<div dir="ltr"><div>Hello list, we faced with some gen_tcp related problems after switch from erlang R15B03 to R16B03-01</div><div><br></div><div>The problem is as following: When server produce data faster then consumer can handle, after </div>
<div>the out server's buffers are full and incoming client's buffers are full gen_tcp:send/2 call on server side blocks forever in erts_internal:port_command/3. After this, even when client consumes all</div><div>
the data and the buffers are empty server process remains to be suspended in that call</div><div><br></div><div>This problem does not occur always, but quite often.</div><div><br></div><div>Some details on implementation are below, I also shrink the example to </div>
<div>this small app so you can check the code: <a href="https://github.com/define-null/tcp_failing_ex">https://github.com/define-null/tcp_failing_ex</a></div><div><br></div><div>Server is implemented in such a way, that it listen on 8899 port, then when client connect to it spawn main srv process and plenty of workers, which start to write to this port after client send some special msg. The main process is responsible for commands from the client and send responses via gen_tcp:send/2, while workers try to write some stream data to the client and use erang:port_command with no-suspend. So workers send only up-to-date data, dropping any in case client is slow. </div>
<div><br></div><div>The behaviour which we see is as following:</div><div>At first phase producer fills OS and erlang driver's buffers. Consumer read data as it arrives and server drop data which it cannot send. So we see buffer size growing on both side out queue of the server and in queue of the client respectively</div>
<div><br></div><div>After some moment in time, i guess when the buffers are completely filled, server try respond to</div><div>ping message of the client, using gen_tcp:send/2 call. After that, it blocks there forever, even after client consumes all the messages. The situation does not change and the srv process remains in the suspended state, while it's incoming buffer begins to grow when client send more ping messages.</div>
<div><br></div><div>Below is the output on the system with two slow clients, where for the first client server's process is already blocked in gen_tcp:send/2 call, while the second is served well.</div><div><br></div>
<div>Every 2.0s: netstat -al | grep 8899 Tue Jun 24 16:34:51 2014</div><div><br></div><div>tcp4 36 0 localhost.8899 localhost.63263 ESTABLISHED</div>
<div>tcp4 0 0 localhost.63263 localhost.8899 ESTABLISHED</div><div>tcp4 0 130990 localhost.8899 localhost.63257 ESTABLISHED</div><div>tcp4 619190 0 localhost.63257 localhost.8899 ESTABLISHED</div>
<div>tcp4 0 0 *.8899 *.* LISTEN</div><div><br></div><div>This is the output for the client process from github example, where we see that</div><div>after send operation (ping msg) no incoming msg come any more.</div>
<div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div>
<div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,28}}: msg</div><div>{{2014,6,24},{16,33,48}} before send</div><div>{{2014,6,24},{16,33,48}} after send ok</div>
<div>{{2014,6,24},{16,34,9}} before send</div><div>{{2014,6,24},{16,34,9}} after send ok</div><div>{{2014,6,24},{16,34,30}} before send</div><div>{{2014,6,24},{16,34,30}} after send ok</div><div>{{2014,6,24},{16,34,51}} before send</div>
<div>{{2014,6,24},{16,34,51}} after send ok</div><div>{{2014,6,24},{16,35,12}} before send</div><div>....</div><div><br></div><div>Server blocked process output:</div><div><br></div><div>{{2014,6,24},{16,33,21}}: <0.95.0> ping</div>
<div>{{2014,6,24},{16,33,21}} bsend: <0.95.0></div><div>{{2014,6,24},{16,33,21}} asend: <0.95.0> ok</div><div>{{2014,6,24},{16,33,48}}: <0.95.0> ping</div><div>{{2014,6,24},{16,33,48}} bsend: <0.95.0></div>
<div>%% (no asend message after it)</div><div><br></div><div>(<a href="mailto:tcp_failing_node@localdomain.me">tcp_failing_node@localdomain.me</a>)1> erlang:process_info(pid(0,95,0)).</div><div>[{current_function,{erts_internal,port_command,3}},</div>
<div> {initial_call,{proc_lib,init_p,5}},</div><div><br></div><div>Bug is not always reproducible, but occurs quite often. The problem is that even</div><div>after server's out buffers are empty data does not arrive to the client, and incoming buffer grow</div>
<div>as client send ping messages to the server. (So erlang:port_command/3 with no-suspend always return false</div><div>when another main server process for this connection is suspended in gen_tcp:send/2)</div><div><br></div>
<div>And then it's getting only worse as already mentioned</div><div><br></div><div>Every 2.0s: netstat -al | grep 8899 Tue Jun 24 16:56:59 2014</div>
<div><br></div><div>tcp4 804 0 localhost.8899 localhost.63263 ESTABLISHED</div><div>tcp4 0 0 localhost.63263 localhost.8899 ESTABLISHED</div><div>tcp4 0 0 localhost.8899 localhost.63257 ESTABLISHED</div>
<div>tcp4 0 0 localhost.63257 localhost.8899 ESTABLISHED</div><div>tcp4 0 0 *.8899 *.* LISTEN</div><div><br></div><div>We faced with this after switching to R16B03 from R15B03, I know there were some changes in port_command handling, i guess why we got such behaviour?</div>
</div>