Hi,

I am involved in the maintenance of an application using ZODB and Zope-2.8.0-final. After some investigation I noticed what could be an old memory leak related to the ObjectWriter (cycle with the pickler) in ZODB, as can be seen running this short program:


import os, os.path
from ZODB import FileStorage, DB
import transaction
import sys
import gc

class ZODBStore:
  def __init__(self, name):
      self.name = name
      self.absPathName = os.path.dirname(os.path.abspath(name))
  def Connect(self):
      self.file = FileStorage.FileStorage(self.name)
      self.db = DB(self.file)
      self.connection = self.db.open()
      self.root = self.connection.root()
      self.transaction = transaction.begin()
  def Disconnect(self):
      self.Commit()
      self.connection.close()
      self.db.close()
      self.file.close()
  def Commit(self):
      self.transaction.commit()
  def Abort(self):
      self.transaction.abort()

if __name__ == "__main__":
  ObjectStore = ZODBStore("test.fs")
  ObjectStore.Connect()
  ObjectStore.root['Employees'] = ['Jack', 'Joe', 'Bob']
  ObjectStore.Disconnect()

  gc.enable()
  gc.set_debug(gc.DEBUG_LEAK)
  resultFile = open('resultConnectionLeak.txt', 'a+')
  gc.collect()
  for x in gc.garbage:
      msg = '>>>> Garbage ' + str(type(x)) + ' ' + str(x) + '\n     '
      msg += x.__class__.__name__ + ' at ' + str(hex(id(x))) + '\n'
      resultFile.write(msg)
resultFile.close() I fixed it by adding a close function to the ObjectWriter in serialize.py and calling it explicitly from connection.py (modification found in the CVS).
def close(self):
  self._p.persistent_id = None
  self._p = None
_______________________________________________
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
http://mail.zope.org/mailman/listinfo/zodb-dev

Reply via email to