bbrowning

require 'benchmark'
require 'json/ext'

NUM_THREADS = 40
NUM_PARSES = 20_000

def run_test(json)
  puts Benchmark.measure {
  threads = NUM_THREADS.times.map do
    Thread.new do

bbrowning

require 'benchmark'
require 'json/ext'
 
NUM_THREADS = 40
NUM_PARSES = 20_000
 
def run_test(json)
puts Benchmark.measure {
threads = NUM_THREADS.times.map do
Thread.new do

bbrowning

I've added more comments to this code to make it more clear what
everything does.

ActiveSupport::Testing::Isolation runs its tests in parallel, and
JRuby has real multithreading. So, what happens under JRuby is
the first test executes, sets the environment variables, and
spawns a child process. In parallel, a second test executes, sets
the environment variables to different values, and spawns another
child process. It's this setting of environment variables that is
a problem - the second test is being run in parallel and sets the

bbrowning

This is in reference to https://plus.google.com/105596541985629444566/posts/27y819XoT2V
to show simple steps I took to try and reproduce the described issue based upon the
information available. This is very simple benchmarking with an almost vanilla Rails 4
application using sqlite as the database so is not useful other than to demonstrate the
out of the box performance with TorqueBox and Trinidad under these situations is very
close.

bbrowning

diff --git a/servlet/src/test/java/io/undertow/servlet/test/streams/ServletInputStreamTestCase.java b/servlet/src/test/java/io/undertow/servlet/test/streams/ServletInputStreamTestCase.java
index fb00498..d423189 100644
--- a/servlet/src/test/java/io/undertow/servlet/test/streams/ServletInputStreamTestCase.java
+++ b/servlet/src/test/java/io/undertow/servlet/test/streams/ServletInputStreamTestCase.java
@@ -95,6 +95,16 @@ public class ServletInputStreamTestCase {
         //}
     }
 
+    @Test
+    public void testAsyncServletInputStreamWithEmptyRequestBody() {

bbrowning

diff --git a/core/src/main/java/io/undertow/conduits/ChunkedStreamSinkConduit.java b/core/src/main/java/io/undertow/conduits/ChunkedStreamSinkConduit.java
index 1e70155..b5fb6b4 100644
--- a/core/src/main/java/io/undertow/conduits/ChunkedStreamSinkConduit.java
+++ b/core/src/main/java/io/undertow/conduits/ChunkedStreamSinkConduit.java
@@ -82,7 +82,8 @@ public class ChunkedStreamSinkConduit extends AbstractStreamSinkConduit<StreamSi
     private int state;
     private int chunkleft = 0;
 
-    private final ByteBuffer chunkingBuffer = ByteBuffer.allocate(14); //14 is the most
+    private final ByteBuffer chunkingBuffer = ByteBuffer.allocate(12); //12 is the most

bbrowning

# Jubille started with:
#    jubilee -p 8080 -e production
#
# Note that the Jubilee console outputs quite a few errors like below during
# the test:
#
# Jan 03, 2014 2:15:31 PM org.vertx.java.core.logging.impl.JULLogDelegate error
# SEVERE: Unhandled exception
# java.nio.channels.ClosedChannelException

bbrowning

run lambda { |env|
  [200, { 'Content-Type' => 'text/plain' }, ["hello world\n"]]
}

bbrowning

source 'https://rubygems.org'

gem 'jruby-openssl', '0.9.3'

bbrowning

1. Make sure your Gemfile.lock is not committed to your OpenShift git repository.

2. Copy the pre_build file to .openshift/action_hooks/pre_build

3. Add .openshift/action_hooks/pre_build to git, set the executable bit, and push your changes

   git add .openshift/action_hooks/pre_build git update-index --chmod=+x .openshift/action_hooks/pre_build git commit -m "Add bundle_install in pre_build hook" git push