def test_thread_deletion_with_short_ttl(db, default_account, default_namespace,
marked_deleted_message, thread,
folder):
handler = DeleteHandler(
account_id=default_account.id,
namespace_id=default_namespace.id,
provider_name=default_account.provider,
uid_accessor=lambda m: m.imapuids,
message_ttl=0,
thread_ttl=120,
)
delete_time = marked_deleted_message.deleted_at
handler.check(delete_time + timedelta(seconds=1))
handler.gc_deleted_threads(delete_time + timedelta(seconds=1))
db.session.expire_all()
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
thread.id
assert thread.deleted_at is not None
handler.check(thread.deleted_at + timedelta(seconds=121))
handler.gc_deleted_threads(thread.deleted_at + timedelta(seconds=121))
db.session.expire_all()
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
with pytest.raises(ObjectDeletedError):
thread.id
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def test_deletion_creates_revision(db, default_account, default_namespace,
marked_deleted_message, thread, folder):
message_id = marked_deleted_message.id
thread_id = thread.id
handler = DeleteHandler(
account_id=default_account.id,
namespace_id=default_namespace.id,
provider_name=default_account.provider,
uid_accessor=lambda m: m.imapuids,
message_ttl=0,
)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.commit()
latest_message_transaction = (db.session.query(Transaction).filter(
Transaction.record_id == message_id,
Transaction.object_type == "message",
Transaction.namespace_id == default_namespace.id,
).order_by(desc(Transaction.id)).first())
assert latest_message_transaction.command == "delete"
latest_thread_transaction = (db.session.query(Transaction).filter(
Transaction.record_id == thread_id,
Transaction.object_type == "thread",
Transaction.namespace_id == default_namespace.id,
).order_by(desc(Transaction.id)).first())
assert latest_thread_transaction.command == "update"
def test_thread_deletion_with_short_ttl(db, default_account, default_namespace,
marked_deleted_message, thread, folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
provider_name=default_account.provider,
uid_accessor=lambda m: m.imapuids,
message_ttl=0, thread_ttl=120)
delete_time = marked_deleted_message.deleted_at
handler.check(delete_time + timedelta(seconds=1))
handler.gc_deleted_threads(delete_time + timedelta(seconds=1))
db.session.expire_all()
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
thread.id
assert thread.deleted_at is not None
handler.check(thread.deleted_at + timedelta(seconds=121))
handler.gc_deleted_threads(thread.deleted_at + timedelta(seconds=121))
db.session.expire_all()
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
with pytest.raises(ObjectDeletedError):
thread.id
def test_deletion_with_short_ttl(db, default_account, default_namespace,
marked_deleted_message, thread, folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# Check that objects were actually deleted
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
with pytest.raises(ObjectDeletedError):
thread.id
def test_non_orphaned_messages_get_unmarked(db, default_account,
default_namespace,
marked_deleted_message, thread,
folder, imapuid):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# message actually has an imapuid associated, so check that the
# DeleteHandler unmarked it.
assert marked_deleted_message.deleted_at is None
def test_deletion_with_short_ttl(db, default_account, default_namespace,
marked_deleted_message, thread, folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# Check that objects were actually deleted
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
with pytest.raises(ObjectDeletedError):
thread.id
def test_non_orphaned_messages_get_unmarked(db, default_account,
default_namespace,
marked_deleted_message, thread,
folder, imapuid):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# message actually has an imapuid associated, so check that the
# DeleteHandler unmarked it.
assert marked_deleted_message.deleted_at is None
def test_deletion_deferred_with_longer_ttl(db, default_account,
default_namespace, message, thread,
folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=1)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Would raise ObjectDeletedError if objects were deleted
message.id
thread.id
def test_deletion_deferred_with_longer_ttl(db, default_account,
default_namespace, message, thread,
folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=1)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Would raise ObjectDeletedError if objects were deleted
message.id
thread.id
def test_non_orphaned_messages_get_unmarked(db, default_account,
default_namespace, message, thread,
folder, imapuid):
message.deleted_at = datetime.utcnow()
db.session.commit()
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check()
# message actually has an imapuid associated, so check that the
# DeleteHandler unmarked it.
assert message.deleted_at is None
def test_non_orphaned_messages_get_unmarked(db, default_account,
default_namespace, message, thread,
folder, imapuid):
message.deleted_at = datetime.utcnow()
db.session.commit()
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check()
# message actually has an imapuid associated, so check that the
# DeleteHandler unmarked it.
assert message.deleted_at is None
def test_deletion_deferred_with_longer_ttl(db, default_account,
default_namespace,
marked_deleted_message, thread,
folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=5)
db.session.commit()
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
# Would raise ObjectDeletedError if objects were deleted
marked_deleted_message.id
thread.id
def test_deletion_with_short_ttl(db, default_account, default_namespace,
message, thread, folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Check that objects were actually deleted
with pytest.raises(ObjectDeletedError):
message.id
with pytest.raises(ObjectDeletedError):
thread.id
def test_deletion_with_short_ttl(db, default_account, default_namespace,
message, thread, folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Check that objects were actually deleted
with pytest.raises(ObjectDeletedError):
message.id
with pytest.raises(ObjectDeletedError):
thread.id
def test_deletion_deferred_with_longer_ttl(db, default_account,
default_namespace,
marked_deleted_message, thread,
folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=5)
db.session.commit()
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
# Would raise ObjectDeletedError if objects were deleted
marked_deleted_message.id
thread.id
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def test_threads_only_deleted_when_no_messages_left(db, default_account,
default_namespace, message,
thread, folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
# Add another message onto the thread
add_fake_message(db.session, default_namespace.id, thread)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Check that the orphaned message was deleted.
with pytest.raises(ObjectDeletedError):
message.id
# Would raise ObjectDeletedError if thread was deleted.
thread.id
def test_threads_only_deleted_when_no_messages_left(db, default_account,
default_namespace, message,
thread, folder, imapuid):
msg_uid = imapuid.msg_uid
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
# Add another message onto the thread
add_fake_message(db.session, default_namespace.id, thread)
remove_deleted_uids(default_account.id, db.session, [msg_uid], folder.id)
handler.check()
# Check that the orphaned message was deleted.
with pytest.raises(ObjectDeletedError):
message.id
# Would raise ObjectDeletedError if thread was deleted.
thread.id
def test_deleted_labels_get_gced(empty_db, default_account, thread, message,
imapuid, folder):
# Check that only the labels without messages attached to them
# get deleted.
default_namespace = default_account.namespace
# Create a label w/ no messages attached.
label = Label.find_or_create(empty_db.session, default_account,
"dangling label")
label.deleted_at = datetime.utcnow()
label.category.deleted_at = datetime.utcnow()
label_id = label.id
empty_db.session.commit()
# Create a label with attached messages.
msg_uid = imapuid.msg_uid
update_metadata(
default_account.id,
folder.id,
folder.canonical_name,
{msg_uid: GmailFlags((), ("label", ), None)},
empty_db.session,
)
label_ids = []
for cat in message.categories:
for l in cat.labels:
label_ids.append(l.id)
handler = DeleteHandler(
account_id=default_account.id,
namespace_id=default_namespace.id,
provider_name=default_account.provider,
uid_accessor=lambda m: m.imapuids,
message_ttl=0,
)
handler.gc_deleted_categories()
empty_db.session.commit()
# Check that the first label got gc'ed
marked_deleted = empty_db.session.query(Label).get(label_id)
assert marked_deleted is None
# Check that the other labels didn't.
for label_id in label_ids:
assert empty_db.session.query(Label).get(label_id) is not None
def test_threads_only_deleted_when_no_messages_left(db, default_account,
default_namespace,
marked_deleted_message,
thread, folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
# Add another message onto the thread
add_fake_message(db.session, default_namespace.id, thread)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# Check that the orphaned message was deleted.
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
# Would raise ObjectDeletedError if thread was deleted.
thread.id
def test_threads_only_deleted_when_no_messages_left(db, default_account,
default_namespace,
marked_deleted_message,
thread, folder):
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
# Add another message onto the thread
add_fake_message(db.session, default_namespace.id, thread)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.expire_all()
# Check that the orphaned message was deleted.
with pytest.raises(ObjectDeletedError):
marked_deleted_message.id
# Would raise ObjectDeletedError if thread was deleted.
thread.id
def test_deleted_labels_get_gced(db, default_account, thread, message,
imapuid, folder):
# Check that only the labels without messages attached to them
# get deleted.
default_namespace = default_account.namespace
# Create a label w/ no messages attached.
label = Label.find_or_create(db.session, default_account, 'dangling label')
label.deleted_at = datetime.utcnow()
label.category.deleted_at = datetime.utcnow()
label_id = label.id
db.session.commit()
# Create a label with attached messages.
msg_uid = imapuid.msg_uid
update_metadata(default_account.id, folder.id, folder.canonical_name,
{msg_uid: GmailFlags((), ('label',), None)}, db.session)
label_ids = []
for cat in message.categories:
for l in cat.labels:
label_ids.append(l.id)
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
provider_name=default_account.provider,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.gc_deleted_categories()
db.session.commit()
# Check that the first label got gc'ed
marked_deleted = db.session.query(Label).get(label_id)
assert marked_deleted is None
# Check that the other labels didn't.
for label_id in label_ids:
assert db.session.query(Label).get(label_id) is not None
def test_deletion_creates_revision(db, default_account, default_namespace,
marked_deleted_message, thread, folder):
message_id = marked_deleted_message.id
thread_id = thread.id
handler = DeleteHandler(account_id=default_account.id,
namespace_id=default_namespace.id,
uid_accessor=lambda m: m.imapuids,
message_ttl=0)
handler.check(marked_deleted_message.deleted_at + timedelta(seconds=1))
db.session.commit()
latest_message_transaction = db.session.query(Transaction). \
filter(Transaction.record_id == message_id,
Transaction.object_type == 'message',
Transaction.namespace_id == default_namespace.id). \
order_by(desc(Transaction.id)).first()
assert latest_message_transaction.command == 'delete'
latest_thread_transaction = db.session.query(Transaction). \
filter(Transaction.record_id == thread_id,
Transaction.object_type == 'thread',
Transaction.namespace_id == default_namespace.id). \
order_by(desc(Transaction.id)).first()
assert latest_thread_transaction.command == 'delete'
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
poll_frequency: Integer
Seconds to wait between polling for the greenlets spawned
refresh_flags_max: Integer
the maximum number of UIDs for which we'll check flags
periodically.
"""
def __init__(self,
account,
heartbeat=1,
refresh_frequency=30,
poll_frequency=30,
retry_fail_classes=[],
refresh_flags_max=2000):
self.refresh_frequency = refresh_frequency
self.poll_frequency = poll_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.refresh_flags_max = refresh_flags_max
provider_supports_condstore = account.provider_info.get(
'condstore', False)
account_supports_condstore = getattr(account, 'supports_condstore',
False)
if provider_supports_condstore or account_supports_condstore:
self.sync_engine_class = CondstoreFolderSyncEngine
else:
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
BaseMailSyncMonitor.__init__(self, account, heartbeat,
retry_fail_classes)
@retry_crispin
def prepare_sync(self):
"""Ensures that canonical tags are created for the account, and gets
and save Folder objects for folders on the IMAP backend. Returns a list
of tuples (folder_name, folder_id) for each folder we want to sync (in
order)."""
with mailsync_session_scope() as db_session:
with _pool(self.account_id).get() as crispin_client:
sync_folders = crispin_client.sync_folders()
save_folder_names(log, self.account_id,
crispin_client.folder_names(), db_session)
sync_folder_names_ids = []
for folder_name in sync_folders:
try:
id_, = db_session.query(Folder.id). \
filter(Folder.name == folder_name,
Folder.account_id == self.account_id).one()
sync_folder_names_ids.append((folder_name, id_))
except NoResultFound:
log.error("Missing Folder object when starting sync",
folder_name=folder_name)
raise MailsyncError(
"Missing Folder '{}' on account {}".format(
folder_name, self.account_id))
return sync_folder_names_ids
def start_new_folder_sync_engines(self, folders=set()):
new_folders = [f for f in self.prepare_sync() if f not in folders]
for folder_name, folder_id in new_folders:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
thread = self.sync_engine_class(
self.account_id, folder_name, folder_id, self.email_address,
self.provider_name, self.poll_frequency, self.syncmanager_lock,
self.refresh_flags_max, self.retry_fail_classes)
self.folder_monitors.start(thread)
while not thread_polling(thread) and \
not thread_finished(thread) and \
not thread.ready():
sleep(self.heartbeat)
# allow individual folder sync monitors to shut themselves down
# after completing the initial sync
if thread_finished(thread) or thread.ready():
log.info('Folder sync engine finished/killed',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
# note: thread is automatically removed from
# self.folder_monitors
else:
folders.add((folder_name, folder_id))
def start_delete_handler(self):
self.delete_handler = DeleteHandler(account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def sync(self):
self.start_delete_handler()
folders = set()
self.start_new_folder_sync_engines(folders)
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines(folders)
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
(DEPRECATED) refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
syncback_frequency: Integer
Seconds to wait between performing consecutive syncback iterations and
checking for new folders to sync.
"""
def __init__(self,
account,
heartbeat=1,
refresh_frequency=30,
syncback_frequency=5):
# DEPRECATED.
# TODO[k]: Remove after sync-syncback integration deploy is complete.
self.refresh_frequency = refresh_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.saved_remote_folders = None
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
self.delete_handler = None
self.syncback_handler = None
self.folder_sync_signals = {}
self.syncback_timestamp = None
self.syncback_frequency = syncback_frequency
BaseMailSyncMonitor.__init__(self, account, heartbeat)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of folder names for the folders we want to sync (in order).
"""
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
if self.saved_remote_folders != remote_folders:
with session_scope(self.namespace_id) as db_session:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
return sync_folders
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {
f.display_name.rstrip()[:MAX_FOLDER_NAME_LENGTH]
for f in raw_folders
}
assert 'inbox' in {f.role for f in raw_folders},\
'Account {} has no detected inbox folder'.\
format(account.email_address)
local_folders = {
f.name: f
for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)
}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info('Folder deleted from remote',
account_id=self.account_id,
name=name)
if local_folders[name].category_id is not None:
cat = db_session.query(Category).get(
local_folders[name].category_id)
if cat is not None:
db_session.delete(cat)
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
Folder.find_or_create(db_session, account, raw_folder.display_name,
raw_folder.role)
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self):
running_monitors = {
monitor.folder_name: monitor
for monitor in self.folder_monitors
}
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).options(
load_only('_sync_status')).get(self.account_id)
s3_resync = account._sync_status.get('s3_resync', False)
for folder_name in self.prepare_sync():
if folder_name in running_monitors:
thread = running_monitors[folder_name]
else:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_name=folder_name)
self._add_sync_signal(folder_name)
thread = self.sync_engine_class(
self.account_id, self.namespace_id, folder_name,
self.email_address, self.provider_name,
self.syncmanager_lock,
self.folder_sync_signals[folder_name])
self.folder_monitors.start(thread)
if s3_resync:
log.info('Starting an S3 monitor',
account_id=self.account_id)
s3_thread = S3FolderSyncEngine(self.account_id,
self.namespace_id,
folder_name,
self.email_address,
self.provider_name,
self.syncmanager_lock, None)
self.folder_monitors.start(s3_thread)
while not thread.state == 'poll' and not thread.ready():
sleep(self.heartbeat)
self.perform_syncback()
if thread.ready():
self._remove_sync_signal[folder_name]
log.info('Folder sync engine exited',
account_id=self.account_id,
folder_name=folder_name,
error=thread.exception)
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
provider_name=self.provider_name,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def perform_syncback(self):
"""
Perform syncback for the account.
Syncback is performed iff all folder syncs are paused, and the previous
syncback occurred more than syncback_frequency seconds ago.
The first condition is checked by the call to _can_syncback().
The second condition is needed because if there are a large number of
pending actions during initial sync, it could repeatedly get interrupted
and put on hold for seconds at a time.
"""
from inbox.syncback.base import SyncbackHandler
if not self._can_syncback():
log.info('Skipping syncback', reason='folder syncs running')
return
if (self.syncback_timestamp
and (datetime.utcnow() - self.syncback_timestamp).seconds <
self.syncback_frequency):
log.info('Skipping syncback',
reason='last syncback < syncback_frequency seconds ago',
syncback_frequency=self.syncback_frequency)
# Reset here so syncs can proceed
self._signal_syncs()
return
if self.syncback_handler is None:
self.syncback_handler = SyncbackHandler(self.account_id,
self.namespace_id,
self.provider_name)
try:
interval = ((datetime.utcnow() - self.syncback_timestamp).seconds
if self.syncback_timestamp else None)
log.info('Performing syncback',
syncback_interval_in_seconds=interval)
self.syncback_handler.send_client_changes()
self.syncback_timestamp = datetime.utcnow()
except Exception:
# Log, set self.folder_sync_signals and then re-raise (so the
# greenlet can be restarted etc.)
log.error('Critical syncback error', exc_info=True)
raise
finally:
# Reset here so syncs can proceed
self._signal_syncs()
def sync(self):
try:
self.start_delete_handler()
self.start_new_folder_sync_engines()
while True:
sleep(self.syncback_frequency)
self.perform_syncback()
self.start_new_folder_sync_engines()
except ValidationError as exc:
log.error('Error authenticating; stopping sync',
exc_info=True,
account_id=self.account_id,
logstash_tag='mark_invalid')
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(str(exc))
def _add_sync_signal(self, folder_name):
self.folder_sync_signals[folder_name] = Event()
self.folder_sync_signals[folder_name].set()
def _remove_sync_signal(self, folder_name):
del self.folder_sync_signals[folder_name]
def _can_syncback(self):
"""
Determine if syncback can occur.
If all folder syncs are paused as indicated by the folder_sync_signals,
returns True. Else, returns False.
"""
return (not self.folder_sync_signals
or all(not signal.is_set()
for signal in self.folder_sync_signals.values()))
def _signal_syncs(self):
""" Indicate that folder syncs can resume. """
for signal in self.folder_sync_signals.values():
signal.set()
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
poll_frequency: Integer
Seconds to wait between polling for the greenlets spawned
refresh_flags_max: Integer
the maximum number of UIDs for which we'll check flags
periodically.
"""
def __init__(self, account,
heartbeat=1, refresh_frequency=30, poll_frequency=30,
retry_fail_classes=[], refresh_flags_max=2000):
self.refresh_frequency = refresh_frequency
self.poll_frequency = poll_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.refresh_flags_max = refresh_flags_max
self.saved_remote_folders = None
provider_supports_condstore = account.provider_info.get('condstore',
False)
account_supports_condstore = getattr(account, 'supports_condstore',
False)
if provider_supports_condstore or account_supports_condstore:
self.sync_engine_class = CondstoreFolderSyncEngine
else:
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
BaseMailSyncMonitor.__init__(self, account, heartbeat,
retry_fail_classes)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of tuples (folder_name, folder_id) for each folder we want to sync
(in order).
"""
with mailsync_session_scope() as db_session:
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
if self.saved_remote_folders != remote_folders:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
sync_folder_names_ids = []
for folder_name in sync_folders:
try:
id_, = db_session.query(Folder.id). \
filter(Folder.name == folder_name,
Folder.account_id == self.account_id).one()
sync_folder_names_ids.append((folder_name, id_))
except NoResultFound:
log.error('Missing Folder object when starting sync',
folder_name=folder_name)
raise MailsyncError("Missing Folder '{}' on account {}"
.format(folder_name, self.account_id))
return sync_folder_names_ids
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {f.display_name.rstrip()[:MAX_FOLDER_NAME_LENGTH]
for f in raw_folders}
assert 'inbox' in {f.role for f in raw_folders},\
'Account {} has no detected inbox folder'.\
format(account.email_address)
local_folders = {f.name: f for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info('Folder deleted from remote', account_id=self.account_id,
name=name)
db_session.delete(local_folders[name])
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
Folder.find_or_create(db_session, account, raw_folder.display_name,
raw_folder.role)
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self, folders=set()):
new_folders = [f for f in self.prepare_sync() if f not in folders]
for folder_name, folder_id in new_folders:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
thread = self.sync_engine_class(self.account_id,
folder_name,
folder_id,
self.email_address,
self.provider_name,
self.poll_frequency,
self.syncmanager_lock,
self.refresh_flags_max,
self.retry_fail_classes)
self.folder_monitors.start(thread)
while not thread_polling(thread) and \
not thread_finished(thread) and \
not thread.ready():
sleep(self.heartbeat)
# allow individual folder sync monitors to shut themselves down
# after completing the initial sync
if thread_finished(thread) or thread.ready():
if thread.exception:
# Exceptions causing the folder sync to exit should not
# clear the heartbeat.
log.info('Folder sync engine exited with error',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name,
error=thread.exception)
else:
log.info('Folder sync engine finished',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
# clear the heartbeat for this folder-thread since it
# exited cleanly.
clear_heartbeat_status(self.account_id, folder_id)
# note: thread is automatically removed from
# self.folder_monitors
else:
folders.add((folder_name, folder_id))
def start_delete_handler(self):
self.delete_handler = DeleteHandler(account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def sync(self):
try:
self.start_delete_handler()
folders = set()
self.start_new_folder_sync_engines(folders)
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines(folders)
except ValidationError as exc:
log.error(
'Error authenticating; stopping sync', exc_info=True,
account_id=self.account_id, logstash_tag='mark_invalid')
with mailsync_session_scope() as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(str(exc))
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
"""
def __init__(self, account,
heartbeat=1, refresh_frequency=30):
self.refresh_frequency = refresh_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.saved_remote_folders = None
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
self.delete_handler = None
BaseMailSyncMonitor.__init__(self, account, heartbeat)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of folder names for the folders we want to sync (in order).
"""
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
if self.saved_remote_folders != remote_folders:
with session_scope(self.namespace_id) as db_session:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
return sync_folders
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {f.display_name.rstrip()[:MAX_FOLDER_NAME_LENGTH]
for f in raw_folders}
assert 'inbox' in {f.role for f in raw_folders},\
'Account {} has no detected inbox folder'.\
format(account.email_address)
local_folders = {f.name: f for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info('Folder deleted from remote', account_id=self.account_id,
name=name)
cat = db_session.query(Category).get(
local_folders[name].category_id)
if cat is not None:
db_session.delete(cat)
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
Folder.find_or_create(db_session, account, raw_folder.display_name,
raw_folder.role)
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self):
running_monitors = {monitor.folder_name: monitor for monitor in
self.folder_monitors}
for folder_name in self.prepare_sync():
if folder_name in running_monitors:
thread = running_monitors[folder_name]
else:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_name=folder_name)
thread = self.sync_engine_class(self.account_id,
self.namespace_id,
folder_name,
self.email_address,
self.provider_name,
self.syncmanager_lock)
self.folder_monitors.start(thread)
while not thread_polling(thread) and not thread.ready():
sleep(self.heartbeat)
if thread.ready():
log.info('Folder sync engine exited',
account_id=self.account_id,
folder_name=folder_name,
error=thread.exception)
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
provider_name=self.provider_name,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def sync(self):
try:
self.start_delete_handler()
self.start_new_folder_sync_engines()
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines()
except ValidationError as exc:
log.error(
'Error authenticating; stopping sync', exc_info=True,
account_id=self.account_id, logstash_tag='mark_invalid')
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(str(exc))
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
"""
def __init__(self, account, heartbeat=1, refresh_frequency=30):
self.refresh_frequency = refresh_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.saved_remote_folders = None
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
self.delete_handler = None
BaseMailSyncMonitor.__init__(self, account, heartbeat)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of folder names for the folders we want to sync (in order).
"""
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
if self.saved_remote_folders != remote_folders:
with session_scope(self.namespace_id) as db_session:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
return sync_folders
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {
sanitize_name(f.display_name)
for f in raw_folders
}
assert "inbox" in {f.role
for f in raw_folders
}, "Account {} has no detected inbox folder".format(
account.email_address)
local_folders = {
f.name: f
for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)
}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info("Folder deleted from remote",
account_id=self.account_id,
name=name)
if local_folders[name].category_id is not None:
cat = db_session.query(Category).get(
local_folders[name].category_id)
if cat is not None:
db_session.delete(cat)
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
folder = Folder.find_or_create(db_session, account,
raw_folder.display_name,
raw_folder.role)
if folder.canonical_name != raw_folder.role:
folder.canonical_name = raw_folder.role
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self):
running_monitors = {
monitor.folder_name: monitor
for monitor in self.folder_monitors
}
for folder_name in self.prepare_sync():
if folder_name in running_monitors:
thread = running_monitors[folder_name]
else:
log.info(
"Folder sync engine started",
account_id=self.account_id,
folder_name=folder_name,
)
thread = self.sync_engine_class(
self.account_id,
self.namespace_id,
folder_name,
self.email_address,
self.provider_name,
self.syncmanager_lock,
)
self.folder_monitors.start(thread)
while thread.state != "poll" and not thread.ready():
sleep(self.heartbeat)
if thread.ready():
log.info(
"Folder sync engine exited",
account_id=self.account_id,
folder_name=folder_name,
error=thread.exception,
)
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
provider_name=self.provider_name,
uid_accessor=lambda m: m.imapuids,
)
self.delete_handler.start()
def sync(self):
try:
self.start_delete_handler()
self.start_new_folder_sync_engines()
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines()
except ValidationError as exc:
log.error(
"Error authenticating; stopping sync",
exc_info=True,
account_id=self.account_id,
logstash_tag="mark_invalid",
)
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(exc)
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
poll_frequency: Integer
Seconds to wait between polling for the greenlets spawned
refresh_flags_max: Integer
the maximum number of UIDs for which we'll check flags
periodically.
"""
def __init__(self,
account,
heartbeat=1,
refresh_frequency=30,
poll_frequency=30,
retry_fail_classes=[],
refresh_flags_max=2000):
self.refresh_frequency = refresh_frequency
self.poll_frequency = poll_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.refresh_flags_max = refresh_flags_max
self.saved_remote_folders = None
provider_supports_condstore = account.provider_info.get(
'condstore', False)
account_supports_condstore = getattr(account, 'supports_condstore',
False)
if provider_supports_condstore or account_supports_condstore:
self.sync_engine_class = CondstoreFolderSyncEngine
else:
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
BaseMailSyncMonitor.__init__(self, account, heartbeat,
retry_fail_classes)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of tuples (folder_name, folder_id) for each folder we want to sync
(in order).
"""
with mailsync_session_scope() as db_session:
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
if self.saved_remote_folders != remote_folders:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
sync_folder_names_ids = []
for folder_name in sync_folders:
try:
id_, = db_session.query(Folder.id). \
filter(Folder.name == folder_name,
Folder.account_id == self.account_id).one()
sync_folder_names_ids.append((folder_name, id_))
except NoResultFound:
log.error('Missing Folder object when starting sync',
folder_name=folder_name)
raise MailsyncError(
"Missing Folder '{}' on account {}".format(
folder_name, self.account_id))
return sync_folder_names_ids
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {
f.display_name.rstrip()[:MAX_FOLDER_NAME_LENGTH]
for f in raw_folders
}
assert 'inbox' in {f.role for f in raw_folders},\
'Account {} has no detected inbox folder'.\
format(account.email_address)
local_folders = {
f.name: f
for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)
}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info('Folder deleted from remote',
account_id=self.account_id,
name=name)
db_session.delete(local_folders[name])
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
Folder.find_or_create(db_session, account, raw_folder.display_name,
raw_folder.role)
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self, folders=set()):
new_folders = [f for f in self.prepare_sync() if f not in folders]
for folder_name, folder_id in new_folders:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
thread = self.sync_engine_class(
self.account_id, folder_name, folder_id, self.email_address,
self.provider_name, self.poll_frequency, self.syncmanager_lock,
self.refresh_flags_max, self.retry_fail_classes)
self.folder_monitors.start(thread)
while not thread_polling(thread) and \
not thread_finished(thread) and \
not thread.ready():
sleep(self.heartbeat)
# allow individual folder sync monitors to shut themselves down
# after completing the initial sync
if thread_finished(thread) or thread.ready():
if thread.exception:
# Exceptions causing the folder sync to exit should not
# clear the heartbeat.
log.info('Folder sync engine exited with error',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name,
error=thread.exception)
else:
log.info('Folder sync engine finished',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
# clear the heartbeat for this folder-thread since it
# exited cleanly.
clear_heartbeat_status(self.account_id, folder_id)
# note: thread is automatically removed from
# self.folder_monitors
else:
folders.add((folder_name, folder_id))
def start_delete_handler(self):
self.delete_handler = DeleteHandler(account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def sync(self):
try:
self.start_delete_handler()
folders = set()
self.start_new_folder_sync_engines(folders)
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines(folders)
except ValidationError as exc:
log.error('Error authenticating; stopping sync',
exc_info=True,
account_id=self.account_id,
logstash_tag='mark_invalid')
with mailsync_session_scope() as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(str(exc))
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
poll_frequency: Integer
Seconds to wait between polling for the greenlets spawned
refresh_flags_max: Integer
the maximum number of UIDs for which we'll check flags
periodically.
"""
def __init__(self, account,
heartbeat=1, refresh_frequency=30, poll_frequency=30,
retry_fail_classes=[], refresh_flags_max=2000):
self.refresh_frequency = refresh_frequency
self.poll_frequency = poll_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.refresh_flags_max = refresh_flags_max
provider_supports_condstore = account.provider_info.get('condstore',
False)
account_supports_condstore = getattr(account, 'supports_condstore',
False)
if provider_supports_condstore or account_supports_condstore:
self.sync_engine_class = CondstoreFolderSyncEngine
else:
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
BaseMailSyncMonitor.__init__(self, account, heartbeat,
retry_fail_classes)
@retry_crispin
def prepare_sync(self):
"""Ensures that canonical tags are created for the account, and gets
and save Folder objects for folders on the IMAP backend. Returns a list
of tuples (folder_name, folder_id) for each folder we want to sync (in
order)."""
with mailsync_session_scope() as db_session:
with _pool(self.account_id).get() as crispin_client:
sync_folders = crispin_client.sync_folders()
save_folder_names(log, self.account_id,
crispin_client.folder_names(), db_session)
sync_folder_names_ids = []
for folder_name in sync_folders:
try:
id_, = db_session.query(Folder.id). \
filter(Folder.name == folder_name,
Folder.account_id == self.account_id).one()
sync_folder_names_ids.append((folder_name, id_))
except NoResultFound:
log.error("Missing Folder object when starting sync",
folder_name=folder_name)
raise MailsyncError("Missing Folder '{}' on account {}"
.format(folder_name, self.account_id))
return sync_folder_names_ids
def start_new_folder_sync_engines(self, folders=set()):
new_folders = [f for f in self.prepare_sync() if f not in folders]
for folder_name, folder_id in new_folders:
log.info('Folder sync engine started',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
thread = self.sync_engine_class(self.account_id,
folder_name,
folder_id,
self.email_address,
self.provider_name,
self.poll_frequency,
self.syncmanager_lock,
self.refresh_flags_max,
self.retry_fail_classes)
self.folder_monitors.start(thread)
while not thread_polling(thread) and \
not thread_finished(thread) and \
not thread.ready():
sleep(self.heartbeat)
# allow individual folder sync monitors to shut themselves down
# after completing the initial sync
if thread_finished(thread) or thread.ready():
log.info('Folder sync engine finished/killed',
account_id=self.account_id,
folder_id=folder_id,
folder_name=folder_name)
# note: thread is automatically removed from
# self.folder_monitors
else:
folders.add((folder_name, folder_id))
def start_delete_handler(self):
self.delete_handler = DeleteHandler(account_id=self.account_id,
namespace_id=self.namespace_id,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def sync(self):
self.start_delete_handler()
folders = set()
self.start_new_folder_sync_engines(folders)
while True:
sleep(self.refresh_frequency)
self.start_new_folder_sync_engines(folders)
class ImapSyncMonitor(BaseMailSyncMonitor):
"""
Top-level controller for an account's mail sync. Spawns individual
FolderSync greenlets for each folder.
Parameters
----------
heartbeat: Integer
Seconds to wait between checking on folder sync threads.
(DEPRECATED) refresh_frequency: Integer
Seconds to wait between checking for new folders to sync.
syncback_frequency: Integer
Seconds to wait between performing consecutive syncback iterations and
checking for new folders to sync.
"""
def __init__(self, account, heartbeat=1, refresh_frequency=30,
syncback_frequency=5):
# DEPRECATED.
# TODO[k]: Remove after sync-syncback integration deploy is complete.
self.refresh_frequency = refresh_frequency
self.syncmanager_lock = BoundedSemaphore(1)
self.saved_remote_folders = None
self.sync_engine_class = FolderSyncEngine
self.folder_monitors = Group()
self.delete_handler = None
self.syncback_handler = None
self.folder_sync_signals = {}
self.syncback_timestamp = None
self.syncback_frequency = syncback_frequency
BaseMailSyncMonitor.__init__(self, account, heartbeat)
@retry_crispin
def prepare_sync(self):
"""
Gets and save Folder objects for folders on the IMAP backend. Returns a
list of folder names for the folders we want to sync (in order).
"""
with connection_pool(self.account_id).get() as crispin_client:
# Get a fresh list of the folder names from the remote
remote_folders = crispin_client.folders()
# The folders we should be syncing
sync_folders = crispin_client.sync_folders()
if self.saved_remote_folders != remote_folders:
with session_scope(self.namespace_id) as db_session:
self.save_folder_names(db_session, remote_folders)
self.saved_remote_folders = remote_folders
return sync_folders
def save_folder_names(self, db_session, raw_folders):
"""
Save the folders present on the remote backend for an account.
* Create Folder objects.
* Delete Folders that no longer exist on the remote.
Notes
-----
Generic IMAP uses folders (not labels).
Canonical folders ('inbox') and other folders are created as Folder
objects only accordingly.
We don't canonicalize folder names to lowercase when saving because
different backends may be case-sensitive or otherwise - code that
references saved folder names should canonicalize if needed when doing
comparisons.
"""
account = db_session.query(Account).get(self.account_id)
remote_folder_names = {f.display_name.rstrip()[:MAX_FOLDER_NAME_LENGTH]
for f in raw_folders}
assert 'inbox' in {f.role for f in raw_folders},\
'Account {} has no detected inbox folder'.\
format(account.email_address)
local_folders = {f.name: f for f in db_session.query(Folder).filter(
Folder.account_id == self.account_id)}
# Delete folders no longer present on the remote.
# Note that the folder with canonical_name='inbox' cannot be deleted;
# remote_folder_names will always contain an entry corresponding to it.
discard = set(local_folders) - remote_folder_names
for name in discard:
log.info('Folder deleted from remote', account_id=self.account_id,
name=name)
if local_folders[name].category_id is not None:
cat = db_session.query(Category).get(
local_folders[name].category_id)
if cat is not None:
db_session.delete(cat)
del local_folders[name]
# Create new folders
for raw_folder in raw_folders:
Folder.find_or_create(db_session, account, raw_folder.display_name,
raw_folder.role)
# Set the should_run bit for existing folders to True (it's True by
# default for new ones.)
for f in local_folders.values():
if f.imapsyncstatus:
f.imapsyncstatus.sync_should_run = True
db_session.commit()
def start_new_folder_sync_engines(self):
running_monitors = {monitor.folder_name: monitor for monitor in
self.folder_monitors}
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).options(
load_only('_sync_status')).get(self.account_id)
s3_resync = account._sync_status.get('s3_resync', False)
for folder_name in self.prepare_sync():
if folder_name in running_monitors:
thread = running_monitors[folder_name]
else:
log.info('Folder sync engine started',
account_id=self.account_id, folder_name=folder_name)
self._add_sync_signal(folder_name)
thread = self.sync_engine_class(self.account_id,
self.namespace_id,
folder_name,
self.email_address,
self.provider_name,
self.syncmanager_lock,
self.folder_sync_signals[folder_name])
self.folder_monitors.start(thread)
if s3_resync:
log.info('Starting an S3 monitor',
account_id=self.account_id)
s3_thread = S3FolderSyncEngine(self.account_id,
self.namespace_id,
folder_name,
self.email_address,
self.provider_name,
self.syncmanager_lock,
None)
self.folder_monitors.start(s3_thread)
while not thread.state == 'poll' and not thread.ready():
sleep(self.heartbeat)
self.perform_syncback()
if thread.ready():
self._remove_sync_signal[folder_name]
log.info('Folder sync engine exited',
account_id=self.account_id,
folder_name=folder_name,
error=thread.exception)
def start_delete_handler(self):
if self.delete_handler is None:
self.delete_handler = DeleteHandler(
account_id=self.account_id,
namespace_id=self.namespace_id,
provider_name=self.provider_name,
uid_accessor=lambda m: m.imapuids)
self.delete_handler.start()
def perform_syncback(self):
"""
Perform syncback for the account.
Syncback is performed iff all folder syncs are paused, and the previous
syncback occurred more than syncback_frequency seconds ago.
The first condition is checked by the call to _can_syncback().
The second condition is needed because if there are a large number of
pending actions during initial sync, it could repeatedly get interrupted
and put on hold for seconds at a time.
"""
from inbox.syncback.base import SyncbackHandler
if not self._can_syncback():
log.info('Skipping syncback', reason='folder syncs running')
return
if (self.syncback_timestamp and
(datetime.utcnow() - self.syncback_timestamp).seconds <
self.syncback_frequency):
log.info('Skipping syncback',
reason='last syncback < syncback_frequency seconds ago',
syncback_frequency=self.syncback_frequency)
# Reset here so syncs can proceed
self._signal_syncs()
return
if self.syncback_handler is None:
self.syncback_handler = SyncbackHandler(self.account_id,
self.namespace_id,
self.provider_name)
try:
interval = ((datetime.utcnow() - self.syncback_timestamp).seconds
if self.syncback_timestamp else None)
log.info('Performing syncback', syncback_interval_in_seconds=interval)
self.syncback_handler.send_client_changes()
self.syncback_timestamp = datetime.utcnow()
except Exception:
# Log, set self.folder_sync_signals and then re-raise (so the
# greenlet can be restarted etc.)
log.error('Critical syncback error', exc_info=True)
raise
finally:
# Reset here so syncs can proceed
self._signal_syncs()
def sync(self):
try:
self.start_delete_handler()
self.start_new_folder_sync_engines()
while True:
sleep(self.syncback_frequency)
self.perform_syncback()
self.start_new_folder_sync_engines()
except ValidationError as exc:
log.error(
'Error authenticating; stopping sync', exc_info=True,
account_id=self.account_id, logstash_tag='mark_invalid')
with session_scope(self.namespace_id) as db_session:
account = db_session.query(Account).get(self.account_id)
account.mark_invalid()
account.update_sync_error(str(exc))
def _add_sync_signal(self, folder_name):
self.folder_sync_signals[folder_name] = Event()
self.folder_sync_signals[folder_name].set()
def _remove_sync_signal(self, folder_name):
del self.folder_sync_signals[folder_name]
def _can_syncback(self):
"""
Determine if syncback can occur.
If all folder syncs are paused as indicated by the folder_sync_signals,
returns True. Else, returns False.
"""
return (not self.folder_sync_signals or
all(not signal.is_set() for signal in
self.folder_sync_signals.values()))
def _signal_syncs(self):
""" Indicate that folder syncs can resume. """
for signal in self.folder_sync_signals.values():
signal.set()
