Commit 227953b9 authored by Jim Fulton's avatar Jim Fulton

Simplify MVCC by determining transaction start time using lastTransaction.

This implements: https://github.com/zopefoundation/ZODB/issues/50

Rather than watching invalidations, simply use 1 + the storages
lastTransaction, which is equivalent to but much simpler than waiting
for the first invalidation after a transaction starts.

More importantly, it means we can always use loadBefore and get away
from load.  We no longer have to worry about ordering of invalidations
and load() results.

Much thanks to NEO for pointing the way toward this simplification!

Implementing this initially caused a deadlock, because DB.open()
called Connection.open() while holding a database lock and
Connection.open() now calls IStotage.lastTransaction(), which acquires a
storage lock. (It's not clear that lastTransaction() really needs a
storage lock.)  Meanwhile, IStotage.tpc_finish() calls a DB function
that requires the DB lock while holding the storage lock.  Fixing this
required moving the call to Connection.open() outside the region where
the DB lock was held.

To debug the problem above, I greatly improved lock-debugging
support. Now all of the ZODB code imports Lock, RLock and Condition
from ZODB.utils. If the DEBUG_LOCKING is set to a non-empty value,
then these are wrapped in such a way that debugging information is
printed as they are used. This made spotting the nature of the
deadlock easier.

Of course, a change this basic broke lots of tests. Most of the
breakage arises from the fact that connections now call
lastTransaction on storages at transaction boundaries.  Lots of tests
didn't clean up databases and connections properly.  I fixed many
tests, but ultimately gave up and added some extra cleanup code that
violated transaction-manager underware (and the underware's privates)
to clear transaction synchonizers in test setup and tear-down.  I plan
to add a transaction manager API for this and to use it in a
subsequent PR.

This tests makes database and connection hygiene a bit more important,
especially for tests, because a connection will continue to interact
with storages if it isn't properly closed, which can lead to errors if
the storage is closed.  I chose not to swallow these errors in
Connection, choosing rather to clean up tests.

The thread debugging and test changes make this PR larger than I would
have liked. Apologies in advance to the reviewers.
parent 2ae41705
......@@ -12,12 +12,12 @@
#
##############################################################################
"""ZODB transfer activity monitoring
"""
$Id$"""
import threading
import time
from . import utils
class ActivityMonitor:
"""ZODB load/store activity monitor
......@@ -31,7 +31,7 @@ class ActivityMonitor:
def __init__(self, history_length=3600):
self.history_length = history_length # Number of seconds
self.log = [] # [(time, loads, stores)]
self.trim_lock = threading.Lock()
self.trim_lock = utils.Lock()
def closedConnection(self, conn):
log = self.log
......
......@@ -18,7 +18,6 @@ its use is not recommended. It's still here for historical reasons.
"""
from __future__ import print_function
import threading
import time
import logging
import sys
......@@ -28,10 +27,10 @@ import zope.interface
from persistent.TimeStamp import TimeStamp
import ZODB.interfaces
from ZODB import POSException
from ZODB.utils import z64, oid_repr, byte_ord, byte_chr
from ZODB.UndoLogCompatible import UndoLogCompatible
from ZODB._compat import dumps, _protocol, py2_hasattr
from . import POSException, utils
from .utils import z64, oid_repr, byte_ord, byte_chr
from .UndoLogCompatible import UndoLogCompatible
from ._compat import dumps, _protocol, py2_hasattr
log = logging.getLogger("ZODB.BaseStorage")
......@@ -85,8 +84,8 @@ class BaseStorage(UndoLogCompatible):
log.debug("create storage %s", self.__name__)
# Allocate locks:
self._lock = threading.RLock()
self.__commit_lock = threading.Lock()
self._lock = utils.RLock()
self.__commit_lock = utils.Lock()
# Comment out the following 4 lines to debug locking:
self._lock_acquire = self._lock.acquire
......@@ -108,45 +107,6 @@ class BaseStorage(UndoLogCompatible):
else:
self._oid = oid
########################################################################
# The following methods are normally overridden on instances,
# except when debugging:
def _lock_acquire(self, *args):
f = sys._getframe(1)
sys.stdout.write("[la(%s:%s)\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
self._lock.acquire(*args)
sys.stdout.write("la(%s:%s)]\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
def _lock_release(self, *args):
f = sys._getframe(1)
sys.stdout.write("[lr(%s:%s)\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
self._lock.release(*args)
sys.stdout.write("lr(%s:%s)]\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
def _commit_lock_acquire(self, *args):
f = sys._getframe(1)
sys.stdout.write("[ca(%s:%s)\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
self.__commit_lock.acquire(*args)
sys.stdout.write("ca(%s:%s)]\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
def _commit_lock_release(self, *args):
f = sys._getframe(1)
sys.stdout.write("[cr(%s:%s)\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
self.__commit_lock.release(*args)
sys.stdout.write("cr(%s:%s)]\n" % (f.f_code.co_filename, f.f_lineno))
sys.stdout.flush()
#
########################################################################
def sortKey(self):
"""Return a string that can be used to sort storage instances.
......
......@@ -12,13 +12,11 @@
#
##############################################################################
"""Database connection support
$Id$"""
"""
from __future__ import print_function
import logging
import sys
import tempfile
import threading
import warnings
import os
import time
......@@ -54,6 +52,7 @@ import six
global_reset_counter = 0
noop = lambda : None
def resetCaches():
"""Causes all connection caches to be reset as connections are reopened.
......@@ -184,7 +183,7 @@ class Connection(ExportImport, object):
# type of an oid is str. TODO: remove the related now-unnecessary
# critical sections (if any -- this needs careful thought).
self._inv_lock = threading.Lock()
self._inv_lock = utils.Lock()
self._invalidated = set()
# Flag indicating whether the cache has been invalidated:
......@@ -253,7 +252,17 @@ class Connection(ExportImport, object):
if obj is not None:
return obj
p, serial = self._storage.load(oid, '')
before = self.before
if before is None:
# Normal case
before = self._txn_time
data = self._storage.loadBefore(oid, before)
if data is None:
# see the comment in setstate
raise ReadConflictError()
p, _, _ = data
obj = self._reader.getGhost(p)
# Avoid infiniate loop if obj tries to load its state before
......@@ -348,17 +357,9 @@ class Connection(ExportImport, object):
if self.before is not None:
# This is a historical connection. Invalidations are irrelevant.
return
self._inv_lock.acquire()
try:
if self._txn_time is None:
self._txn_time = tid
elif (tid is not None) and (tid < self._txn_time):
raise AssertionError("invalidations out of order, %r < %r"
% (tid, self._txn_time))
with self._inv_lock:
self._invalidated.update(oids)
finally:
self._inv_lock.release()
def invalidateCache(self):
self._inv_lock.acquire()
......@@ -404,7 +405,6 @@ class Connection(ExportImport, object):
def sync(self):
"""Manually update the view on the database."""
self.transaction_manager.abort()
self._storage_sync()
def getDebugInfo(self):
"""Returns a tuple with different items for debugging the
......@@ -534,7 +534,6 @@ class Connection(ExportImport, object):
invalidated = dict.fromkeys(self._invalidated)
self._invalidated = set()
self._txn_time = None
if self._invalidatedCache:
self._invalidatedCache = False
invalidated = self._cache.cache_data.copy()
......@@ -825,18 +824,18 @@ class Connection(ExportImport, object):
# We don't do anything before a commit starts.
pass
# Call the underlying storage's sync() method (if any), and process
# pending invalidations regardless. Of course this should only be
# called at transaction boundaries.
def _storage_sync(self, *ignored):
def newTransaction(self, transaction=None):
self._readCurrent.clear()
sync = getattr(self._storage, 'sync', 0)
if sync:
sync()
getattr(self._storage, 'sync', noop)()
if self.opened:
self._txn_time = p64(u64(self._storage.lastTransaction()) + 1)
# Nope that we flush invalidation *after* setting transaction
# time, because invalidating persistent classes causes data to
# be loaded.
self._flush_invalidations()
afterCompletion = _storage_sync
newTransaction = _storage_sync
afterCompletion = newTransaction
# Transaction-manager synchronization -- ISynchronizer
##########################################################################
......@@ -866,63 +865,26 @@ class Connection(ExportImport, object):
raise
try:
self._setstate(obj, oid)
except ConflictError:
raise
except:
self._log.exception("Couldn't load state for %s %s",
className(obj), oid_repr(oid))
raise
def _setstate(self, obj, oid):
# Helper for setstate(), which provides logging of failures.
# We accept the oid param, which must be the same as obj._p_oid,
# as a performance optimization for the pure-Python persistent implementation
# where accessing an attribute involves __getattribute__ calls
# The control flow is complicated here to avoid loading an
# object revision that we are sure we aren't going to use. As
# a result, invalidation tests occur before and after the
# load. We can only be sure about invalidations after the
# load.
# If an object has been invalidated, among the cases to consider:
# - Try MVCC
# - Raise ConflictError.
if self.before is not None:
# Load data that was current before the time we have.
before = self.before
t = self._storage.loadBefore(oid, before)
if t is None:
raise POSKeyError() # historical connection!
p, serial, end = t
else:
# There is a harmless data race with self._invalidated. A
# dict update could go on in another thread, but we don't care
# because we have to check again after the load anyway.
if before is None:
# Normal case
if self._invalidatedCache:
raise ReadConflictError()
before = self._txn_time
data = self._storage.loadBefore(oid, before)
if data is None:
# We had data (by definition, since we have a
# reference to it), but it's gone. It must have
# updated since this transaction, and been packed
# away, cuz the tests are mean. The best we can do is
# raise a ReadConflictError and try again
raise ReadConflictError()
if (oid in self._invalidated):
self._load_before_or_conflict(obj)
return
p, serial = self._storage.load(oid, '')
p, serial, _ = data
self._load_count += 1
self._inv_lock.acquire()
try:
invalid = oid in self._invalidated
finally:
self._inv_lock.release()
if invalid:
self._load_before_or_conflict(obj)
return
self._reader.setGhostState(obj, p)
obj._p_serial = serial
self._cache.update_object_size_estimation(oid, len(p))
......@@ -933,42 +895,12 @@ class Connection(ExportImport, object):
obj._p_blob_uncommitted = None
obj._p_blob_committed = self._storage.loadBlob(oid, serial)
def _load_before_or_conflict(self, obj):
"""Load non-current state for obj or raise ReadConflictError."""
if not self._setstate_noncurrent(obj):
self._register(obj)
self._conflicts[obj._p_oid] = True
raise ReadConflictError(object=obj)
def _setstate_noncurrent(self, obj):
"""Set state using non-current data.
Return True if state was available, False if not.
"""
try:
# Load data that was current before the commit at txn_time.
t = self._storage.loadBefore(obj._p_oid, self._txn_time)
except KeyError:
return False
if t is None:
return False
data, start, end = t
# The non-current transaction must have been written before
# txn_time. It must be current at txn_time, but could have
# been modified at txn_time.
assert start < self._txn_time, (u64(start), u64(self._txn_time))
assert end is not None
assert self._txn_time <= end, (u64(self._txn_time), u64(end))
self._reader.setGhostState(obj, data)
obj._p_serial = start
# MVCC Blob support
if isinstance(obj, Blob):
obj._p_blob_uncommitted = None
obj._p_blob_committed = self._storage.loadBlob(obj._p_oid, start)
return True
except ConflictError:
raise
except:
self._log.exception("Couldn't load state for %s %s",
className(obj), oid_repr(oid))
raise
def register(self, obj):
"""Register obj with the current transaction manager.
......@@ -1044,20 +976,20 @@ class Connection(ExportImport, object):
register for afterCompletion() calls.
"""
self.opened = time.time()
if transaction_manager is None:
transaction_manager = transaction.manager
self.transaction_manager = transaction_manager
transaction_manager.registerSynch(self)
self.opened = time.time()
if self._reset_counter != global_reset_counter:
# New code is in place. Start a new cache.
self._resetCache()
else:
self._flush_invalidations()
transaction_manager.registerSynch(self)
self.newTransaction()
if self._cache is not None:
self._cache.incrgc() # This is a good time to do some GC
......@@ -1083,6 +1015,7 @@ class Connection(ExportImport, object):
self._reader._cache = cache
def _release_resources(self):
assert not self.opened
for c in six.itervalues(self.connections):
if c._mvcc_storage:
if c._storage is not None:
......@@ -1129,7 +1062,7 @@ class Connection(ExportImport, object):
def savepoint(self):
if self._savepoint_storage is None:
tmpstore = TmpStore(self._normal_storage)
tmpstore = TmpStore(self._normal_storage, self._txn_time)
self._savepoint_storage = tmpstore
self._storage = self._savepoint_storage
......@@ -1178,7 +1111,7 @@ class Connection(ExportImport, object):
self._creating.update(src.creating)
for oid in oids:
data, serial = src.load(oid, src)
data, serial, _ = src.loadBefore(oid, self._txn_time)
obj = self._cache.get(oid, None)
if obj is not None:
self._cache.update_object_size_estimation(
......@@ -1251,12 +1184,10 @@ class TmpStore:
"""A storage-like thing to support savepoints."""
def __init__(self, storage):
def __init__(self, storage, before):
self._storage = storage
for method in (
'getName', 'new_oid', 'getSize', 'sortKey', 'loadBefore',
'isReadOnly'
):
self._before = before
for method in 'getName', 'new_oid', 'getSize', 'sortKey', 'isReadOnly':
setattr(self, method, getattr(storage, method))
self._file = tempfile.TemporaryFile(prefix='TmpStore')
......@@ -1278,10 +1209,11 @@ class TmpStore:
remove_committed_dir(self._blob_dir)
self._blob_dir = None
def load(self, oid, version):
def loadBefore(self, oid, before):
assert before == self._before
pos = self.index.get(oid)
if pos is None:
return self._storage.load(oid, '')
return self._storage.loadBefore(oid, before)
self._file.seek(pos)
h = self._file.read(8)
oidlen = u64(h)
......@@ -1291,7 +1223,7 @@ class TmpStore:
h = self._file.read(16)
size = u64(h[8:])
serial = h[:8]
return self._file.read(size), serial
return self._file.read(size), serial, None
def store(self, oid, serial, data, version, transaction):
# we have this funny signature so we can reuse the normal non-commit
......
......@@ -13,13 +13,15 @@
##############################################################################
"""Database objects
"""
from __future__ import print_function
import sys
import threading
import logging
import datetime
import time
import warnings
from . import utils
from ZODB.broken import find_global
from ZODB.utils import z64
from ZODB.Connection import Connection
......@@ -179,6 +181,7 @@ class ConnectionPool(AbstractConnectionPool):
(available and available[0][0] < threshhold)
):
t, c = available.pop(0)
assert not c.opened
self.all.remove(c)
c._release_resources()
......@@ -213,6 +216,7 @@ class ConnectionPool(AbstractConnectionPool):
to_remove = ()
for (t, c) in self.available:
assert not c.opened
if t < threshhold:
to_remove += (c,)
self.all.remove(c)
......@@ -405,7 +409,7 @@ class DB(object):
storage = ZODB.MappingStorage.MappingStorage(**storage_args)
# Allocate lock.
x = threading.RLock()
x = utils.RLock()
self._a = x.acquire
self._r = x.release
......@@ -559,15 +563,17 @@ class DB(object):
# sys.getrefcount(ob) returns. But, in addition to that,
# the cache holds an extra reference on non-ghost objects,
# and we also want to pretend that doesn't exist.
# If we have no way to get a refcount, we return False to symbolize
# that. As opposed to None, this has the advantage of being usable
# as a number (0) in case clients depended on that.
# If we have no way to get a refcount, we return False
# to symbolize that. As opposed to None, this has the
# advantage of being usable as a number (0) in case
# clients depended on that.
detail.append({
'conn_no': cn,
'oid': oid,
'id': id,
'klass': "%s%s" % (module, ob.__class__.__name__),
'rc': rc(ob) - 3 - (ob._p_changed is not None) if rc else False,
'rc': (rc(ob) - 3 - (ob._p_changed is not None)
if rc else False),
'state': ob._p_changed,
#'references': con.references(oid),
})
......@@ -628,7 +634,11 @@ class DB(object):
@self._connectionMap
def _(c):
if c.opened:
c.transaction_manager.abort()
# Note that this will modify out pool, but this is safe, because
# _connectionMap makes a list of the pool to iterate over
c.close()
c.afterCompletion = c.newTransaction = c.close = noop
c._release_resources()
......@@ -753,18 +763,18 @@ class DB(object):
assert result is not None
# open the connection.
result.open(transaction_manager)
# A good time to do some cache cleanup.
# (note we already have the lock)
self.pool.availableGC()
self.historical_pool.availableGC()
return result
finally:
self._r()
result.open(transaction_manager)
return result
def connectionDebugInfo(self):
result = []
t = time.time()
......@@ -986,7 +996,7 @@ class ContextManager:
self.tm.abort()
self.conn.close()
resource_counter_lock = threading.Lock()
resource_counter_lock = utils.Lock()
resource_counter = 0
class TransactionalUndo(object):
......
......@@ -19,11 +19,11 @@ to be layered over a base database.
The base storage must not change.
"""
from __future__ import print_function
import os
import random
import weakref
import tempfile
import threading
import ZODB.BaseStorage
import ZODB.blob
import ZODB.interfaces
......@@ -71,7 +71,7 @@ class DemoStorage(object):
self._issued_oids = set()
self._stored_oids = set()
self._commit_lock = threading.Lock()
self._commit_lock = ZODB.utils.Lock()
self._transaction = None
if name is None:
......@@ -319,6 +319,7 @@ class DemoStorage(object):
@ZODB.utils.locked
def tpc_abort(self, transaction):
if transaction is not self._transaction:
print('WTF', transaction, self._transaction)
return
self._stored_oids = set()
self._transaction = None
......
......@@ -20,7 +20,6 @@ import contextlib
import errno
import logging
import os
import threading
import time
from struct import pack
from struct import unpack
......@@ -32,6 +31,8 @@ from zc.lockfile import LockFile
from zope.interface import alsoProvides
from zope.interface import implementer
from .. import utils
from ZODB.blob import BlobStorageMixin
from ZODB.blob import link_or_copy
from ZODB.blob import remove_committed
......@@ -2046,7 +2047,7 @@ class FilePool:
self.name = file_name
self._files = []
self._out = []
self._cond = threading.Condition()
self._cond = utils.Condition()
@contextlib.contextmanager
def write_lock(self):
......
......@@ -19,7 +19,6 @@ storage without distracting storage details.
import BTrees
import time
import threading
import ZODB.BaseStorage
import ZODB.interfaces
import ZODB.POSException
......@@ -40,10 +39,10 @@ class MappingStorage(object):
self._transactions = BTrees.OOBTree.OOBTree() # {tid->TransactionRecord}
self._ltid = ZODB.utils.z64
self._last_pack = None
_lock = threading.RLock()
_lock = ZODB.utils.RLock()
self._lock_acquire = _lock.acquire
self._lock_release = _lock.release
self._commit_lock = threading.Lock()
self._commit_lock = ZODB.utils.Lock()
self._opened = True
self._transaction = None
self._oid = 0
......
......@@ -29,6 +29,8 @@ import transaction
import zope.interface
import zope.interface.verify
from .. import utils
ZERO = b'\0'*8
class BasicStorage:
......@@ -345,7 +347,7 @@ class BasicStorage:
results = {}
started.wait()
attempts = []
attempts_cond = threading.Condition()
attempts_cond = utils.Condition()
def update_attempts():
with attempts_cond:
......
......@@ -65,6 +65,7 @@ class ZODBClientThread(TestThread):
for i in range(self.commits):
self.commit(d, i)
self.test.assertEqual(sorted(d.keys()), list(range(self.commits)))
conn.close()
def commit(self, d, num):
d[num] = time.time()
......
......@@ -130,3 +130,10 @@ class MVCCMappingStorage(MappingStorage):
MappingStorage.pack(self, t, referencesf, gc)
finally:
self._commit_lock.release()
@ZODB.utils.locked(MappingStorage.opened)
def lastTransaction(self):
if self._transactions:
return self._transactions.maxKey()
else:
return ZODB.utils.z64
......@@ -25,7 +25,8 @@ from ZODB.serialize import referencesf
from ZODB.tests.MinPO import MinPO
from ZODB.tests.MTStorage import TestThread
from ZODB.tests.StorageTestBase import snooze
from ZODB._compat import loads, PersistentPickler, Pickler, Unpickler, BytesIO, _protocol
from ZODB._compat import (loads, PersistentPickler, Pickler, Unpickler,
BytesIO, _protocol)
import transaction
import ZODB.interfaces
import ZODB.tests.util
......@@ -270,6 +271,8 @@ class PackableStorage(PackableStorageBase):
self._sanity_check()
db.close()
def checkPackWhileWriting(self):
self._PackWhileWriting(pack_now=False)
......@@ -312,6 +315,8 @@ class PackableStorage(PackableStorageBase):
self._sanity_check()
db.close()
def checkPackWithMultiDatabaseReferences(self):
databases = {}
db = DB(self._storage, databases=databases, database_name='')
......@@ -327,6 +332,9 @@ class PackableStorage(PackableStorageBase):
db.pack(time.time()+1)
# some valid storages always return 0 for len()
self.assertTrue(len(self._storage) in (0, 1))
conn.close()
otherdb.close()
db.close()
def checkPackAllRevisions(self):
self._initroot()
......@@ -718,7 +726,7 @@ class ClientThread(TestThread):
def __init__(self, db, choices, loop_trip, timer, thread_id):
TestThread.__init__(self)
self.root = db.open().root()
self.db = db
self.choices = choices
self.loop_trip = loop_trip
self.millis = timer.elapsed_millis
......@@ -737,6 +745,8 @@ class ClientThread(TestThread):
def runtest(self):
from random import choice
conn = self.db.open()
root = conn.root()
for j in range(self.loop_trip):
assign_worked = False
......@@ -745,7 +755,7 @@ class ClientThread(TestThread):
try:
index = choice(self.choices)
alist.extend([self.millis(), index])
self.root[index].value = MinPO(j)
root[index].value = MinPO(j)
assign_worked = True
transaction.commit()
alist.append(self.millis())
......@@ -756,6 +766,8 @@ class ClientThread(TestThread):
transaction.abort()
alist.append(assign_worked)
conn.close()
class ElapsedTimer:
def __init__(self, start_time):
self.start_time = start_time
......@@ -776,5 +788,5 @@ def IExternalGC_suite(factory):
return doctest.DocFileSuite(
'IExternalGC.test',
setUp=setup, tearDown=zope.testing.setupstack.tearDown,
setUp=setup, tearDown=ZODB.tests.util.tearDown,
checker=ZODB.tests.util.checker)
......@@ -23,7 +23,7 @@ import sys
import time
import transaction
from ZODB.utils import u64
from ZODB.utils import u64, z64
from ZODB.tests.MinPO import MinPO
from ZODB._compat import PersistentPickler, Unpickler, BytesIO, _protocol
import ZODB.tests.util
......@@ -153,8 +153,8 @@ class StorageTestBase(ZODB.tests.util.TestCase):
self._storage.close()
def tearDown(self):
self._close()
ZODB.tests.util.TestCase.tearDown(self)
self._close()
def _dostore(self, oid=None, revid=None, data=None,
already_pickled=0, user=None, description=None):
......
......@@ -130,4 +130,4 @@ revision as well as the entire directory:
Clean up our blob directory and database:
>>> blob_storage.close()
>>> database.close()
......@@ -49,3 +49,5 @@ writing and expect the file to be in the blob temporary directory::
True
>>> w.close()
>>> database.close()
......@@ -160,3 +160,5 @@ knowledge that the underlying storage's pack method is also called:
>>> blob_storage._blobs_pack_is_in_progress
False
>>> base_storage.pack = base_pack
>>> database.close()
......@@ -61,7 +61,7 @@ While it's boring, it's important to verify that the same relationships
hold if the default pool size is overridden.
>>> handler.clear()
>>> st.close()
>>> db.close()
>>> st = Storage()
>>> PS = 2 # smaller pool size
>>> db = DB(st, pool_size=PS)
......@@ -117,7 +117,7 @@ We can change the pool size on the fly:
Enough of that.
>>> handler.clear()
>>> st.close()
>>> db.close()
More interesting is the stack-like nature of connection reuse. So long as
we keep opening new connections, and keep them alive, all connections
......@@ -256,7 +256,7 @@ Nothing in that last block should have logged any msgs:
If "too many" connections are open, then closing one may kick an older
closed one out of the available connection stack.
>>> st.close()
>>> db.close()
>>> st = Storage()
>>> db = DB(st, pool_size=3)
>>> conns = [db.open() for dummy in range(6)]
......@@ -324,7 +324,7 @@ gc to reclaim the Connection and its cache eventually works, but that can
take "a long time" and caches can hold on to many objects, and limited
resources (like RDB connections), for the duration.
>>> st.close()
>>> db.close()
>>> st = Storage()
>>> db = DB(st, pool_size=2)
>>> conn0 = db.open()
......
......@@ -25,10 +25,11 @@ Make a change locally:
>>> rt = cn.root()
>>> rt['a'] = 1
Sync should not have been called yet.
Sync is called when a connection is open, as that starts a new transaction:
>>> st.sync_called # False before 3.4
False
>>> st.sync_called
True
>>> st.sync_called = False
``sync()`` is called by the Connection's ``afterCompletion()`` hook after the
......@@ -77,7 +78,9 @@ path at some point when serving pages.
>>> rt = cn.root() # make a change
>>> rt['c'] = 3
>>> st.sync_called
False
True
>>> st.sync_called = False
Now ensure that ``cn.afterCompletion() -> st.sync()`` gets called by commit
despite that the `Connection` registered after the transaction began:
......@@ -96,7 +99,8 @@ And try the same thing with a non-threaded transaction manager:
>>> rt = cn.root() # make a change
>>> rt['d'] = 4
>>> st.sync_called
False
True
>>> st.sync_called = False
>>> tm.commit()
>>> st.sync_called
True
......
......@@ -22,7 +22,7 @@ import unittest
import transaction
import ZODB.tests.util
from ZODB.config import databaseFromString
from ZODB.utils import p64
from ZODB.utils import p64, u64, z64
from persistent import Persistent
from zope.interface.verify import verifyObject
from zope.testing import loggingsupport, renormalizing
......@@ -522,16 +522,24 @@ class InvalidationTests(unittest.TestCase):
create one from an int.
>>> cn.invalidate(p64(1), {p1._p_oid: 1})
>>> cn._txn_time
'\x00\x00\x00\x00\x00\x00\x00\x01'
Transaction start times are based on storage's last
transaction. (Previousely, they were based on the first
invalidation seen in a transaction.)
>>> cn._txn_time == p64(u64(db.storage.lastTransaction()) + 1)
True
>>> p1._p_oid in cn._invalidated
True
>>> p2._p_oid in cn._invalidated
False
>>> cn.invalidate(p64(10), {p2._p_oid: 1, p64(76): 1})
>>> cn._txn_time
'\x00\x00\x00\x00\x00\x00\x00\x01'
>>> cn._txn_time == p64(u64(db.storage.lastTransaction()) + 1)
True
>>> p1._p_oid in cn._invalidated
True
>>> p2._p_oid in cn._invalidated
......@@ -560,6 +568,7 @@ class InvalidationTests(unittest.TestCase):
>>> cn._invalidated
set([])
>>> db.close()
"""
def doctest_invalidateCache():
......@@ -1289,6 +1298,9 @@ class StubStorage:
raise TypeError('StubStorage does not support versions.')
return self._data[oid]
def loadBefore(self, oid, tid):
return self._data[oid] + (None, )
def store(self, oid, serial, p, version, transaction):
if version != '':
raise TypeError('StubStorage does not support versions.')
......@@ -1304,6 +1316,9 @@ class StubStorage:
# storage
return None
def lastTransaction(self):
return z64
class TestConnectionInterface(unittest.TestCase):
......
......@@ -125,7 +125,7 @@ def connectionDebugInfo():
... now += .1
... return now
>>> real_time = time.time
>>> if isinstance(time,type):
>>> if isinstance(time, type):
... time.time = staticmethod(faux_time) # Jython
... else:
... time.time = faux_time
......@@ -151,7 +151,7 @@ def connectionDebugInfo():
>>> before
[None, '\x03zY\xd8\xc0m9\xdd', None]
>>> opened
['2008-12-04T20:40:44Z (1.40s)', '2008-12-04T20:40:45Z (0.30s)', None]
['2008-12-04T20:40:44Z (1.30s)', '2008-12-04T20:40:46Z (0.10s)', None]
>>> infos
['test info (2)', ' (0)', ' (0)']
......
......@@ -75,12 +75,12 @@ class DemoStorageTests(
db = DB(self._storage) # creates object 0. :)
self.assertEqual(len(self._storage), 1)
self.assertTrue(self._storage)
conn = db.open()
with db.transaction() as conn:
for i in range(10):
conn.root()[i] = conn.root().__class__()
transaction.commit()
self.assertEqual(len(self._storage), 11)
self.assertTrue(self._storage)
db.close()
def checkLoadBeforeUndo(self):
pass # we don't support undo yet
......
......@@ -35,6 +35,7 @@ from ZODB.tests import ReadOnlyStorage, RecoveryStorage
from ZODB.tests.StorageTestBase import MinPO, zodb_pickle
from ZODB._compat import dump, dumps, _protocol
from . import util
class FileStorageTests(
StorageTestBase.StorageTestBase,
......@@ -696,7 +697,7 @@ def test_suite():
suite.addTest(unittest.makeSuite(klass, "check"))
suite.addTest(doctest.DocTestSuite(
setUp=zope.testing.setupstack.setUpDirectory,
tearDown=zope.testing.setupstack.tearDown,
tearDown=util.tearDown,
checker=ZODB.tests.util.checker))
suite.addTest(ZODB.tests.testblob.storage_reusable_suite(
'BlobFileStorage',
......
......@@ -33,6 +33,7 @@ from ZODB.tests import (
)
class MVCCTests:
def checkClosingNestedDatabasesWorks(self):
# This tests for the error described in
# https://github.com/zopefoundation/ZODB/issues/45
......@@ -42,7 +43,6 @@ class MVCCTests:
db1.close()
db2.close()
def checkCrossConnectionInvalidation(self):
# Verify connections see updated state at txn boundaries.
# This will fail if the Connection doesn't poll for changes.
......
......@@ -38,6 +38,8 @@ __test__ = dict(
>>> list(conn2.root()[0].keys())
[]
>>> db2.close()
>>> db1.close()
""",
)
......
......@@ -64,7 +64,7 @@ Now we see two transactions and two changed objects.
Clean up.
>>> st.close()
>>> db.close()
"""
import re
......@@ -87,6 +87,6 @@ checker = renormalizing.RENormalizing([
def test_suite():
return doctest.DocTestSuite(
setUp=zope.testing.setupstack.setUpDirectory,
tearDown=zope.testing.setupstack.tearDown,
tearDown=ZODB.tests.util.tearDown,
optionflags=doctest.REPORT_NDIFF,
checker=ZODB.tests.util.checker + checker)
......@@ -99,10 +99,13 @@ class MinimalMemoryStorage(BaseStorage, object):
del self._txn
def _finish(self, tid, u, d, e):
with self._lock:
self._lock_acquire()
try:
self._index.update(self._txn.index)
self._cur.update(self._txn.cur())
self._ltid = self._tid
finally:
self._lock_release()
def loadBefore(self, the_oid, the_tid):
# It's okay if loadBefore() is really expensive, because this
......@@ -121,6 +124,9 @@ class MinimalMemoryStorage(BaseStorage, object):
end_tid = None
else:
end_tid = tids[j]
self.hook(the_oid, self._cur[the_oid], '')
return self._index[(the_oid, tid)], tid, end_tid
def loadSerial(self, oid, serial):
......
......@@ -54,6 +54,8 @@ except NameError:
import io
file_type = io.BufferedReader
from . import util
def new_time():
"""Create a _new_ time stamp.
......@@ -334,6 +336,7 @@ class RecoveryBlobStorage(BlobTestBase,
transaction.commit()
self._dst.copyTransactionsFrom(self._storage)
self.compare(self._storage, self._dst)
db.close()
def gc_blob_removes_uncommitted_data():
......@@ -446,7 +449,6 @@ def packing_with_uncommitted_data_non_undoing():
Clean up:
>>> database.close()
"""
def packing_with_uncommitted_data_undoing():
......@@ -545,13 +547,14 @@ def loadblob_tmpstore():
>>> transaction.commit()
>>> blob_oid = root['blob']._p_oid
>>> tid = connection._storage.lastTransaction()
>>> _txn_time = connection._txn_time
Now we open a database with a TmpStore in front:
>>> database.close()
>>> from ZODB.Connection import TmpStore
>>> tmpstore = TmpStore(blob_storage)
>>> tmpstore = TmpStore(blob_storage, _txn_time)
We can access the blob correctly:
......@@ -609,7 +612,7 @@ def do_not_depend_on_cwd():
>>> with conn.root()['blob'].open() as fp: fp.read()
'data'
>>> bs.close()
>>> db.close()
"""
def savepoint_isolation():
......@@ -700,9 +703,11 @@ def savepoint_cleanup():
>>> db.close()
"""
def lp440234_Setting__p_changed_of_a_Blob_w_no_uncomitted_changes_is_noop():
r"""
>>> conn = ZODB.connection('data.fs', blob_dir='blobs')
>>> db = ZODB.DB('data.fs', blob_dir='blobs')
>>> conn = db.open()
>>> blob = ZODB.blob.Blob(b'blah')
>>> conn.add(blob)
>>> transaction.commit()
......@@ -714,7 +719,7 @@ def lp440234_Setting__p_changed_of_a_Blob_w_no_uncomitted_changes_is_noop():
>>> old_serial == blob._p_serial
True
>>> conn.close()
>>> db.close()
"""
def setUp(test):
......@@ -757,7 +762,7 @@ def storage_reusable_suite(prefix, factory,
"blob_connection.txt",
"blob_importexport.txt",
"blob_transaction.txt",
setUp=setup, tearDown=zope.testing.setupstack.tearDown,
setUp=setup, tearDown=util.tearDown,
checker=zope.testing.renormalizing.RENormalizing([
# Py3k renders bytes where Python2 used native strings...
(re.compile(r"^b'"), "'"),
......@@ -780,15 +785,16 @@ def storage_reusable_suite(prefix, factory,
if test_packing:
suite.addTest(doctest.DocFileSuite(
"blob_packing.txt",
setUp=setup, tearDown=zope.testing.setupstack.tearDown,
setUp=setup, tearDown=util.tearDown,
))
suite.addTest(doctest.DocTestSuite(
setUp=setup, tearDown=zope.testing.setupstack.tearDown,
checker = ZODB.tests.util.checker + \
setUp=setup, tearDown=util.tearDown,
checker = (
ZODB.tests.util.checker +
zope.testing.renormalizing.RENormalizing([
(re.compile(r'\%(sep)s\%(sep)s' % dict(sep=os.path.sep)), '/'),
(re.compile(r'\%(sep)s' % dict(sep=os.path.sep)), '/'),
]),
])),
))
def create_storage(self, name='data', blob_dir=None):
......@@ -823,7 +829,7 @@ def test_suite():
"blob_tempdir.txt",
"blobstorage_packing.txt",
setUp=setUp,
tearDown=zope.testing.setupstack.tearDown,
tearDown=util.tearDown,
optionflags=doctest.ELLIPSIS,
checker=ZODB.tests.util.checker,
))
......@@ -831,7 +837,7 @@ def test_suite():
"blob_layout.txt",
optionflags=doctest.ELLIPSIS|doctest.NORMALIZE_WHITESPACE,
setUp=setUp,
tearDown=zope.testing.setupstack.tearDown,
tearDown=util.tearDown,
checker=ZODB.tests.util.checker +
zope.testing.renormalizing.RENormalizing([
(re.compile(r'\%(sep)s\%(sep)s' % dict(sep=os.path.sep)), '/'),
......
......@@ -33,9 +33,17 @@ This note includes doctests that explain how MVCC is implemented (and
test that the implementation is correct). The tests use a
MinimalMemoryStorage that implements MVCC support, but not much else.
***IMPORTANT***: The MVCC approach has changed since these tests were
originally written. The new approach is much simpler because we no
longer call load to get the current state of an object. We call
loadBefore instead, having gotten a transaction time at the start of a
transaction. As a result, the rhythm of the tests is a little odd,
because the probe a complex dance that doesn't exist any more.
>>> from ZODB.tests.test_storage import MinimalMemoryStorage
>>> from ZODB import DB
>>> db = DB(MinimalMemoryStorage())
>>> st = MinimalMemoryStorage()
>>> db = DB(st)
We will use two different connections with different transaction managers
to make sure that the connections act independently, even though they'll
......@@ -59,6 +67,10 @@ Now open a second connection.
>>> tm2 = transaction.TransactionManager()
>>> cn2 = db.open(transaction_manager=tm2)
>>> from ZODB.utils import p64, u64
>>> cn2._txn_time == p64(u64(st.lastTransaction()) + 1)
True
>>> txn_time2 = cn2._txn_time
Connection high-water mark
--------------------------
......@@ -67,22 +79,20 @@ The ZODB Connection tracks a transaction high-water mark, which
bounds the latest transaction id that can be read by the current
transaction and still present a consistent view of the database.
Transactions with ids up to but not including the high-water mark
are OK to read. When a transaction commits, the database sends
invalidations to all the other connections; the invalidation contains
the transaction id and the oids of modified objects. The Connection
stores the high-water mark in _txn_time, which is set to None until
an invalidation arrives.
are OK to read. At the beginning of a transaction, a connection
sets the high-water mark to just over the last transaction time the
storage has seen.
>>> cn = db.open()
>>> print(cn._txn_time)
None
>>> cn._txn_time == p64(u64(st.lastTransaction()) + 1)
True
>>> cn.invalidate(100, dict.fromkeys([1, 2]))
>>> cn._txn_time
100
>>> cn._txn_time == p64(u64(st.lastTransaction()) + 1)
True
>>> cn.invalidate(200, dict.fromkeys([1, 2]))
>>> cn._txn_time
100
>>> cn._txn_time == p64(u64(st.lastTransaction()) + 1)
True
A connection's high-water mark is set to the transaction id taken from
the first invalidation processed by the connection. Transaction ids are
......@@ -95,8 +105,8 @@ but that doesn't work unless an object is modified. sync() will abort
a transaction and process invalidations.
>>> cn.sync()
>>> print(cn._txn_time) # the high-water mark got reset to None
None
>>> cn._txn_time == p64(u64(st.lastTransaction()) + 1)
True
Basic functionality
-------------------
......@@ -109,9 +119,9 @@ will modify "a." The other transaction will then modify "b" and commit.
>>> tm1.get().commit()
>>> txn = db.lastTransaction()
The second connection has its high-water mark set now.
The second connection already has its high-water mark set.
>>> cn2._txn_time == txn
>>> cn2._txn_time == txn_time2
True
It is safe to read "b," because it was not modified by the concurrent
......@@ -153,22 +163,23 @@ ConflictError: database conflict error (oid 0x01, class ZODB.tests.MinPO.MinPO)
This example will demonstrate that we can commit a transaction if we only
modify current revisions.
>>> print(cn2._txn_time)
None
>>> cn2._txn_time == p64(u64(st.lastTransaction()) + 1)
True
>>> txn_time2 = cn2._txn_time
>>> r1 = cn1.root()
>>> r1["a"].value = 3
>>> tm1.get().commit()
>>> txn = db.lastTransaction()
>>> cn2._txn_time == txn
>>> cn2._txn_time == txn_time2
True
>>> r2["b"].value = r2["a"].value + 1
>>> r2["b"].value
3
>>> tm2.get().commit()
>>> print(cn2._txn_time)
None
>>> cn2._txn_time == p64(u64(st.lastTransaction()) + 1)
True
Object cache
------------
......@@ -302,22 +313,18 @@ same things now.
>>> r2["a"].value, r2["b"].value
(42, 43)
>>> db.close()
Late invalidation
-----------------
The combination of ZEO and MVCC adds more complexity. Since
invalidations are delivered asynchronously by ZEO, it is possible for
an invalidation to arrive just after a request to load the invalidated
object is sent. The connection can't use the just-loaded data,
because the invalidation arrived first. The complexity for MVCC is
that it must check for invalidated objects after it has loaded them,
just in case.
The combination of ZEO and MVCC used to add more complexity. That's
why ZODB no-longer calls load. :)
Rather than add all the complexity of ZEO to these tests, the
MinimalMemoryStorage has a hook. We'll write a subclass that will
deliver an invalidation when it loads an object. The hook allows us
to test the Connection code.
deliver an invalidation when it loads (or loadBefore's) an object.
The hook allows us to test the Connection code.
>>> class TestStorage(MinimalMemoryStorage):
... def __init__(self):
......@@ -351,6 +358,12 @@ non-current revision to load.
>>> oid = r1["b"]._p_oid
>>> ts.hooked[oid] = 1
This test is kinda screwy because it depends on an old approach that
has changed. We'll hack the _txn_time to get the original expected
result, even though what's going on now is much simpler.
>>> cn1._txn_time = ts.lastTransaction()
Once the oid is hooked, an invalidation will be delivered the next
time it is activated. The code below activates the object, then
confirms that the hook worked and that the old state was retrieved.
......@@ -367,6 +380,8 @@ True
>>> r1["b"].value
0
>>> db.close()
No earlier revision available
-----------------------------
......@@ -395,14 +410,13 @@ section above, this is no older state to retrieve.
False
>>> r1["b"]._p_state
-1
>>> cn1._txn_time = ts.lastTransaction()
>>> r1["b"]._p_activate()
Traceback (most recent call last):
...
ReadConflictError: database read conflict error (oid 0x02, class ZODB.tests.MinPO.MinPO)
>>> oid in cn1._invalidated
True
>>> ts.count
1
ReadConflictError: database read conflict error
>>> db.close()
"""
import doctest
import re
......
......@@ -61,6 +61,7 @@ checker = renormalizing.RENormalizing([
])
def setUp(test, name='test'):
clear_transaction_syncs()
transaction.abort()
d = tempfile.mkdtemp(prefix=name)
zope.testing.setupstack.register(test, zope.testing.setupstack.rmtree, d)
......@@ -71,7 +72,9 @@ def setUp(test, name='test'):
os.chdir(d)
zope.testing.setupstack.register(test, transaction.abort)
tearDown = zope.testing.setupstack.tearDown
def tearDown(test):
zope.testing.setupstack.tearDown(test)
clear_transaction_syncs()
class TestCase(unittest.TestCase):
......@@ -186,3 +189,18 @@ def mess_with_time(test=None, globs=None, now=1278864701.5):
time.time = staticmethod(faux_time) # jython
else:
time.time = faux_time
def clear_transaction_syncs():
"""Clear data managers registered with the global transaction manager
Many tests don't clean up synchronizer's registered with the
global transaction managers, which can wreak havoc with following
tests, now that connections interact with their storages at
transaction boundaries. We need to make sure that we clear any
registered data managers.
For now, we'll use the transaction manager's
underware. Eventually, an transaction managers need to grow an API
for this.
"""
transaction.manager._synchs.data.clear()
......@@ -11,10 +11,12 @@
# FOR A PARTICULAR PURPOSE
#
##############################################################################
from __future__ import print_function
import os
import struct
import sys
import time
import threading
import warnings
from binascii import hexlify, unhexlify
from struct import pack, unpack
......@@ -308,3 +310,65 @@ class locked(object):
def __call__(self, func):
return Locked(func, preconditions=self.preconditions)
if os.environ.get('DEBUG_LOCKING'):
class Lock:
lock_class = threading.Lock
def __init__(self):
self._lock = self.lock_class()
def pr(self, name, a=None, kw=None):
f = sys._getframe(2)
if f.f_code.co_filename.endswith('ZODB/utils.py'):
f = sys._getframe(3)
f = '%s:%s' % (f.f_code.co_filename, f.f_lineno)
print(id(self), self._lock, threading.get_ident(), f, name,
a if a else '', kw if kw else '')
def acquire(self, *a, **kw):
self.pr('acquire', a, kw)
return self._lock.acquire(*a, **kw)
def release(self):
self.pr('release')
return self._lock.release()
def __enter__(self):
self.pr('acquire')
return self._lock.acquire()
def __exit__(self, *ignored):
self.pr('release')
return self._lock.release()
class RLock(Lock):
lock_class = threading.RLock
class Condition(Lock):
lock_class = threading.Condition
def wait(self, *a, **kw):
self.pr('wait', a, kw)
return self._lock.wait(*a, **kw)
def wait_for(self, *a, **kw):
self.pr('wait_for', a, kw)
return self._lock.wait_for(*a, **kw)
def notify(self, *a, **kw):
self.pr('notify', a, kw)
return self._lock.notify(*a, **kw)
def notify_all(self):
self.pr('notify_all')
return self._lock.notify_all()
notifyAll = notify_all
else:
from threading import Condition, Lock, RLock
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment