On 03/15/2012 07:20 PM, mitts daki wrote:
> Thanks lot Lucas,
>
> This helps let me try this but have another question
>
> is there a way can I have resource control on this? is it possible?
It's possible, certainly. I was thinking, you could create a multiple
machine test abstraction that knows the number of machines needed to
run it, and keeps a list of the machines already gathered so far. It'd
be something roughly along the lines:
class TestMM(object):
"""
Shell class for a test that might run in multiple machines.
"""
def __init__(self, name, machines_required):
self.name = name
self.machines_required = machines_required
self.hosts = []
self.lock = open("/tmp/test-mm-lock", "w+")
def host_checkin(self, host):
return_value = False
fcntl.lockf(self.lock, fcntl.LOCK_EX)
if not self.ready_to_execute():
self.hosts.append(host)
return_value = True
fcntl.lockf(lockfile, fcntl.LOCK_UN)
return return_value
def ready_to_execute(self):
return not len(self.hosts) < self.machines_required
class TestPool(object):
def __init__(self):
self.available_tests = [ list with 2000 TestMM class instances ]
self.lock = open("/tmp/test-pool-lock", "w+")
def get_test():
fcntl.lockf(self.lock, fcntl.LOCK_EX)
test = self.available_tests.pop()
fcntl.lockf(self.lock, fcntl.LOCK_UN)
return test
test_pool = TestPool()
TIME_WAIT = 10
def run(machine):
host = hosts.create_host(machine)
at = autotest_remote.Autotest(host)
while test_pool.available_tests:
t = test_pool.get_test()
if not t.host_checkin(host):
continue
while not t.ready_to_execute():
sleep TIME_WAIT
at.run_test(t.name)
job.parallel_simple(run, machines)
Again, this is not working code. You still have things to figure out and
adjust, this is just an idea based on the requirements you've given me.
I figured out it was better than explaining things in abstract terms.
You'll have to study better:
* Mechanisms to sync test threads, such as barriers
* Interfaces of functions such as at.run_test
* How to turn your 2000 functional tests into tests runnable by autotest
* Deal with the stupid bugs that might be lurking around on this
sketch I wrote and I did not notice :)
I hope this all was useful,
Lucas
_______________________________________________
Autotest mailing list
Autotest@test.kernel.org
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
