def test_persist_across_reboots(self):
"""Ensure persist_across_reboots works"""
persistent_cache.set_value('key', 'a')
persistent_cache.set_value('key2', 'b', persist_across_reboots=True)
persistent_cache.clear_values()
self.assertEqual(persistent_cache.get_value('key'), None)
self.assertEqual(persistent_cache.get_value('key2'), 'b')
def update_heartbeat(force_update=False):
"""Updates heartbeat with current timestamp and log data."""
# Check if the heartbeat was recently updated. If yes, bail out.
last_modified_time = persistent_cache.get_value(
HEARTBEAT_LAST_UPDATE_KEY, constructor=datetime.datetime.utcfromtimestamp)
if (not force_update and last_modified_time and not dates.time_has_expired(
last_modified_time, seconds=data_types.HEARTBEAT_WAIT_INTERVAL)):
return 0
bot_name = environment.get_value('BOT_NAME')
current_time = datetime.datetime.utcnow()
try:
heartbeat = ndb.Key(data_types.Heartbeat, bot_name).get()
if not heartbeat:
heartbeat = data_types.Heartbeat()
heartbeat.bot_name = bot_name
heartbeat.key = ndb.Key(data_types.Heartbeat, bot_name)
heartbeat.task_payload = tasks.get_task_payload()
heartbeat.task_end_time = tasks.get_task_end_time()
heartbeat.last_beat_time = current_time
heartbeat.source_version = utils.current_source_version()
heartbeat.put()
persistent_cache.set_value(
HEARTBEAT_LAST_UPDATE_KEY, time.time(), persist_across_reboots=True)
except:
logs.log_error('Unable to update heartbeat.')
return 0
return 1
def test_clear_all(self):
"""Ensure clear all works."""
persistent_cache.set_value('key', 'a')
persistent_cache.set_value('key2', 'b', persist_across_reboots=True)
persistent_cache.clear_values(clear_all=True)
self.assertEqual(persistent_cache.get_value('key'), None)
self.assertEqual(persistent_cache.get_value('key2'), None)
def wait_until_good_state():
"""Check battery and make sure it is charged beyond minimum level and
temperature thresholds."""
# Battery levels are not applicable on GCE.
if adb.is_gce():
return
# Make sure device is online.
adb.wait_for_device()
# Skip battery check if done recently.
last_battery_check_time = persistent_cache.get_value(
LAST_BATTERY_CHECK_TIME_KEY,
constructor=datetime.datetime.utcfromtimestamp)
if last_battery_check_time and not dates.time_has_expired(
last_battery_check_time, seconds=BATTERY_CHECK_INTERVAL):
return
# Initialize variables.
battery_level_threshold = environment.get_value("LOW_BATTERY_LEVEL_THRESHOLD",
LOW_BATTERY_LEVEL_THRESHOLD)
battery_temperature_threshold = environment.get_value(
"MAX_BATTERY_TEMPERATURE_THRESHOLD", MAX_BATTERY_TEMPERATURE_THRESHOLD)
device_restarted = False
while True:
battery_information = get_battery_level_and_temperature()
if battery_information is None:
logs.log_error("Failed to get battery information, skipping check.")
return
battery_level = battery_information["level"]
battery_temperature = battery_information["temperature"]
logs.log("Battery information: level (%d%%), temperature (%.1f celsius)." %
(battery_level, battery_temperature))
if (battery_level >= battery_level_threshold and
battery_temperature <= battery_temperature_threshold):
persistent_cache.set_value(LAST_BATTERY_CHECK_TIME_KEY, time.time())
return
logs.log("Battery in bad battery state, putting device in sleep mode.")
if not device_restarted:
adb.reboot()
device_restarted = True
# Change thresholds to expected levels (only if they were below minimum
# thresholds).
if battery_level < battery_level_threshold:
battery_level_threshold = EXPECTED_BATTERY_LEVEL
if battery_temperature > battery_temperature_threshold:
battery_temperature_threshold = EXPECTED_BATTERY_TEMPERATURE
# Stopping shell should help with shutting off a lot of services that would
# otherwise use up the battery. However, we need to turn it back on to get
# battery status information.
adb.stop_shell()
time.sleep(BATTERY_CHARGE_INTERVAL)
adb.start_shell()
def track_task_start(task, task_duration):
"""Cache task information."""
persistent_cache.set_value(TASK_PAYLOAD_KEY, task.payload())
persistent_cache.set_value(TASK_END_TIME_KEY, time.time() + task_duration)
# Don't wait on |run_heartbeat|, update task information as soon as it starts.
from datastore import data_handler
data_handler.update_heartbeat(force_update=True)
def test_get_invalid(self):
"""Ensure it returns default_value when constructor fails"""
time_now = datetime.datetime.utcnow()
persistent_cache.set_value('key', 'random')
self.assertEqual(persistent_cache.get_value('key'), 'random')
self.assertEqual(
persistent_cache.get_value(
'key',
default_value=time_now,
constructor=datetime.datetime.utcfromtimestamp), time_now)
def test_set_get_datetime(self):
"""Ensure it works with datetime value"""
epoch = datetime.datetime.utcfromtimestamp(0)
end_time = datetime.datetime.utcfromtimestamp(10)
diff_time = end_time - epoch
persistent_cache.set_value('key', diff_time.total_seconds())
self.assertEqual(
persistent_cache.get_value(
'key', constructor=datetime.datetime.utcfromtimestamp),
end_time)
def update_tests_if_needed():
"""Updates layout tests every day."""
data_directory = environment.get_value('FUZZ_DATA')
error_occured = False
expected_task_duration = 60 * 60 # 1 hour.
retry_limit = environment.get_value('FAIL_RETRIES')
temp_archive = os.path.join(data_directory, 'temp.zip')
tests_url = environment.get_value('WEB_TESTS_URL')
# Check if we have a valid tests url.
if not tests_url:
return
# Layout test updates are usually disabled to speedup local testing.
if environment.get_value('LOCAL_DEVELOPMENT'):
return
# |UPDATE_WEB_TESTS| env variable can be used to control our update behavior.
if not environment.get_value('UPDATE_WEB_TESTS'):
return
last_modified_time = persistent_cache.get_value(
LAYOUT_TEST_LAST_UPDATE_KEY,
constructor=datetime.datetime.utcfromtimestamp)
if (last_modified_time is not None and not dates.time_has_expired(
last_modified_time, days=LAYOUT_TEST_UPDATE_INTERVAL_DAYS)):
return
logs.log('Updating layout tests.')
tasks.track_task_start(tasks.Task('update_tests', '', ''),
expected_task_duration)
# Download and unpack the tests archive.
for _ in xrange(retry_limit):
try:
shell.remove_directory(data_directory, recreate=True)
storage.copy_file_from(tests_url, temp_archive)
archive.unpack(temp_archive, data_directory, trusted=True)
shell.remove_file(temp_archive)
error_occured = False
break
except:
logs.log_error(
'Could not retrieve and unpack layout tests archive. Retrying.'
)
error_occured = True
if not error_occured:
persistent_cache.set_value(LAYOUT_TEST_LAST_UPDATE_KEY,
time.time(),
persist_across_reboots=True)
tasks.track_task_end()
def put(self, key, value):
"""Put (key, value) into cache."""
# Lock to avoid race condition in pop.
self.lock.acquire()
if len(self.keys) >= self.capacity:
key_to_remove = self.keys.pop(0)
persistent_cache.delete_value(key_to_remove)
persistent_cache.set_value(key, value)
self.keys.append(key)
self.lock.release()
def remove_unused_builds():
"""Remove any builds that are no longer in use by this bot."""
builds_directory = environment.get_value('BUILDS_DIR')
last_checked_time = persistent_cache.get_value(
LAST_UNUSED_BUILD_CHECK_KEY,
constructor=datetime.datetime.utcfromtimestamp)
if (last_checked_time is not None and
not dates.time_has_expired(last_checked_time, days=1)):
return
# Initialize the map with all of our build directories.
build_in_use_map = {}
for build_directory in os.listdir(builds_directory):
absolute_build_directory = os.path.join(builds_directory, build_directory)
if os.path.isdir(absolute_build_directory):
build_in_use_map[absolute_build_directory] = False
# Platforms for jobs may come from the queue override, but use the default
# if no override is present.
job_platform = environment.get_platform_group()
jobs_for_platform = ndb_utils.get_all_from_query(
data_types.Job.query(data_types.Job.platform == job_platform))
for job in jobs_for_platform:
job_environment = job.get_environment()
# Do not attempt to process any incomplete job definitions.
if not job_environment:
continue
for key, value in job_environment.iteritems():
if 'BUILD_BUCKET_PATH' in key:
bucket_path = value
elif key == 'CUSTOM_BINARY' and value != 'False':
bucket_path = None
else:
continue
# If we made it to this point, this build is potentially in use.
build_directory = _get_build_directory(bucket_path, job.name)
if build_directory in build_in_use_map:
build_in_use_map[build_directory] = True
for build_directory, in_use in build_in_use_map.iteritems():
if in_use:
continue
# Remove the build.
logs.log('Removing unused build directory: %s' % build_directory)
shell.remove_directory(build_directory)
persistent_cache.set_value(LAST_UNUSED_BUILD_CHECK_KEY, time.time())
def add_test_accounts_if_needed():
"""Add test account to work with GmsCore, etc."""
last_test_account_check_time = persistent_cache.get_value(
constants.LAST_TEST_ACCOUNT_CHECK_KEY,
constructor=datetime.datetime.utcfromtimestamp)
needs_test_account_update = (last_test_account_check_time is None
or dates.time_has_expired(
last_test_account_check_time,
seconds=ADD_TEST_ACCOUNT_CHECK_INTERVAL))
if not needs_test_account_update:
return
config = db_config.get()
if not config:
return
test_account_email = config.test_account_email
test_account_password = config.test_account_password
if not test_account_email or not test_account_password:
return
adb.run_as_root()
wifi.configure(force_enable=True)
if not app.is_installed(ADD_TEST_ACCOUNT_PKG_NAME):
logs.log('Installing helper apk for adding test account.')
android_directory = environment.get_platform_resources_directory()
add_test_account_apk_path = os.path.join(android_directory,
ADD_TEST_ACCOUNT_APK_NAME)
app.install(add_test_account_apk_path)
logs.log('Trying to add test account.')
output = adb.run_shell_command(
'am instrument -e account %s -e password %s -w %s' %
(test_account_email, test_account_password,
ADD_TEST_ACCOUNT_CALL_PATH),
timeout=ADD_TEST_ACCOUNT_TIMEOUT)
if not output or test_account_email not in output:
logs.log('Failed to add test account, probably due to wifi issues.')
return
logs.log('Test account added successfully.')
persistent_cache.set_value(constants.LAST_TEST_ACCOUNT_CHECK_KEY,
time.time())
def remove_running_handle(handle):
"""Remove a handle from the tracked set."""
new_handle_list = list(set(get_running_handles()) - set([handle]))
persistent_cache.set_value(
HANDLE_CACHE_KEY, new_handle_list, persist_across_reboots=True)
def configure_system_build_properties():
"""Modifies system build properties in /system/build.prop for better boot
speed and power use."""
# Check md5 checksum of build.prop to see if already updated,
# in which case exit. If build.prop does not exist, something
# is very wrong with the device, so bail.
old_md5 = persistent_cache.get_value(constants.BUILD_PROP_MD5_KEY)
current_md5 = adb.get_file_checksum(BUILD_PROP_PATH)
if current_md5 is None:
logs.log_error('Unable to find %s on device.' % BUILD_PROP_PATH)
return
if old_md5 == current_md5:
return
# Pull to tmp file.
bot_tmp_directory = environment.get_value('BOT_TMPDIR')
old_build_prop_path = os.path.join(bot_tmp_directory, 'old.prop')
adb.run_command(['pull', BUILD_PROP_PATH, old_build_prop_path])
if not os.path.exists(old_build_prop_path):
logs.log_error('Unable to fetch %s from device.' % BUILD_PROP_PATH)
return
# Write new build.prop.
new_build_prop_path = os.path.join(bot_tmp_directory, 'new.prop')
old_build_prop_file_content = open(old_build_prop_path, 'r')
new_build_prop_file_content = open(new_build_prop_path, 'w')
new_content_notification = '### CHANGED OR ADDED PROPERTIES ###'
for line in old_build_prop_file_content:
property_name = line.split('=')[0].strip()
if property_name in BUILD_PROPERTIES:
continue
if new_content_notification in line:
continue
new_build_prop_file_content.write(line)
new_build_prop_file_content.write(new_content_notification + '\n')
for flag, value in six.iteritems(BUILD_PROPERTIES):
new_build_prop_file_content.write('%s=%s\n' % (flag, value))
old_build_prop_file_content.close()
new_build_prop_file_content.close()
# Keep verified boot disabled for M and higher releases. This makes it easy
# to modify system's app_process to load asan libraries.
build_version = settings.get_build_version()
if is_build_at_least(build_version, 'M'):
adb.run_as_root()
adb.run_command('disable-verity')
reboot()
# Make /system writable.
adb.run_as_root()
adb.remount()
# Remove seccomp policies (on N and higher) as ASan requires extra syscalls.
if is_build_at_least(build_version, 'N'):
policy_files = adb.run_shell_command(
['find', '/system/etc/seccomp_policy/', '-type', 'f'])
for policy_file in policy_files.splitlines():
adb.run_shell_command(['rm', policy_file.strip()])
# Push new build.prop and backup to device.
logs.log('Pushing new build properties file on device.')
adb.run_command(
['push', '-p', old_build_prop_path, BUILD_PROP_BACKUP_PATH])
adb.run_command(['push', '-p', new_build_prop_path, BUILD_PROP_PATH])
adb.run_shell_command(['chmod', '644', BUILD_PROP_PATH])
# Set persistent cache key containing and md5sum.
current_md5 = adb.get_file_checksum(BUILD_PROP_PATH)
persistent_cache.set_value(constants.BUILD_PROP_MD5_KEY, current_md5)
def test_set_get_string(self):
"""Ensure it works with string value"""
persistent_cache.set_value('key', 'value')
self.assertEqual(persistent_cache.get_value('key'), 'value')
def flash_to_latest_build_if_needed():
"""Wipes user data, resetting the device to original factory state."""
if environment.get_value('LOCAL_DEVELOPMENT'):
# Don't reimage local development devices.
return
run_timeout = environment.get_value('RUN_TIMEOUT')
if run_timeout:
# If we have a run timeout, then we are already scheduled to bail out and
# will be probably get re-imaged. E.g. using frameworks like Tradefed.
return
# Check if a flash is needed based on last recorded flash time.
last_flash_time = persistent_cache.get_value(
constants.LAST_FLASH_TIME_KEY,
constructor=datetime.datetime.utcfromtimestamp)
needs_flash = last_flash_time is None or dates.time_has_expired(
last_flash_time, seconds=FLASH_INTERVAL)
if not needs_flash:
return
is_google_device = settings.is_google_device()
if is_google_device is None:
logs.log_error('Unable to query device. Reimaging failed.')
adb.bad_state_reached()
elif not is_google_device:
# We can't reimage these, skip.
logs.log('Non-Google device found, skipping reimage.')
return
# Check if both |BUILD_BRANCH| and |BUILD_TARGET| environment variables
# are set. If not, we don't have enough data for reimaging and hence
# we bail out.
branch = environment.get_value('BUILD_BRANCH')
target = environment.get_value('BUILD_TARGET')
if not target:
# We default to userdebug configuration.
build_params = settings.get_build_parameters()
if build_params:
target = build_params.get('target') + '-userdebug'
# Cache target in environment. This is also useful for cases when
# device is bricked and we don't have this information available.
environment.set_value('BUILD_TARGET', target)
if not branch or not target:
logs.log_warn(
'BUILD_BRANCH and BUILD_TARGET are not set, skipping reimage.')
return
image_directory = environment.get_value('IMAGES_DIR')
build_info = fetch_artifact.get_latest_artifact_info(branch, target)
if not build_info:
logs.log_error('Unable to fetch information on latest build artifact for '
'branch %s and target %s.' % (branch, target))
return
if environment.is_android_cuttlefish():
download_latest_build(build_info, FLASH_CUTTLEFISH_REGEXES, image_directory)
adb.recreate_cuttlefish_device()
adb.connect_to_cuttlefish_device()
else:
download_latest_build(build_info, FLASH_IMAGE_REGEXES, image_directory)
# We do one device flash at a time on one host, otherwise we run into
# failures and device being stuck in a bad state.
flash_lock_key_name = 'flash:%s' % socket.gethostname()
if not locks.acquire_lock(flash_lock_key_name, by_zone=True):
logs.log_error('Failed to acquire lock for reimaging, exiting.')
return
logs.log('Reimaging started.')
logs.log('Rebooting into bootloader mode.')
for _ in range(FLASH_RETRIES):
adb.run_as_root()
adb.run_command(['reboot-bootloader'])
time.sleep(FLASH_REBOOT_BOOTLOADER_WAIT)
adb.run_fastboot_command(['oem', 'off-mode-charge', '0'])
adb.run_fastboot_command(['-w', 'reboot-bootloader'])
for partition, partition_image_filename in FLASH_IMAGE_FILES:
partition_image_file_path = os.path.join(image_directory,
partition_image_filename)
adb.run_fastboot_command(
['flash', partition, partition_image_file_path])
if partition in ['bootloader', 'radio']:
adb.run_fastboot_command(['reboot-bootloader'])
# Disable ramdump to avoid capturing ramdumps during kernel crashes.
# This causes device lockup of several minutes during boot and we intend
# to analyze them ourselves.
adb.run_fastboot_command(['oem', 'ramdump', 'disable'])
adb.run_fastboot_command('reboot')
time.sleep(FLASH_REBOOT_WAIT)
if adb.get_device_state() == 'device':
break
logs.log_error('Reimaging failed, retrying.')
locks.release_lock(flash_lock_key_name, by_zone=True)
if adb.get_device_state() != 'device':
logs.log_error('Unable to find device. Reimaging failed.')
adb.bad_state_reached()
logs.log('Reimaging finished.')
# Reset all of our persistent keys after wipe.
persistent_cache.delete_value(constants.BUILD_PROP_MD5_KEY)
persistent_cache.delete_value(constants.LAST_TEST_ACCOUNT_CHECK_KEY)
persistent_cache.set_value(constants.LAST_FLASH_BUILD_KEY, build_info)
persistent_cache.set_value(constants.LAST_FLASH_TIME_KEY, time.time())
def test_delete(self):
"""Ensure it deletes key"""
persistent_cache.set_value('key', 'value')
persistent_cache.delete_value('key')
self.assertEqual(persistent_cache.get_value('key'), None)
def flash_to_latest_build_if_needed():
"""Wipes user data, resetting the device to original factory state."""
if environment.get_value('LOCAL_DEVELOPMENT'):
# Don't reimage local development devices.
return
run_timeout = environment.get_value('RUN_TIMEOUT')
if run_timeout:
# If we have a run timeout, then we are already scheduled to bail out and
# will be probably get re-imaged. E.g. using frameworks like Tradefed.
return
# Check if a flash is needed based on last recorded flash time.
last_flash_time = persistent_cache.get_value(
LAST_FLASH_TIME_KEY, constructor=datetime.datetime.utcfromtimestamp)
needs_flash = last_flash_time is None or dates.time_has_expired(
last_flash_time, seconds=adb.FLASH_INTERVAL)
if not needs_flash:
return
build_info = {}
if adb.is_gce():
adb.recreate_gce_device()
else:
# Physical device.
is_google_device = google_device()
if is_google_device is None:
logs.log_error('Unable to query device. Reimaging failed.')
adb.bad_state_reached()
elif not is_google_device:
# We can't reimage these, skip.
logs.log('Non-Google device found, skipping reimage.')
return
else:
# For Google devices.
# Check if both |BUILD_BRANCH| and |BUILD_TARGET| environment variables
# are set. If not, we don't have enough data for reimaging and hence
# we bail out.
branch = environment.get_value('BUILD_BRANCH')
target = environment.get_value('BUILD_TARGET')
if not target:
# We default to userdebug configuration.
build_params = get_build_parameters()
if build_params:
target = build_params.get('target') + '-userdebug'
# Cache target in environment. This is also useful for cases when
# device is bricked and we don't have this information available.
environment.set_value('BUILD_TARGET', target)
if not branch or not target:
logs.log_warn(
'BUILD_BRANCH and BUILD_TARGET are not set, skipping reimage.'
)
return
# Download the latest build artifact for this branch and target.
build_info = fetch_artifact.get_latest_artifact_info(
branch, target)
if not build_info:
logs.log_error(
'Unable to fetch information on latest build artifact for '
'branch %s and target %s.' % (branch, target))
return
# Check if our local build matches the latest build. If not, we will
# download it.
build_id = build_info['bid']
target = build_info['target']
image_directory = environment.get_value('IMAGES_DIR')
last_build_info = persistent_cache.get_value(LAST_FLASH_BUILD_KEY)
if not last_build_info or last_build_info['bid'] != build_id:
# Clean up the images directory first.
shell.remove_directory(image_directory, recreate=True)
# We have a new build, download the build artifacts for it.
for image_regex in FLASH_IMAGE_REGEXES:
image_file_path = fetch_artifact.get(
build_id, target, image_regex, image_directory)
if not image_file_path:
logs.log_error(
'Failed to download image artifact %s for '
'branch %s and target %s.' %
(image_file_path, branch, target))
return
if image_file_path.endswith('.zip'):
archive.unpack(image_file_path, image_directory)
# We do one device flash at a time on one host, otherwise we run into
# failures and device being stuck in a bad state.
flash_lock_key_name = 'flash:%s' % socket.gethostname()
if not locks.acquire_lock(flash_lock_key_name, by_zone=True):
logs.log_error(
'Failed to acquire lock for reimaging, exiting.')
return
logs.log('Reimaging started.')
logs.log('Rebooting into bootloader mode.')
for _ in xrange(FLASH_RETRIES):
adb.run_as_root()
adb.run_adb_command(['reboot-bootloader'])
time.sleep(FLASH_REBOOT_BOOTLOADER_WAIT)
adb.run_fastboot_command(['oem', 'off-mode-charge', '0'])
adb.run_fastboot_command(['-w', 'reboot-bootloader'])
for partition, partition_image_filename in FLASH_IMAGE_FILES:
partition_image_file_path = os.path.join(
image_directory, partition_image_filename)
adb.run_fastboot_command(
['flash', partition, partition_image_file_path])
if partition in ['bootloader', 'radio']:
adb.run_fastboot_command(['reboot-bootloader'])
adb.run_fastboot_command('reboot')
time.sleep(FLASH_REBOOT_WAIT)
if adb.get_device_state() == 'device':
break
logs.log_error('Reimaging failed, retrying.')
locks.release_lock(flash_lock_key_name, by_zone=True)
if adb.get_device_state() != 'device':
logs.log_error('Unable to find device. Reimaging failed.')
adb.bad_state_reached()
logs.log('Reimaging finished.')
# Reset all of our persistent keys after wipe.
persistent_cache.delete_value(BUILD_PROP_MD5_KEY)
persistent_cache.delete_value(LAST_TEST_ACCOUNT_CHECK_KEY)
persistent_cache.set_value(LAST_FLASH_BUILD_KEY, build_info)
persistent_cache.set_value(LAST_FLASH_TIME_KEY, time.time())
def recreate_instance_with_disks(instance_name,
project,
zone,
additional_metadata=None,
wait_for_completion=False):
"""Recreate an instance and its disk."""
# Get existing instance information.
# First, try to get instance info from cache.
# TODO(ochang): Make this more general in case anything else needs to use
# this method (e.g. appengine).
instance_info = persistent_cache.get_value(GCE_INSTANCE_INFO_KEY)
if instance_info is None:
instance_info = _get_instance_info(instance_name, project, zone)
# Bail out if we don't have a valid instance information.
if (not instance_info or 'disks' not in instance_info or
not instance_info['disks']):
logs.log_error(
'Failed to get disk info from existing instance, bailing on instance '
'recreation.')
return False
# Add any additional metadata required for instance booting.
if additional_metadata:
for key, value in six.iteritems(additional_metadata):
items = instance_info.setdefault('metadata', {}).setdefault('items', [])
_add_metadata_key_value(items, key, value)
# Cache the latest instance information.
persistent_cache.set_value(
GCE_INSTANCE_INFO_KEY, instance_info, persist_across_reboots=True)
# Delete the instance.
if not _do_instance_operation(
'delete', instance_name, project, zone, wait_for_completion=True):
logs.log_error('Failed to delete instance.')
return False
# Get existing disks information, and recreate.
api = _get_api()
disks = instance_info['disks']
for disk in disks:
disk_source = disk['source']
disk_name = disk_source.split('/')[-1]
disk_info_func = api.disks().get(disk=disk_name, project=project, zone=zone)
disk_info = _execute_api_call_with_retries(disk_info_func)
if 'sourceImage' not in disk_info or 'sizeGb' not in disk_info:
logs.log_error(
'Failed to get source image and size from existing disk, bailing on '
'instance recreation.')
return False
size_gb = disk_info['sizeGb']
source_image = disk_info['sourceImage']
# Recreate the disk.
if not delete_disk(disk_name, project, zone, wait_for_completion=True):
logs.log_error('Failed to delete disk.')
return False
if not create_disk(
disk_name,
source_image,
size_gb,
project,
zone,
wait_for_completion=True):
logs.log_error('Failed to recreate disk.')
return False
# Recreate the instance with the exact same configurations, but not
# necessarily the same IPs.
try:
del instance_info['networkInterfaces'][0]['accessConfigs'][0]['natIP']
except:
# This is not a failure. When a bot is stopped, it has no ip/interface.
pass
try:
del instance_info['networkInterfaces'][0]['networkIP']
except:
# This is not a failure. When a bot is stopped, it has no ip/interface.
pass
operation = api.instances().insert(
body=instance_info, project=project, zone=zone)
return _do_operation_with_retries(
operation, project, zone, wait_for_completion=wait_for_completion)
def add_running_handle(handle):
"""Record a handle as potentially needing to be cleaned up on restart."""
new_handle_list = list(set(get_running_handles()) | set([handle]))
persistent_cache.set_value(
HANDLE_CACHE_KEY, new_handle_list, persist_across_reboots=True)