diff --git a/tools/pretty_tox.sh b/tools/pretty_tox.sh
new file mode 100755
index 0000000000..55dd140429
--- /dev/null
+++ b/tools/pretty_tox.sh
@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+
+set -o pipefail
+
+TESTRARGS=$1
+python -m neutron.openstack.common.lockutils python setup.py testr --slowest --testr-args="--subunit $TESTRARGS" | $(dirname $0)/subunit-trace.py -f
diff --git a/tools/subunit-trace.py b/tools/subunit-trace.py
new file mode 100755
index 0000000000..73f2f1051c
--- /dev/null
+++ b/tools/subunit-trace.py
@@ -0,0 +1,307 @@
+#!/usr/bin/env python
+
+# Copyright 2014 Hewlett-Packard Development Company, L.P.
+# Copyright 2014 Samsung Electronics
+# All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"); you may
+# not use this file except in compliance with the License. You may obtain
+# a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
+# License for the specific language governing permissions and limitations
+# under the License.
+
+"""Trace a subunit stream in reasonable detail and high accuracy."""
+
+import argparse
+import functools
+import os
+import re
+import sys
+
+import mimeparse
+import subunit
+import testtools
+
+DAY_SECONDS = 60 * 60 * 24
+FAILS = []
+RESULTS = {}
+
+
+class Starts(testtools.StreamResult):
+
+    def __init__(self, output):
+        super(Starts, self).__init__()
+        self._output = output
+
+    def startTestRun(self):
+        self._neednewline = False
+        self._emitted = set()
+
+    def status(self, test_id=None, test_status=None, test_tags=None,
+               runnable=True, file_name=None, file_bytes=None, eof=False,
+               mime_type=None, route_code=None, timestamp=None):
+        super(Starts, self).status(
+            test_id, test_status,
+            test_tags=test_tags, runnable=runnable, file_name=file_name,
+            file_bytes=file_bytes, eof=eof, mime_type=mime_type,
+            route_code=route_code, timestamp=timestamp)
+        if not test_id:
+            if not file_bytes:
+                return
+            if not mime_type or mime_type == 'test/plain;charset=utf8':
+                mime_type = 'text/plain; charset=utf-8'
+            primary, sub, parameters = mimeparse.parse_mime_type(mime_type)
+            content_type = testtools.content_type.ContentType(
+                primary, sub, parameters)
+            content = testtools.content.Content(
+                content_type, lambda: [file_bytes])
+            text = content.as_text()
+            if text and text[-1] not in '\r\n':
+                self._neednewline = True
+            self._output.write(text)
+        elif test_status == 'inprogress' and test_id not in self._emitted:
+            if self._neednewline:
+                self._neednewline = False
+                self._output.write('\n')
+            worker = ''
+            for tag in test_tags or ():
+                if tag.startswith('worker-'):
+                    worker = '(' + tag[7:] + ') '
+            if timestamp:
+                timestr = timestamp.isoformat()
+            else:
+                timestr = ''
+                self._output.write('%s: %s%s [start]\n' %
+                                   (timestr, worker, test_id))
+            self._emitted.add(test_id)
+
+
+def cleanup_test_name(name, strip_tags=True, strip_scenarios=False):
+    """Clean up the test name for display.
+
+    By default we strip out the tags in the test because they don't help us
+    in identifying the test that is run to it's result.
+
+    Make it possible to strip out the testscenarios information (not to
+    be confused with tempest scenarios) however that's often needed to
+    indentify generated negative tests.
+    """
+    if strip_tags:
+        tags_start = name.find('[')
+        tags_end = name.find(']')
+        if tags_start > 0 and tags_end > tags_start:
+            newname = name[:tags_start]
+            newname += name[tags_end + 1:]
+            name = newname
+
+    if strip_scenarios:
+        tags_start = name.find('(')
+        tags_end = name.find(')')
+        if tags_start > 0 and tags_end > tags_start:
+            newname = name[:tags_start]
+            newname += name[tags_end + 1:]
+            name = newname
+
+    return name
+
+
+def get_duration(timestamps):
+    start, end = timestamps
+    if not start or not end:
+        duration = ''
+    else:
+        delta = end - start
+        duration = '%d.%06ds' % (
+            delta.days * DAY_SECONDS + delta.seconds, delta.microseconds)
+    return duration
+
+
+def find_worker(test):
+    for tag in test['tags']:
+        if tag.startswith('worker-'):
+            return int(tag[7:])
+    return 'NaN'
+
+
+# Print out stdout/stderr if it exists, always
+def print_attachments(stream, test, all_channels=False):
+    """Print out subunit attachments.
+
+    Print out subunit attachments that contain content. This
+    runs in 2 modes, one for successes where we print out just stdout
+    and stderr, and an override that dumps all the attachments.
+    """
+    channels = ('stdout', 'stderr')
+    for name, detail in test['details'].items():
+        # NOTE(sdague): the subunit names are a little crazy, and actually
+        # are in the form pythonlogging:'' (with the colon and quotes)
+        name = name.split(':')[0]
+        if detail.content_type.type == 'test':
+            detail.content_type.type = 'text'
+        if (all_channels or name in channels) and detail.as_text():
+            title = "Captured %s:" % name
+            stream.write("\n%s\n%s\n" % (title, ('~' * len(title))))
+            # indent attachment lines 4 spaces to make them visually
+            # offset
+            for line in detail.as_text().split('\n'):
+                stream.write("    %s\n" % line)
+
+
+def show_outcome(stream, test, print_failures=False, failonly=False):
+    global RESULTS
+    status = test['status']
+    # TODO(sdague): ask lifeless why on this?
+    if status == 'exists':
+        return
+
+    worker = find_worker(test)
+    name = cleanup_test_name(test['id'])
+    duration = get_duration(test['timestamps'])
+
+    if worker not in RESULTS:
+        RESULTS[worker] = []
+    RESULTS[worker].append(test)
+
+    # don't count the end of the return code as a fail
+    if name == 'process-returncode':
+        return
+
+    if status == 'fail':
+        FAILS.append(test)
+        stream.write('{%s} %s [%s] ... FAILED\n' % (
+            worker, name, duration))
+        if not print_failures:
+            print_attachments(stream, test, all_channels=True)
+    elif not failonly:
+        if status == 'success':
+            stream.write('{%s} %s [%s] ... ok\n' % (
+                worker, name, duration))
+            print_attachments(stream, test)
+        elif status == 'skip':
+            stream.write('{%s} %s ... SKIPPED: %s\n' % (
+                worker, name, test['details']['reason'].as_text()))
+        else:
+            stream.write('{%s} %s [%s] ... %s\n' % (
+                worker, name, duration, test['status']))
+            if not print_failures:
+                print_attachments(stream, test, all_channels=True)
+
+    stream.flush()
+
+
+def print_fails(stream):
+    """Print summary failure report.
+
+    Currently unused, however there remains debate on inline vs. at end
+    reporting, so leave the utility function for later use.
+    """
+    if not FAILS:
+        return
+    stream.write("\n==============================\n")
+    stream.write("Failed %s tests - output below:" % len(FAILS))
+    stream.write("\n==============================\n")
+    for f in FAILS:
+        stream.write("\n%s\n" % f['id'])
+        stream.write("%s\n" % ('-' * len(f['id'])))
+        print_attachments(stream, f, all_channels=True)
+    stream.write('\n')
+
+
+def count_tests(key, value):
+    count = 0
+    for k, v in RESULTS.items():
+        for item in v:
+            if key in item:
+                if re.search(value, item[key]):
+                    count += 1
+    return count
+
+
+def run_time():
+    runtime = 0.0
+    for k, v in RESULTS.items():
+        for test in v:
+            runtime += float(get_duration(test['timestamps']).strip('s'))
+    return runtime
+
+
+def worker_stats(worker):
+    tests = RESULTS[worker]
+    num_tests = len(tests)
+    delta = tests[-1]['timestamps'][1] - tests[0]['timestamps'][0]
+    return num_tests, delta
+
+
+def print_summary(stream):
+    stream.write("\n======\nTotals\n======\n")
+    stream.write("Run: %s in %s sec.\n" % (count_tests('status', '.*'),
+                                           run_time()))
+    stream.write(" - Passed: %s\n" % count_tests('status', 'success'))
+    stream.write(" - Skipped: %s\n" % count_tests('status', 'skip'))
+    stream.write(" - Failed: %s\n" % count_tests('status', 'fail'))
+
+    # we could have no results, especially as we filter out the process-codes
+    if RESULTS:
+        stream.write("\n==============\nWorker Balance\n==============\n")
+
+        for w in range(max(RESULTS.keys()) + 1):
+            if w not in RESULTS:
+                stream.write(
+                    " - WARNING: missing Worker %s! "
+                    "Race in testr accounting.\n" % w)
+            else:
+                num, time = worker_stats(w)
+                stream.write(" - Worker %s (%s tests) => %ss\n" %
+                             (w, num, time))
+
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--no-failure-debug', '-n', action='store_true',
+                        dest='print_failures', help='Disable printing failure '
+                        'debug information in realtime')
+    parser.add_argument('--fails', '-f', action='store_true',
+                        dest='post_fails', help='Print failure debug '
+                        'information after the stream is proccesed')
+    parser.add_argument('--failonly', action='store_true',
+                        dest='failonly', help="Don't print success items",
+                        default=(
+                            os.environ.get('TRACE_FAILONLY', False)
+                            is not False))
+    return parser.parse_args()
+
+
+def main():
+    args = parse_args()
+    stream = subunit.ByteStreamToStreamResult(
+        sys.stdin, non_subunit_name='stdout')
+    starts = Starts(sys.stdout)
+    outcomes = testtools.StreamToDict(
+        functools.partial(show_outcome, sys.stdout,
+                          print_failures=args.print_failures,
+                          failonly=args.failonly
+                      ))
+    summary = testtools.StreamSummary()
+    result = testtools.CopyStreamResult([starts, outcomes, summary])
+    result.startTestRun()
+    try:
+        stream.run(result)
+    finally:
+        result.stopTestRun()
+    if count_tests('status', '.*') == 0:
+        print("The test run didn't actually run any tests")
+        return 1
+    if args.post_fails:
+        print_fails(sys.stdout)
+    print_summary(sys.stdout)
+    return (0 if summary.wasSuccessful() else 1)
+
+
+if __name__ == '__main__':
+    sys.exit(main())
diff --git a/tox.ini b/tox.ini
index 442a6348b3..bd5838b042 100644
--- a/tox.ini
+++ b/tox.ini
@@ -12,8 +12,11 @@ usedevelop = True
 install_command = pip install -U {opts} {packages}
 deps = -r{toxinidir}/requirements.txt
        -r{toxinidir}/test-requirements.txt
+whitelist_externals = bash
 commands =
-  python -m neutron.openstack.common.lockutils python setup.py testr --slowest --testr-args='{posargs}'
+  bash tools/pretty_tox.sh '{posargs}'
+# there is also secret magic in pretty_tox.sh which lets you run in a fail only
+# mode. To do this define the TRACE_FAILONLY environmental variable.
 
 [testenv:hashtest]
 # This is the same as default environment, but with a random PYTHONHASHSEED.