Commit 1a896e9c authored by Jim Fulton's avatar Jim Fulton

Refactored the connection pool implementation. (I have a feeling that

it could be made simpler still.)
parent 6c88dc5c
......@@ -103,22 +103,21 @@ class AbstractConnectionPool(object):
def getTimeout(self):
return self._timeout
timeout = property(getTimeout, setTimeout)
timeout = property(getTimeout, lambda self, v: self.setTimeout(v))
size = property(getSize, setSize)
size = property(getSize, lambda self, v: self.setSize(v))
class ConnectionPool(AbstractConnectionPool):
def __init__(self, size, timeout=None):
def __init__(self, size, timeout=time()):
super(ConnectionPool, self).__init__(size, timeout)
# A stack of connections available to hand out. This is a subset
# of self.all. push() and repush() add to this, and may remove
# the oldest available connections if the pool is too large.
# pop() pops this stack. There are never more than size entries
# in this stack. The keys are time.time() values of the push or
# repush calls.
self.available = BTrees.OOBTree.Bucket()
# in this stack.
self.available = []
def push(self, c):
"""Register a new available connection.
......@@ -127,10 +126,10 @@ class ConnectionPool(AbstractConnectionPool):
stack even if we're over the pool size limit.
"""
assert c not in self.all
assert c not in self.available.values()
assert c not in self.available
self._reduce_size(strictly_less=True)
self.all.add(c)
self.available[time()] = c
self.available.append((time(), c))
n = len(self.all)
limit = self.size
if n > limit:
......@@ -147,43 +146,42 @@ class ConnectionPool(AbstractConnectionPool):
older available connections.
"""
assert c in self.all
assert c not in self.available.values()
assert c not in self.available
self._reduce_size(strictly_less=True)
self.available[time()] = c
self.available.append((time(), c))
def _reduce_size(self, strictly_less=False):
"""Throw away the oldest available connections until we're under our
target size (strictly_less=False, the default) or no more than that
(strictly_less=True).
"""
if self.timeout is None:
threshhold = None
else:
threshhold = time() - self.timeout
threshhold = time() - self.timeout
target = self.size
if strictly_less:
target -= 1
for t, c in list(self.available.items()):
if (len(self.available) > target or
threshhold is not None and t < threshhold):
del self.available[t]
self.all.remove(c)
# While application code may still hold a reference to `c`,
# there's little useful that can be done with this Connection
# anymore. Its cache may be holding on to limited resources,
# and we replace the cache with an empty one now so that we
# don't have to wait for gc to reclaim it. Note that it's not
# possible for DB.open() to return `c` again: `c` can never be
# in an open state again.
# TODO: Perhaps it would be better to break the reference
# cycles between `c` and `c._cache`, so that refcounting
# reclaims both right now. But if user code _does_ have a
# strong reference to `c` now, breaking the cycle would not
# reclaim `c` now, and `c` would be left in a user-visible
# crazy state.
c._resetCache()
else:
break
available = self.available
while (
(len(available) > target)
or
(available and available[0][0] < threshhold)
):
t, c = available.pop(0)
self.all.remove(c)
# While application code may still hold a reference to `c`,
# there's little useful that can be done with this Connection
# anymore. Its cache may be holding on to limited resources,
# and we replace the cache with an empty one now so that we
# don't have to wait for gc to reclaim it. Note that it's not
# possible for DB.open() to return `c` again: `c` can never be
# in an open state again.
# TODO: Perhaps it would be better to break the reference
# cycles between `c` and `c._cache`, so that refcounting
# reclaims both right now. But if user code _does_ have a
# strong reference to `c` now, breaking the cycle would not
# reclaim `c` now, and `c` would be left in a user-visible
# crazy state.
c._resetCache()
def reduce_size(self):
self._reduce_size()
......@@ -197,7 +195,7 @@ class ConnectionPool(AbstractConnectionPool):
"""
result = None
if self.available:
result = self.available.pop(self.available.maxKey())
_, result = self.available.pop()
# Leave it in self.all, so we can still get at it for statistics
# while it's alive.
assert result in self.all
......@@ -212,19 +210,15 @@ class ConnectionPool(AbstractConnectionPool):
If a connection is no longer viable because it has timed out, it is
garbage collected."""
if self.timeout is None:
threshhold = None
else:
threshhold = time() - self.timeout
for t, c in tuple(self.available.items()):
if threshhold is not None and t < threshhold:
threshhold = time() - self.timeout
for t, c in list(self.available):
if t < threshhold:
del self.available[t]
self.all.remove(c)
c._resetCache()
else:
c.cacheGC()
class KeyedConnectionPool(AbstractConnectionPool):
# this pool keeps track of keyed connections all together. It makes
# it possible to make assertions about total numbers of keyed connections.
......@@ -233,99 +227,68 @@ class KeyedConnectionPool(AbstractConnectionPool):
# see the comments in ConnectionPool for method descriptions.
def __init__(self, size, timeout=None):
def __init__(self, size, timeout=time()):
super(KeyedConnectionPool, self).__init__(size, timeout)
# key: {time.time: connection}
self.available = BTrees.family32.OO.Bucket()
# time.time: key
self.closed = BTrees.family32.OO.Bucket()
self.pools = {}
def setSize(self, v):
self._size = v
for pool in self.pools.values():
pool.setSize(v)
def setTimeout(self, v):
self._timeout = v
for pool in self.pools.values():
pool.setTimeout(v)
def push(self, c, key):
assert c not in self.all
available = self.available.get(key)
if available is None:
available = self.available[key] = BTrees.family32.OO.Bucket()
else:
assert c not in available.values()
self._reduce_size(strictly_less=True)
self.all.add(c)
t = time()
available[t] = c
self.closed[t] = key
n = len(self.all)
limit = self.size
if n > limit:
reporter = logger.warn
if n > 2 * limit:
reporter = logger.critical
reporter("DB.open() has %s open connections with a size "
"of %s", n, limit)
pool = self.pools.get(key)
if pool is None:
pool = self.pools[key] = ConnectionPool(self.size, self.timeout)
pool.push(c)
def repush(self, c, key):
assert c in self.all
self._reduce_size(strictly_less=True)
available = self.available.get(key)
if available is None:
available = self.available[key] = BTrees.family32.OO.Bucket()
else:
assert c not in available.values()
t = time()
available[t] = c
self.closed[t] = key
self.pools[key].repush(c)
def _reduce_size(self, strictly_less=False):
if self.timeout is None:
threshhold = None
else:
threshhold = time() - self.timeout
target = self.size
if strictly_less:
target -= 1
for t, key in tuple(self.closed.items()):
if (len(self.available) > target or
threshhold is not None and t < threshhold):
del self.closed[t]
c = self.available[key].pop(t)
if not self.available[key]:
del self.available[key]
self.all.remove(c)
c._resetCache()
else:
break
for key, pool in list(self.pools.items()):
pool._reduce_size(strictly_less)
if not pool.all:
del self.pools[key]
def reduce_size(self):
self._reduce_size()
def pop(self, key):
result = None
available = self.available.get(key)
if available:
t = available.maxKey()
result = available.pop(t)
del self.closed[t]
if not available:
del self.available[key]
assert result in self.all
return result
pool = self.pools.get(key)
if pool is not None:
return pool.pop()
def map(self, f):
self.all.map(f)
for pool in self.pools.itervalues():
pool.map(f)
def availableGC(self):
if self.timeout is None:
threshhold = None
else:
threshhold = time() - self.timeout
for t, key in tuple(self.closed.items()):
if threshhold is not None and t < threshhold:
del self.closed[t]
c = self.available[key].pop(t)
if not self.available[key]:
del self.available[key]
self.all.remove(c)
c._resetCache()
else:
self.available[key][t].cacheGC()
for key, pool in self.pools.items():
pool.availableGC()
if not pool.all:
del self.pools[key]
@property
def test_all(self):
result = set()
for pool in self.pools.itervalues():
result.update(pool.all)
return frozenset(result)
@property
def test_available(self):
result = []
for pool in self.pools.itervalues():
result.extend(pool.available)
return tuple(result)
def toTimeStamp(dt):
utc_struct = dt.utctimetuple()
......
......@@ -13,11 +13,7 @@ development continues on a "development" head.
A database can be opened historically ``at`` or ``before`` a given transaction
serial or datetime. Here's a simple example. It should work with any storage
that supports ``loadBefore``. Unfortunately that does not include
MappingStorage, so we use a FileStorage instance. Also unfortunately, as of
this writing there is no reliable way to determine if a storage truly
implements loadBefore, or if it simply returns None (as in BaseStorage), other
than reading code.
that supports ``loadBefore``.
We'll begin our example with a fairly standard set up. We
......@@ -28,11 +24,8 @@ We'll begin our example with a fairly standard set up. We
- modify the database again; and
- commit a transaction.
>>> import ZODB.FileStorage
>>> storage = ZODB.FileStorage.FileStorage(
... 'HistoricalConnectionTests.fs', create=True)
>>> import ZODB
>>> db = ZODB.DB(storage)
>>> import ZODB.MappingStorage
>>> db = ZODB.MappingStorage.DB()
>>> conn = db.open()
>>> import persistent.mapping
......@@ -42,14 +35,13 @@ We'll begin our example with a fairly standard set up. We
>>> import transaction
>>> transaction.commit()
We wait for some ttime to pass, and then make some other changes.
We wait for some time to pass, record he time, and then make some other changes.
>>> import time
>>> t = time.time()
>>> while time.time() <= t:
... time.sleep(.001)
>>> import datetime
>>> now = datetime.datetime.utcnow()
......@@ -164,186 +156,80 @@ historical connection should be kept.
>>> db.getHistoricalTimeout()
400
All three of these values can be specified in a ZConfig file. We're using
mapping storage for simplicity, but remember, as we said at the start of this
document, mapping storage will not work for historical connections (and in fact
may seem to work but then fail confusingly) because it does not implement
loadBefore.
All three of these values can be specified in a ZConfig file.
>>> import ZODB.config
>>> db2 = ZODB.config.databaseFromString('''
... <zodb>
... <mappingstorage/>
... historical-pool-size 5
... historical-pool-size 3
... historical-cache-size 1500
... historical-timeout 6m
... </zodb>
... ''')
>>> db2.getHistoricalPoolSize()
5
3
>>> db2.getHistoricalCacheSize()
1500
>>> db2.getHistoricalTimeout()
360
Let's actually look at these values at work by shining some light into what
has been a black box up to now. We'll actually do some white box examination
of what is going on in the database, pools and connections.
Historical connections are held in a single connection pool with mappings
from the ``before`` TID to available connections. First we'll put a new
pool on the database so we have a clean slate.
>>> historical_conn.close()
>>> from ZODB.DB import KeyedConnectionPool
>>> db.historical_pool = KeyedConnectionPool(
... db.historical_pool.size, db.historical_pool.timeout)
Now lets look what happens to the pool when we create and close an historical
connection.
The pool lets us reuse connections. To see this, we'll open some
connections, close them, and then open them again:
>>> pool = db.historical_pool
>>> len(pool.all)
0
>>> len(pool.available)
0
>>> historical_conn = db.open(
... transaction_manager=transaction1, before=serial)
>>> len(pool.all)
1
>>> len(pool.available)
0
>>> historical_conn in pool.all
True
>>> historical_conn.close()
>>> len(pool.all)
1
>>> len(pool.available)
1
>>> pool.available.keys()[0] == serial
True
>>> len(pool.available.values()[0])
1
>>> conns1 = [db2.open(before=serial) for i in range(4)]
>>> _ = [c.close() for c in conns1]
>>> conns2 = [db2.open(before=serial) for i in range(4)]
Now we'll open and close two for the same serial to see what happens to the
data structures.
Now let's look at what we got. The first connection in conns 2 is the
last connection in conns1, because it was the last connection closed.
>>> historical_conn is db.open(
... transaction_manager=transaction1, before=serial)
>>> conns2[0] is conns1[-1]
True
>>> len(pool.all)
1
>>> len(pool.available)
0
>>> transaction2 = transaction.TransactionManager()
>>> historical_conn2 = db.open(
... transaction_manager=transaction2, before=serial)
>>> len(pool.all)
2
>>> len(pool.available)
0
>>> historical_conn2.close()
>>> len(pool.all)
2
>>> len(pool.available)
1
>>> len(pool.available.values()[0])
1
>>> historical_conn.close()
>>> len(pool.all)
2
>>> len(pool.available)
1
>>> len(pool.available.values()[0])
2
If you change the historical cache size, that changes the size of the
persistent cache on our connection.
Also for the next two:
>>> historical_conn._cache.cache_size
2000
>>> db.setHistoricalCacheSize(1500)
>>> historical_conn._cache.cache_size
1500
>>> (conns2[1] is conns1[-2]), (conns2[2] is conns1[-3])
(True, True)
Now let's look at pool sizes. We'll set it to two, then open and close three
connections. We should end up with only two available connections.
But not for the last:
>>> db.setHistoricalPoolSize(2)
>>> conns2[3] is conns1[-4]
False
>>> historical_conn = db.open(
... transaction_manager=transaction1, before=serial)
>>> historical_conn2 = db.open(
... transaction_manager=transaction2, before=serial)
>>> transaction3 = transaction.TransactionManager()
>>> historical_conn3 = db.open(
... transaction_manager=transaction3, at=historical_serial)
>>> len(pool.all)
3
>>> len(pool.available)
0
Because the pool size was set to 3.
>>> historical_conn3.close()
>>> len(pool.all)
3
>>> len(pool.available)
1
>>> len(pool.available.values()[0])
1
Connections are also discarded if they haven't been used in a while.
To see this, let's close two of the connections:
>>> historical_conn2.close()
>>> len(pool.all)
3
>>> len(pool.available)
2
>>> len(pool.available.values()[0])
1
>>> len(pool.available.values()[1])
1
>>> conns2[0].close(); conns2[1].close()
>>> historical_conn.close()
>>> len(pool.all)
2
>>> len(pool.available)
1
>>> len(pool.available.values()[0])
2
We'l also set the historical timeout to be very low:
Notice it dumped the one that was closed at the earliest time.
>>> db2.setHistoricalTimeout(.01)
>>> time.sleep(.1)
>>> conns2[2].close(); conns2[3].close()
Finally, we'll look at the timeout. We'll need to monkeypatch ``time`` for
this. (The funky __import__ of DB is because some ZODB __init__ shenanigans
make the DB class mask the DB module.)
Now, when we open 4 connections:
>>> db.getHistoricalTimeout()
400
>>> import time
>>> delta = 200
>>> def stub_time():
... return time.time() + delta
...
>>> DB_module = __import__('ZODB.DB', globals(), locals(), ['chicken'])
>>> original_time = DB_module.time
>>> DB_module.time = stub_time
>>> conns1 = [db2.open(before=serial) for i in range(4)]
>>> historical_conn = db.open(before=serial)
We'll see that only the last 2 connections from conn2 are in the
result:
>>> len(pool.all)
2
>>> len(pool.available)
1
>>> [c in conns1 for c in conns2]
[False, False, True, True]
A close or an open will do garbage collection on the timed out connections.
>>> delta += 200
>>> historical_conn.close()
If you change the historical cache size, that changes the size of the
persistent cache on our connection.
>>> len(pool.all)
1
>>> len(pool.available)
1
>>> len(pool.available.values()[0])
1
>>> historical_conn._cache.cache_size
2000
>>> db.setHistoricalCacheSize(1500)
>>> historical_conn._cache.cache_size
1500
Invalidations
=============
......
......@@ -239,12 +239,12 @@ Closing connections adds them to the stack:
Closing another one will purge the one with MARKER 0 from the stack
(since it was the first added to the stack):
>>> [c.MARKER for c in pool.available.values()]
>>> [c.MARKER for (t, c) in pool.available]
[0, 1, 2]
>>> conns[0].close() # MARKER 3
>>> len(pool.available), len(pool.all)
(3, 5)
>>> [c.MARKER for c in pool.available.values()]
>>> [c.MARKER for (t, c) in pool.available]
[1, 2, 3]
Similarly for the other two:
......@@ -252,7 +252,7 @@ Similarly for the other two:
>>> conns[1].close(); conns[2].close()
>>> len(pool.available), len(pool.all)
(3, 3)
>>> [c.MARKER for c in pool.available.values()]
>>> [c.MARKER for (t, c) in pool.available]
[3, 4, 5]
Reducing the pool size may also purge the oldest closed connections:
......@@ -260,7 +260,7 @@ Reducing the pool size may also purge the oldest closed connections:
>>> db.setPoolSize(2) # gets rid of MARKER 3
>>> len(pool.available), len(pool.all)
(2, 2)
>>> [c.MARKER for c in pool.available.values()]
>>> [c.MARKER for (t, c) in pool.available]
[4, 5]
Since MARKER 5 is still the last one added to the stack, it will be the
......
......@@ -25,10 +25,7 @@ def setUp(test):
def tearDown(test):
test.globs['db'].close()
test.globs['db2'].close()
test.globs['storage'].close()
# the DB class masks the module because of __init__ shenanigans
DB_module = __import__('ZODB.DB', globals(), locals(), ['chicken'])
DB_module.time = test.globs['original_time']
module.tearDown(test)
ZODB.tests.util.tearDown(test)
......
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