DB- stat() always return DB_LOCK_DEADLOCK, why?

Question

HI,
when I run the test program below(windows or linux) concurrently, only the last launched process instance running properly, the previous instances always failed when executing db.stat(0), a bsddb.db.DBLockDeadlockError exception throwed.
why db_stat(0) can not recover when a DBLockDeadlockError occurred? how i can do? pls help me, thanks.
this only occurred when using hash mode, btree mode always running successfully.
#!/usr/bin/python2.5
import sys, os
from bsddb import db as bdb
import time
def open_db(home, dbtype):
  envflags = bdb.DB_INIT_TXN | bdb.DB_CREATE | bdb.DB_THREAD | \
      bdb.DB_INIT_LOCK | bdb.DB_INIT_MPOOL | bdb.DB_SYSTEM_MEM
  env = bdb.DBEnv()
  env.set_lk_detect(bdb.DB_LOCK_MINWRITE)
  env.open(home, envflags, 0)
  print 'init env ok ...'
  db = bdb.DB(env)
  db.open('test.db', dbtype=dbtype, flags=bdb.DB_CREATE)
  print 'init db ok ...'
  return env, db
def test_bdb(env, db):
  import random, time
  MAX_NUM = 30000
  get_key = lambda : 'KEY_%d' % int(random.random() * MAX_NUM)
  while True:
    try:
      db.put(get_key(), 'VALUE_%f' % random.random())
      print db.get(get_key())
      #st = db.stat(bdb.DB_FAST_STAT)
      # bsddb.db.DBLockDeadlockError: (-30995, 'DB_LOCK_DEADLOCK: Locker killed to resolve a deadlock')
      # when another process launched, always throws this exception, util the post running process terminated
      st = db.stat(0)
      print 'size = %d' % st['ndata']
    except Exception, ex:
      print 'error:%s' % str(ex)
      time.sleep(0.2)
def main():
  dbhome = "dbhome"
  dbtype = bdb.DB_HASH  # hash mode, only one instance can execute db.stat(0) successfully
  # dbtype = bdb.DB_BTREE   # btree always ok
  env, db = open_db(dbhome, dbtype)
  test_bdb(env, db)
  db.close()
  env.close()
if __name__ == '__main__':
  main()Edited by: 903138 on 2011-12-19 上午2:23

Answer 1

Hello,
How is the program handling deadlock? The DB_LOCK_DEADLOCK
error error indicates that there is a deadlock and the thread
of control receiving the DB_LOCK_DEADLOCK error was selected
to discard its locks in order to resolve the problem. When
the application receives a DB_LOCK_DEADLOCK return, the correct
action is to close any cursors involved in the operation and abort any
enclosing transaction. A common course of action is attempt the
transaction again. Is the application currently handling
deadlocks in this way?
Thank you,
Sandra