def testMasterKilledWithOneStandalone(self):
yield self.agency.initiate()
yield self.wait_for(self.agency.is_idle, 20)
yield self.wait_for_slave()
self.info('terminating master')
pid = run.get_pid(os.path.curdir)
run.term_pid(pid)
yield self.wait_for_master_gone()
yield self.wait_for_master_back()
# we should have a pid now
yield self.wait_for_pid(self.pid_path)
def testBackupAgency(self):
pid_path = os.path.join(os.path.curdir, "feat.pid")
hostname = self.agency.get_hostname()
yield self.spawn_agency()
yield self.wait_for_pid(pid_path)
def host_descriptor():
def check(host_desc):
return host_desc.instance_id == 1
d = self.db.get_document(hostname)
d.addCallbacks(check, failure.Failure.trap, errbackArgs=(NotFoundError,))
return d
yield self.wait_for(host_descriptor, 5)
yield common.delay(None, 5)
yield self.agency.initiate()
yield self.wait_for_slave()
pid = run.get_pid(os.path.curdir)
run.term_pid(pid)
# now cleanup the stale descriptors the way the monitor agent would
yield self.wait_for_master()
yield host_restart.do_cleanup(self.db, hostname)
def has_host():
m = self.agency._get_host_medium()
return m is not None and m.is_ready()
yield self.wait_for(has_host, 15)
host_desc = yield self.db.get_document(hostname)
# for host agent the instance id should not increase
# (this is only the case for agents run by host agent)
self.assertEqual(1, host_desc.instance_id)
yield self.wait_for_backup()
slave = self.agency._broker.slaves.values()[0]
self.info("killing slave %s", slave.slave_id)
d = slave.callRemote("shutdown", stop_process=True)
self.assertFailure(d, pb.PBConnectionLost)
yield d
yield common.delay(None, 0.5)
yield self.wait_for_backup()
slave2 = self.agency._broker.slaves.values()[0]
self.assertNotEqual(slave.slave_id, slave2.slave_id)
def testBackupAgency(self):
pid_path = os.path.join(os.path.curdir, 'feat.pid')
hostname = unicode(socket.gethostbyaddr(socket.gethostname())[0])
yield self.spawn_agency()
yield self.wait_for_pid(pid_path)
def host_descriptor():
def check(host_desc):
return host_desc.instance_id == 1
d = self.db.get_document(hostname)
d.addCallbacks(check, failure.Failure.trap,
errbackArgs=(NotFoundError, ))
return d
yield self.wait_for(host_descriptor, 5)
yield common.delay(None, 5)
yield self.agency.initiate()
yield self.wait_for_slave()
pid = run.get_pid(os.path.curdir)
run.term_pid(pid)
yield self.wait_for_master()
def has_host():
m = self.agency._get_host_medium()
return m is not None and m.is_ready()
yield self.wait_for(has_host, 10)
host_desc = yield self.db.get_document(hostname)
self.assertEqual(2, host_desc.instance_id)
yield self.wait_for_backup()
slave = self.agency._broker.slaves.values()[0]
self.info('killing slave')
d = slave.callRemote('shutdown', stop_process=True)
self.assertFailure(d, pb.PBConnectionLost)
yield d
yield common.delay(None, 0.5)
yield self.wait_for_backup()
slave2 = self.agency._broker.slaves.values()[0]
self.assertNotEqual(slave.slave_id, slave2.slave_id)
def testLockAlreadyTaken(self):
self.lock_fd = open(self.lock_path, 'rb+')
if not fcntl.lock(self.lock_fd):
self.fail("Could not take the lock")
yield self.agency.initiate()
yield self.wait_for_slave()
pid = run.get_pid(os.path.curdir)
run.term_pid(pid)
yield self.wait_for_master_gone()
yield common.delay(None, 10)
pid = run.get_pid(os.path.curdir)
self.assertTrue(pid is None)
# remove the lock so that the broker in our
# agency can connect and stop retrying, overwise the test
# will finish in undefined way (this is part of the teardown)
fcntl.unlock(self.lock_fd)