def setUp(self):
self.ctx = mocks.FakeContext()
self.cfg = self.ctx.GetConfig("ec_id")
self.rpc = CreateRpcRunnerMock()
self.proc = mcpu.Processor(self.ctx, "ec_id", enable_locks=False)
self.op = self.OpTest()
self.calc_timeout = lambda: 42
def _Query(self, op):
"""Runs the specified opcode and returns the result.
"""
# Queries don't have a job id
proc = mcpu.Processor(self.server.context, None, enable_locks=False)
# TODO: Executing an opcode using locks will acquire them in blocking mode.
# Consider using a timeout for retries.
return proc.ExecOpCode(op, None)
def main():
debug = int(os.environ["GNT_DEBUG"])
logname = pathutils.GetLogFilename("jobs")
utils.SetupLogging(logname, "job-startup", debug=debug)
(job_id, llock, secret_params_serialized) = _SetupJob()
secret_params = ""
if secret_params_serialized:
secret_params_json = serializer.LoadJson(secret_params_serialized)
secret_params = RestorePrivateValueWrapping(secret_params_json)
utils.SetupLogging(logname, "job-%s" % (job_id,), debug=debug)
try:
logging.debug("Preparing the context and the configuration")
context = masterd.GanetiContext(llock)
logging.debug("Registering signal handlers")
cancel = [False]
prio_change = [False]
def _TermHandler(signum, _frame):
logging.info("Killed by signal %d", signum)
cancel[0] = True
signal.signal(signal.SIGTERM, _TermHandler)
def _HupHandler(signum, _frame):
logging.debug("Received signal %d, old flag was %s, will set to True",
signum, mcpu.sighupReceived)
mcpu.sighupReceived[0] = True
signal.signal(signal.SIGHUP, _HupHandler)
def _User1Handler(signum, _frame):
logging.info("Received signal %d, indicating priority change", signum)
prio_change[0] = True
signal.signal(signal.SIGUSR1, _User1Handler)
job = context.jobqueue.SafeLoadJobFromDisk(job_id, False)
job.SetPid(os.getpid())
if secret_params:
for i in range(0, len(secret_params)):
if hasattr(job.ops[i].input, "osparams_secret"):
job.ops[i].input.osparams_secret = secret_params[i]
execfun = mcpu.Processor(context, job_id, job_id).ExecOpCode
proc = _JobProcessor(context.jobqueue, execfun, job)
result = _JobProcessor.DEFER
while result != _JobProcessor.FINISHED:
result = proc()
if result == _JobProcessor.WAITDEP and not cancel[0]:
# Normally, the scheduler should avoid starting a job where the
# dependencies are not yet finalised. So warn, but wait an continue.
logging.warning("Got started despite a dependency not yet finished")
time.sleep(5)
if cancel[0]:
logging.debug("Got cancel request, cancelling job %d", job_id)
r = context.jobqueue.CancelJob(job_id)
job = context.jobqueue.SafeLoadJobFromDisk(job_id, False)
proc = _JobProcessor(context.jobqueue, execfun, job)
logging.debug("CancelJob result for job %d: %s", job_id, r)
cancel[0] = False
if prio_change[0]:
logging.debug("Received priority-change request")
try:
fname = os.path.join(pathutils.LUXID_MESSAGE_DIR, "%d.prio" % job_id)
new_prio = int(utils.ReadFile(fname))
utils.RemoveFile(fname)
logging.debug("Changing priority of job %d to %d", job_id, new_prio)
r = context.jobqueue.ChangeJobPriority(job_id, new_prio)
job = context.jobqueue.SafeLoadJobFromDisk(job_id, False)
proc = _JobProcessor(context.jobqueue, execfun, job)
logging.debug("Result of changing priority of %d to %d: %s", job_id,
new_prio, r)
except Exception: # pylint: disable=W0703
logging.warning("Informed of priority change, but could not"
" read new priority")
prio_change[0] = False
except Exception: # pylint: disable=W0703
logging.exception("Exception when trying to run job %d", job_id)
finally:
logging.debug("Job %d finalized", job_id)
logging.debug("Removing livelock file %s", llock.GetPath())
os.remove(llock.GetPath())
sys.exit(0)