diff options
Diffstat (limited to 'crosperf/machine_manager.py')
| -rw-r--r-- | crosperf/machine_manager.py | 709 |
1 files changed, 709 insertions, 0 deletions
diff --git a/crosperf/machine_manager.py b/crosperf/machine_manager.py new file mode 100644 index 00000000..2fdf141b --- /dev/null +++ b/crosperf/machine_manager.py @@ -0,0 +1,709 @@ +# Copyright (c) 2013 The Chromium OS Authors. All rights reserved. +# Use of this source code is governed by a BSD-style license that can be +# found in the LICENSE file. +"""Machine Manager module.""" + +from __future__ import print_function + +import collections +import file_lock_machine +import hashlib +import image_chromeos +import math +import os.path +import re +import sys +import threading +import time + +import test_flag +from cros_utils import command_executer +from cros_utils import logger + +CHECKSUM_FILE = '/usr/local/osimage_checksum_file' + + +class BadChecksum(Exception): + """Raised if all machines for a label don't have the same checksum.""" + pass + + +class BadChecksumString(Exception): + """Raised if all machines for a label don't have the same checksum string.""" + pass + + +class MissingLocksDirectory(Exception): + """Raised when cannot find/access the machine locks directory.""" + + +class CrosCommandError(Exception): + """Raised when an error occurs running command on DUT.""" + + +class CrosMachine(object): + """The machine class.""" + + def __init__(self, name, chromeos_root, log_level, cmd_exec=None): + self.name = name + self.image = None + # We relate a dut with a label if we reimage the dut using label or we + # detect at the very beginning that the dut is running this label. + self.label = None + self.checksum = None + self.locked = False + self.released_time = time.time() + self.test_run = None + self.chromeos_root = chromeos_root + self.log_level = log_level + self.cpuinfo = None + self.machine_id = None + self.checksum_string = None + self.meminfo = None + self.phys_kbytes = None + self.ce = cmd_exec or command_executer.GetCommandExecuter( + log_level=self.log_level) + self.SetUpChecksumInfo() + + def SetUpChecksumInfo(self): + if not self.IsReachable(): + self.machine_checksum = None + return + self._GetMemoryInfo() + self._GetCPUInfo() + self._ComputeMachineChecksumString() + self._GetMachineID() + self.machine_checksum = self._GetMD5Checksum(self.checksum_string) + self.machine_id_checksum = self._GetMD5Checksum(self.machine_id) + + def IsReachable(self): + command = 'ls' + ret = self.ce.CrosRunCommand(command, + machine=self.name, + chromeos_root=self.chromeos_root) + if ret: + return False + return True + + def _ParseMemoryInfo(self): + line = self.meminfo.splitlines()[0] + usable_kbytes = int(line.split()[1]) + # This code is from src/third_party/test/files/client/bin/base_utils.py + # usable_kbytes is system's usable DRAM in kbytes, + # as reported by memtotal() from device /proc/meminfo memtotal + # after Linux deducts 1.5% to 9.5% for system table overhead + # Undo the unknown actual deduction by rounding up + # to next small multiple of a big power-of-two + # eg 12GB - 5.1% gets rounded back up to 12GB + mindeduct = 0.005 # 0.5 percent + maxdeduct = 0.095 # 9.5 percent + # deduction range 1.5% .. 9.5% supports physical mem sizes + # 6GB .. 12GB in steps of .5GB + # 12GB .. 24GB in steps of 1 GB + # 24GB .. 48GB in steps of 2 GB ... + # Finer granularity in physical mem sizes would require + # tighter spread between min and max possible deductions + + # increase mem size by at least min deduction, without rounding + min_kbytes = int(usable_kbytes / (1.0 - mindeduct)) + # increase mem size further by 2**n rounding, by 0..roundKb or more + round_kbytes = int(usable_kbytes / (1.0 - maxdeduct)) - min_kbytes + # find least binary roundup 2**n that covers worst-cast roundKb + mod2n = 1 << int(math.ceil(math.log(round_kbytes, 2))) + # have round_kbytes <= mod2n < round_kbytes*2 + # round min_kbytes up to next multiple of mod2n + phys_kbytes = min_kbytes + mod2n - 1 + phys_kbytes -= phys_kbytes % mod2n # clear low bits + self.phys_kbytes = phys_kbytes + + def _GetMemoryInfo(self): + #TODO yunlian: when the machine in rebooting, it will not return + #meminfo, the assert does not catch it either + command = 'cat /proc/meminfo' + ret, self.meminfo, _ = self.ce.CrosRunCommandWOutput( + command, + machine=self.name, + chromeos_root=self.chromeos_root) + assert ret == 0, 'Could not get meminfo from machine: %s' % self.name + if ret == 0: + self._ParseMemoryInfo() + + def _GetCPUInfo(self): + command = 'cat /proc/cpuinfo' + ret, self.cpuinfo, _ = self.ce.CrosRunCommandWOutput( + command, + machine=self.name, + chromeos_root=self.chromeos_root) + assert ret == 0, 'Could not get cpuinfo from machine: %s' % self.name + + def _ComputeMachineChecksumString(self): + self.checksum_string = '' + exclude_lines_list = ['MHz', 'BogoMIPS', 'bogomips'] + for line in self.cpuinfo.splitlines(): + if not any(e in line for e in exclude_lines_list): + self.checksum_string += line + self.checksum_string += ' ' + str(self.phys_kbytes) + + def _GetMD5Checksum(self, ss): + if ss: + return hashlib.md5(ss).hexdigest() + else: + return '' + + def _GetMachineID(self): + command = 'dump_vpd_log --full --stdout' + _, if_out, _ = self.ce.CrosRunCommandWOutput( + command, + machine=self.name, + chromeos_root=self.chromeos_root) + b = if_out.splitlines() + a = [l for l in b if 'Product' in l] + if len(a): + self.machine_id = a[0] + return + command = 'ifconfig' + _, if_out, _ = self.ce.CrosRunCommandWOutput( + command, + machine=self.name, + chromeos_root=self.chromeos_root) + b = if_out.splitlines() + a = [l for l in b if 'HWaddr' in l] + if len(a): + self.machine_id = '_'.join(a) + return + a = [l for l in b if 'ether' in l] + if len(a): + self.machine_id = '_'.join(a) + return + assert 0, 'Could not get machine_id from machine: %s' % self.name + + def __str__(self): + l = [] + l.append(self.name) + l.append(str(self.image)) + l.append(str(self.checksum)) + l.append(str(self.locked)) + l.append(str(self.released_time)) + return ', '.join(l) + + +class MachineManager(object): + """Lock, image and unlock machines locally for benchmark runs. + + This class contains methods and calls to lock, unlock and image + machines and distribute machines to each benchmark run. The assumption is + that all of the machines for the experiment have been globally locked + (using an AFE server) in the ExperimentRunner, but the machines still need + to be locally locked/unlocked (allocated to benchmark runs) to prevent + multiple benchmark runs within the same experiment from trying to use the + same machine at the same time. + """ + + def __init__(self, + chromeos_root, + acquire_timeout, + log_level, + locks_dir, + cmd_exec=None, + lgr=None): + self._lock = threading.RLock() + self._all_machines = [] + self._machines = [] + self.image_lock = threading.Lock() + self.num_reimages = 0 + self.chromeos_root = None + self.machine_checksum = {} + self.machine_checksum_string = {} + self.acquire_timeout = acquire_timeout + self.log_level = log_level + self.locks_dir = locks_dir + self.ce = cmd_exec or command_executer.GetCommandExecuter( + log_level=self.log_level) + self.logger = lgr or logger.GetLogger() + + if self.locks_dir and not os.path.isdir(self.locks_dir): + raise MissingLocksDirectory('Cannot access locks directory: %s' % + self.locks_dir) + + self._initialized_machines = [] + self.chromeos_root = chromeos_root + + def RemoveNonLockedMachines(self, locked_machines): + for m in self._all_machines: + if m.name not in locked_machines: + self._all_machines.remove(m) + + for m in self._machines: + if m.name not in locked_machines: + self._machines.remove(m) + + def GetChromeVersion(self, machine): + """Get the version of Chrome running on the DUT.""" + + cmd = '/opt/google/chrome/chrome --version' + ret, version, _ = self.ce.CrosRunCommandWOutput( + cmd, + machine=machine.name, + chromeos_root=self.chromeos_root) + if ret != 0: + raise CrosCommandError("Couldn't get Chrome version from %s." % + machine.name) + + if ret != 0: + version = '' + return version.rstrip() + + def ImageMachine(self, machine, label): + checksum = label.checksum + + if checksum and (machine.checksum == checksum): + return + chromeos_root = label.chromeos_root + if not chromeos_root: + chromeos_root = self.chromeos_root + image_chromeos_args = [image_chromeos.__file__, '--no_lock', + '--chromeos_root=%s' % chromeos_root, + '--image=%s' % label.chromeos_image, + '--image_args=%s' % label.image_args, '--remote=%s' % + machine.name, '--logging_level=%s' % self.log_level] + if label.board: + image_chromeos_args.append('--board=%s' % label.board) + + # Currently can't image two machines at once. + # So have to serialized on this lock. + save_ce_log_level = self.ce.log_level + if self.log_level != 'verbose': + self.ce.log_level = 'average' + + with self.image_lock: + if self.log_level != 'verbose': + self.logger.LogOutput('Pushing image onto machine.') + self.logger.LogOutput('Running image_chromeos.DoImage with %s' % + ' '.join(image_chromeos_args)) + retval = 0 + if not test_flag.GetTestMode(): + retval = image_chromeos.DoImage(image_chromeos_args) + if retval: + cmd = 'reboot && exit' + if self.log_level != 'verbose': + self.logger.LogOutput('reboot & exit.') + self.ce.CrosRunCommand(cmd, + machine=machine.name, + chromeos_root=self.chromeos_root) + time.sleep(60) + if self.log_level != 'verbose': + self.logger.LogOutput('Pushing image onto machine.') + self.logger.LogOutput('Running image_chromeos.DoImage with %s' % + ' '.join(image_chromeos_args)) + retval = image_chromeos.DoImage(image_chromeos_args) + if retval: + raise RuntimeError("Could not image machine: '%s'." % machine.name) + else: + self.num_reimages += 1 + machine.checksum = checksum + machine.image = label.chromeos_image + machine.label = label + + if not label.chrome_version: + label.chrome_version = self.GetChromeVersion(machine) + + self.ce.log_level = save_ce_log_level + return retval + + def ComputeCommonCheckSum(self, label): + # Since this is used for cache lookups before the machines have been + # compared/verified, check here to make sure they all have the same + # checksum (otherwise the cache lookup may not be valid). + common_checksum = None + for machine in self.GetMachines(label): + # Make sure the machine's checksums are calculated. + if not machine.machine_checksum: + machine.SetUpChecksumInfo() + cs = machine.machine_checksum + # If this is the first machine we've examined, initialize + # common_checksum. + if not common_checksum: + common_checksum = cs + # Make sure this machine's checksum matches our 'common' checksum. + if cs != common_checksum: + raise BadChecksum('Machine checksums do not match!') + self.machine_checksum[label.name] = common_checksum + + def ComputeCommonCheckSumString(self, label): + # The assumption is that this function is only called AFTER + # ComputeCommonCheckSum, so there is no need to verify the machines + # are the same here. If this is ever changed, this function should be + # modified to verify that all the machines for a given label are the + # same. + for machine in self.GetMachines(label): + if machine.checksum_string: + self.machine_checksum_string[label.name] = machine.checksum_string + break + + def _TryToLockMachine(self, cros_machine): + with self._lock: + assert cros_machine, "Machine can't be None" + for m in self._machines: + if m.name == cros_machine.name: + return + locked = True + if self.locks_dir: + locked = file_lock_machine.Machine(cros_machine.name, + self.locks_dir).Lock(True, + sys.argv[0]) + if locked: + self._machines.append(cros_machine) + command = 'cat %s' % CHECKSUM_FILE + ret, out, _ = self.ce.CrosRunCommandWOutput( + command, + chromeos_root=self.chromeos_root, + machine=cros_machine.name) + if ret == 0: + cros_machine.checksum = out.strip() + elif self.locks_dir: + self.logger.LogOutput("Couldn't lock: %s" % cros_machine.name) + + # This is called from single threaded mode. + def AddMachine(self, machine_name): + with self._lock: + for m in self._all_machines: + assert m.name != machine_name, 'Tried to double-add %s' % machine_name + + if self.log_level != 'verbose': + self.logger.LogOutput('Setting up remote access to %s' % machine_name) + self.logger.LogOutput('Checking machine characteristics for %s' % + machine_name) + cm = CrosMachine(machine_name, self.chromeos_root, self.log_level) + if cm.machine_checksum: + self._all_machines.append(cm) + + def RemoveMachine(self, machine_name): + with self._lock: + self._machines = [m for m in self._machines if m.name != machine_name] + if self.locks_dir: + res = file_lock_machine.Machine(machine_name, + self.locks_dir).Unlock(True) + if not res: + self.logger.LogError("Could not unlock machine: '%s'." % machine_name) + + def ForceSameImageToAllMachines(self, label): + machines = self.GetMachines(label) + for m in machines: + self.ImageMachine(m, label) + m.SetUpChecksumInfo() + + def AcquireMachine(self, label): + image_checksum = label.checksum + machines = self.GetMachines(label) + check_interval_time = 120 + with self._lock: + # Lazily external lock machines + while self.acquire_timeout >= 0: + for m in machines: + new_machine = m not in self._all_machines + self._TryToLockMachine(m) + if new_machine: + m.released_time = time.time() + if self.GetAvailableMachines(label): + break + sleep_time = max(1, min(self.acquire_timeout, check_interval_time)) + time.sleep(sleep_time) + self.acquire_timeout -= sleep_time + + if self.acquire_timeout < 0: + self.logger.LogFatal('Could not acquire any of the ' + "following machines: '%s'" % + ', '.join(machine.name for machine in machines)) + +### for m in self._machines: +### if (m.locked and time.time() - m.released_time < 10 and +### m.checksum == image_checksum): +### return None + unlocked_machines = [machine + for machine in self.GetAvailableMachines(label) + if not machine.locked] + for m in unlocked_machines: + if image_checksum and m.checksum == image_checksum: + m.locked = True + m.test_run = threading.current_thread() + return m + for m in unlocked_machines: + if not m.checksum: + m.locked = True + m.test_run = threading.current_thread() + return m + # This logic ensures that threads waiting on a machine will get a machine + # with a checksum equal to their image over other threads. This saves time + # when crosperf initially assigns the machines to threads by minimizing + # the number of re-images. + # TODO(asharif): If we centralize the thread-scheduler, we wont need this + # code and can implement minimal reimaging code more cleanly. + for m in unlocked_machines: + if time.time() - m.released_time > 15: + # The release time gap is too large, so it is probably in the start + # stage, we need to reset the released_time. + m.released_time = time.time() + elif time.time() - m.released_time > 8: + m.locked = True + m.test_run = threading.current_thread() + return m + return None + + def GetAvailableMachines(self, label=None): + if not label: + return self._machines + return [m for m in self._machines if m.name in label.remote] + + def GetMachines(self, label=None): + if not label: + return self._all_machines + return [m for m in self._all_machines if m.name in label.remote] + + def ReleaseMachine(self, machine): + with self._lock: + for m in self._machines: + if machine.name == m.name: + assert m.locked, 'Tried to double-release %s' % m.name + m.released_time = time.time() + m.locked = False + m.status = 'Available' + break + + def Cleanup(self): + with self._lock: + # Unlock all machines (via file lock) + for m in self._machines: + res = file_lock_machine.Machine(m.name, self.locks_dir).Unlock(True) + + if not res: + self.logger.LogError("Could not unlock machine: '%s'." % m.name) + + def __str__(self): + with self._lock: + l = ['MachineManager Status:'] + [str(m) for m in self._machines] + return '\n'.join(l) + + def AsString(self): + with self._lock: + stringify_fmt = '%-30s %-10s %-4s %-25s %-32s' + header = stringify_fmt % ('Machine', 'Thread', 'Lock', 'Status', + 'Checksum') + table = [header] + for m in self._machines: + if m.test_run: + test_name = m.test_run.name + test_status = m.test_run.timeline.GetLastEvent() + else: + test_name = '' + test_status = '' + + try: + machine_string = stringify_fmt % (m.name, test_name, m.locked, + test_status, m.checksum) + except ValueError: + machine_string = '' + table.append(machine_string) + return 'Machine Status:\n%s' % '\n'.join(table) + + def GetAllCPUInfo(self, labels): + """Get cpuinfo for labels, merge them if their cpuinfo are the same.""" + dic = collections.defaultdict(list) + for label in labels: + for machine in self._all_machines: + if machine.name in label.remote: + dic[machine.cpuinfo].append(label.name) + break + output_segs = [] + for key, v in dic.iteritems(): + output = ' '.join(v) + output += '\n-------------------\n' + output += key + output += '\n\n\n' + output_segs.append(output) + return ''.join(output_segs) + + def GetAllMachines(self): + return self._all_machines + + +class MockCrosMachine(CrosMachine): + """Mock cros machine class.""" + # pylint: disable=super-init-not-called + + MEMINFO_STRING = """MemTotal: 3990332 kB +MemFree: 2608396 kB +Buffers: 147168 kB +Cached: 811560 kB +SwapCached: 0 kB +Active: 503480 kB +Inactive: 628572 kB +Active(anon): 174532 kB +Inactive(anon): 88576 kB +Active(file): 328948 kB +Inactive(file): 539996 kB +Unevictable: 0 kB +Mlocked: 0 kB +SwapTotal: 5845212 kB +SwapFree: 5845212 kB +Dirty: 9384 kB +Writeback: 0 kB +AnonPages: 173408 kB +Mapped: 146268 kB +Shmem: 89676 kB +Slab: 188260 kB +SReclaimable: 169208 kB +SUnreclaim: 19052 kB +KernelStack: 2032 kB +PageTables: 7120 kB +NFS_Unstable: 0 kB +Bounce: 0 kB +WritebackTmp: 0 kB +CommitLimit: 7840376 kB +Committed_AS: 1082032 kB +VmallocTotal: 34359738367 kB +VmallocUsed: 364980 kB +VmallocChunk: 34359369407 kB +DirectMap4k: 45824 kB +DirectMap2M: 4096000 kB +""" + + CPUINFO_STRING = """processor: 0 +vendor_id: GenuineIntel +cpu family: 6 +model: 42 +model name: Intel(R) Celeron(R) CPU 867 @ 1.30GHz +stepping: 7 +microcode: 0x25 +cpu MHz: 1300.000 +cache size: 2048 KB +physical id: 0 +siblings: 2 +core id: 0 +cpu cores: 2 +apicid: 0 +initial apicid: 0 +fpu: yes +fpu_exception: yes +cpuid level: 13 +wp: yes +flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer xsave lahf_lm arat epb xsaveopt pln pts dts tpr_shadow vnmi flexpriority ept vpid +bogomips: 2594.17 +clflush size: 64 +cache_alignment: 64 +address sizes: 36 bits physical, 48 bits virtual +power management: + +processor: 1 +vendor_id: GenuineIntel +cpu family: 6 +model: 42 +model name: Intel(R) Celeron(R) CPU 867 @ 1.30GHz +stepping: 7 +microcode: 0x25 +cpu MHz: 1300.000 +cache size: 2048 KB +physical id: 0 +siblings: 2 +core id: 1 +cpu cores: 2 +apicid: 2 +initial apicid: 2 +fpu: yes +fpu_exception: yes +cpuid level: 13 +wp: yes +flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer xsave lahf_lm arat epb xsaveopt pln pts dts tpr_shadow vnmi flexpriority ept vpid +bogomips: 2594.17 +clflush size: 64 +cache_alignment: 64 +address sizes: 36 bits physical, 48 bits virtual +power management: +""" + + def __init__(self, name, chromeos_root, log_level): + self.name = name + self.image = None + self.checksum = None + self.locked = False + self.released_time = time.time() + self.test_run = None + self.chromeos_root = chromeos_root + self.checksum_string = re.sub(r'\d', '', name) + #In test, we assume "lumpy1", "lumpy2" are the same machine. + self.machine_checksum = self._GetMD5Checksum(self.checksum_string) + self.log_level = log_level + self.label = None + self.ce = command_executer.GetCommandExecuter(log_level=self.log_level) + self._GetCPUInfo() + + def IsReachable(self): + return True + + def _GetMemoryInfo(self): + self.meminfo = self.MEMINFO_STRING + self._ParseMemoryInfo() + + def _GetCPUInfo(self): + self.cpuinfo = self.CPUINFO_STRING + + +class MockMachineManager(MachineManager): + """Mock machine manager class.""" + + def __init__(self, chromeos_root, acquire_timeout, log_level, locks_dir): + super(MockMachineManager, self).__init__( + chromeos_root, acquire_timeout, log_level, locks_dir) + + def _TryToLockMachine(self, cros_machine): + self._machines.append(cros_machine) + cros_machine.checksum = '' + + def AddMachine(self, machine_name): + with self._lock: + for m in self._all_machines: + assert m.name != machine_name, 'Tried to double-add %s' % machine_name + cm = MockCrosMachine(machine_name, self.chromeos_root, self.log_level) + assert cm.machine_checksum, ('Could not find checksum for machine %s' % + machine_name) + # In Original MachineManager, the test is 'if cm.machine_checksum:' - if a + # machine is unreachable, then its machine_checksum is None. Here we + # cannot do this, because machine_checksum is always faked, so we directly + # test cm.IsReachable, which is properly mocked. + if cm.IsReachable(): + self._all_machines.append(cm) + + def GetChromeVersion(self, machine): + return 'Mock Chrome Version R50' + + def AcquireMachine(self, label): + for machine in self._all_machines: + if not machine.locked: + machine.locked = True + return machine + return None + + def ImageMachine(self, machine_name, label): + if machine_name or label: + return 0 + return 1 + + def ReleaseMachine(self, machine): + machine.locked = False + + def GetMachines(self, label=None): + return self._all_machines + + def GetAvailableMachines(self, label=None): + return self._all_machines + + def ForceSameImageToAllMachines(self, label=None): + return 0 + + def ComputeCommonCheckSum(self, label=None): + common_checksum = 12345 + for machine in self.GetMachines(label): + machine.machine_checksum = common_checksum + self.machine_checksum[label.name] = common_checksum + + def GetAllMachines(self): + return self._all_machines |