const boost::shared_ptr<apache::thrift::server::TAcceptQueueEntry> &) 
(function_obj_ptr=..., a0=<optimized out>, a1=...)
    at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/function/function_template.hpp:159
#12 0x00000000010d3e59 in operator() (a1=..., a0=1, this=0x7f3279ea9510) at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/function/function_template.hpp:770
#13 
impala::ThreadPool<boost::shared_ptr<apache::thrift::server::TAcceptQueueEntry> 
>::WorkerThread (this=0x7f3279ea94c0, thread_id=1) at ../util/thread-pool.h:166
#14 0x000000000144f8f2 in operator() (this=0x7f3277ea5b40) at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/function/function_template.hpp:770
#15 impala::Thread::SuperviseThread(std::__cxx11::basic_string<char, 
std::char_traits<char>, std::allocator<char> > const&, 
std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > 
const&, boost::function<void ()>, impala::ThreadDebugInfo const*, 
impala::Promise<long, (impala::PromiseMode)0>*) (name=..., category=..., 
functor=..., parent_thread_info=<optimized out>, thread_started=0x7f3279ea9110) 
at thread.cc:360
#16 0x0000000001450d6b in operator()<void (*)(const 
std::__cxx11::basic_string<char>&, const std::__cxx11::basic_string<char>&, 
boost::function<void()>, const impala::ThreadDebugInfo*, impala::Promise<long 
int>*), boost::_bi::list0> (a=<synthetic pointer>,
    f=@0x1417ccf8: 0x144f5f0 
<impala::Thread::SuperviseThread(std::__cxx11::basic_string<char, 
std::char_traits<char>, std::allocator<char> > const&, 
std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > 
const&, boost::function<void ()>, impala::ThreadDebugInfo const*, 
impala::Promise<long, (impala::PromiseMode)0>*)>, this=0x1417cd00) at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/bind/bind.hpp:531
#17 operator() (this=0x1417ccf8) at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/bind/bind.hpp:1222
#18 boost::detail::thread_data<boost::_bi::bind_t<void, void 
(*)(std::__cxx11::basic_string<char, std::char_traits<char>, 
std::allocator<char> > const&, std::__cxx11::basic_string<char, 
std::char_traits<char>, std::allocator<char> > const&, boost::function<void 
()>, impala::ThreadDebugInfo const*, impala::Promise<long, 
(impala::PromiseMode)0>*), 
boost::_bi::list5<boost::_bi::value<std::__cxx11::basic_string<char, 
std::char_traits<char>, std::allocator<char> > >, 
boost::_bi::value<std::__cxx11::basic_string<char, std::char_traits<char>, 
std::allocator<char> > >, boost::_bi::value<boost::function<void ()> >, 
boost::_bi::value<impala::ThreadDebugInfo*>, 
boost::_bi::value<impala::Promise<long, (impala::PromiseMode)0>*> > > >::run() 
(this=0x1417cb40)
    at 
../../../toolchain/toolchain-packages-gcc7.5.0/boost-1.61.0-p2/include/boost/thread/detail/thread.hpp:116
#19 0x0000000001ca17c2 in thread_proxy ()
#20 0x00007f3b9a7f2dd5 in start_thread () from ./lib64/libpthread.so.0
#21 0x00007f3b972f8ead in clone () from ./lib64/libc.so.6
{noformat}


As discussed in IMPALA-7638, we need to distinguish between a client that is 
slow to finish sasl negotiation (due to slow Kerberos negotiation) from a 
client that is never going to do anything beyond the tcp handshake and time 
these out much sooner than the sasl_connect_tcp_timeout_ms duration.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org
For additional commands, e-mail: issues-all-h...@impala.apache.org

Reply via email to