Fix last_devices to be quieter, and improve device affinity.

build/android/pylib/perf/setup.py

 # Copyright 2013 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 """Generates test runner factory and tests for performance tests."""
 
 import json
 import fnmatch
+import hashlib
 import logging
 import os
 import shutil
 
 from pylib import android_commands
 from pylib import constants
 from pylib import forwarder
 from pylib.device import device_list
 from pylib.perf import test_runner
 from pylib.utils import test_environment
 
 
-def _GetAllDevices():
-  devices_path = os.path.join(os.environ.get('CHROMIUM_OUT_DIR', 'out'),
-                              device_list.LAST_DEVICES_FILENAME)
-  try:
-    devices = device_list.GetPersistentDeviceList(devices_path)
-  except IOError as e:
-    logging.error('Unable to find %s [%s]', devices_path, e)
-    devices = android_commands.GetAttachedDevices()

[jbudorick, 2015/05/23 01:06:49]

How is this still around...

[luqui, 2015/05/27 20:01:12]

Looks like I forgot to sync before starting work on this :-(

-  return sorted(devices)
-
-
 def _GetStepsDictFromSingleStep(test_options):
   # Running a single command, build the tests structure.
   steps_dict = {
     'version': 1,
     'steps': {
         'single_step': {
           'device_affinity': 0,
           'cmd': test_options.single_step
         },
     }
   }
   return steps_dict
 
 
 def _GetStepsDict(test_options):
   if test_options.single_step:
     return _GetStepsDictFromSingleStep(test_options)
   if test_options.steps:
     with file(test_options.steps, 'r') as f:
       steps = json.load(f)
 
       # Already using the new format.
       assert steps['version'] == 1
       return steps
 
 
+def DistributeAffinities(devices):
+  """Create a mapping from device to affinity using a round-robin hash scheme.
+  This is a way to make sure that affinity -> device mapping is relatively

[jbudorick, 2015/05/23 01:06:49]

nit: line break after the first line of the docstring

also, this should include Args: and Returns:

[luqui, 2015/05/27 20:01:12]

Done.

+  stable while still maximizing the use of resources."""

[jbudorick, 2015/05/23 01:06:49]

I'm not sure this is stable if you lose one of the lexicographically early
devices. That's why the persistent device list exists.

Also, this docstring should better explain what's going on.

[luqui, 2015/05/27 20:01:12]

IIUC the persistent device list basically messes up everything if a new device
is added.

You gave me an excuse to do some math, so I added a bunch of docs explaining the
scheme and analyzing its stability.

+  if not devices:
+    return {}
+
+  device_cycle = []
+  while len(device_cycle) < test_runner.NUM_DEVICE_AFFINITIES:
+    device_cycle.extend(devices)
+  device_cycle = device_cycle[:test_runner.NUM_DEVICE_AFFINITIES]
+
+  affinity_to_device = {}
+
+  # Hash and rehash each device ID until we get a unique affinity.
+  for device in device_cycle:
+    cur_hash = device
+    while True:
+      cur_hash = hashlib.md5(cur_hash).hexdigest()
+      affinity = int(int(cur_hash, 16) % test_runner.NUM_DEVICE_AFFINITIES)
+      if affinity not in affinity_to_device:
+        affinity_to_device[affinity] = device
+        break
+
+  # Invert the dictionary we just built.
+  device_to_affinity = {}
+  for k, v in affinity_to_device.iteritems():
+    if v not in device_to_affinity:
+      device_to_affinity[v] = []
+    device_to_affinity[v].append(k)
+
+  return device_to_affinity
+
 def Setup(test_options):
   """Create and return the test runner factory and tests.
 
   Args:
     test_options: A PerformanceOptions object.
 
   Returns:
     A tuple of (TestRunnerFactory, tests, devices).
   """
   # TODO(bulach): remove this once the bot side lands. BUG=318369
   constants.SetBuildType('Release')
   if os.path.exists(constants.PERF_OUTPUT_DIR):
     shutil.rmtree(constants.PERF_OUTPUT_DIR)
   os.makedirs(constants.PERF_OUTPUT_DIR)
 
   # Before running the tests, kill any leftover server.
   test_environment.CleanupLeftoverProcesses()
   forwarder.Forwarder.UseMultiprocessing()
 
-  # We want to keep device affinity, so return all devices ever seen.
-  all_devices = _GetAllDevices()
+  all_devices = android_commands.GetAttachedDevices()

[jbudorick, 2015/05/23 01:06:49]

device_utils.DeviceUtils.HealthyDevices()

if you want serials,

  [d.adb.GetDeviceSerial()
   for d in device_utils.DeviceUtils.HealthyDevices()]

android_commands is deprecated.

also, you will absolutely want to sort this.

[luqui, 2015/05/27 20:01:12]

Done.

+  affinity_map = DistributeAffinities(all_devices)
 
   steps_dict = _GetStepsDict(test_options)
   sorted_step_names = sorted(steps_dict['steps'].keys())
 
   if test_options.test_filter:
     sorted_step_names = fnmatch.filter(sorted_step_names,
                                        test_options.test_filter)
 
   flaky_steps = []
   if test_options.flaky_steps:
     with file(test_options.flaky_steps, 'r') as f:
       flaky_steps = json.load(f)
 
-  def TestRunnerFactory(device, shard_index):
+  def TestRunnerFactory(device, _):
+    affinities = affinity_map[device]
     return test_runner.TestRunner(
-        test_options, device, shard_index, len(all_devices),
-        steps_dict, flaky_steps)
+        test_options, device, affinities, steps_dict, flaky_steps)
 
   return (TestRunnerFactory, sorted_step_names, all_devices)