def _report_unschedulable(self, unschedulable, err_msg=None):
if not unschedulable:
return
self._log(
logging.ERROR,
f"There are {len(unschedulable)} tests which cannot be run due to insufficient cluster resources"
)
for tc in unschedulable:
if err_msg:
msg = err_msg
else:
msg = f"Test {tc.test_id} requires more resources than are available in the whole cluster. " \
f"{self.cluster.all().nodes.attempt_remove_spec(tc.expected_cluster_spec)}"
self._log(logging.ERROR, msg)
result = TestResult(tc,
self.test_counter,
self.session_context,
test_status=FAIL,
summary=msg,
start_time=time.time(),
stop_time=time.time())
self.results.append(result)
result.report()
self.test_counter += 1
def run_all_tests(self):
self.receiver.start()
self.results.start_time = time.time()
# Report tests which cannot be run
if len(self.scheduler.unschedulable) > 0:
self._log(
logging.ERROR,
"There are %d tests which cannot be run due to insufficient cluster resources"
% len(self.scheduler.unschedulable))
for tc in self.scheduler.unschedulable:
msg = "Test %s expects more nodes than are available in the entire cluster: " % tc.test_id
msg += "expected_num_nodes: %s, " % str(tc.expected_node_spec)
msg += "cluster size: %s." % str(self.cluster.node_spec)
self._log(logging.ERROR, msg)
result = TestResult(tc,
self.test_counter,
self.session_context,
test_status=FAIL,
summary=msg,
start_time=time.time(),
stop_time=time.time())
self.results.append(result)
result.report()
self.test_counter += 1
# Run the tests!
self._log(
logging.INFO, "starting test run with session id %s..." %
self.session_context.session_id)
self._log(logging.INFO, "running %d tests..." % len(self.scheduler))
while self._ready_to_trigger_more_tests or self._expect_client_requests:
try:
while self._ready_to_trigger_more_tests:
next_test_context = self.scheduler.next()
self._preallocate_subcluster(next_test_context)
self._run_single_test(next_test_context)
if self._expect_client_requests:
try:
event = self.receiver.recv()
self._handle(event)
except Exception as e:
err_str = "Exception receiving message: %s: %s" % (str(
type(e)), str(e))
err_str += "\n" + traceback.format_exc(limit=16)
self._log(logging.ERROR, err_str)
# All processes are on the same machine, so treat communication failure as a fatal error
raise
except KeyboardInterrupt:
# If SIGINT is received, stop triggering new tests, and let the currently running tests finish
self._log(
logging.INFO,
"Received KeyboardInterrupt. Now waiting for currently running tests to finish..."
)
self.stop_testing = True
for proc in self._client_procs.values():
proc.join()
self.receiver.close()
return self.results
def run(self):
self.log(logging.INFO, "Loading test %s" % str(self.test_metadata))
self.test_context = self._collect_test_context(**self.test_metadata)
self.test_context.test_index = self.test_index
self.send(self.message.running())
if self.test_context.ignore:
# Skip running this test, but keep track of the fact that we ignored it
result = TestResult(self.test_context,
self.test_index,
self.session_context,
test_status=IGNORE,
start_time=time.time(),
stop_time=time.time())
result.report()
# Tell the server we are finished
self.send(self.message.finished(result=result))
return
# Results from this test, as well as logs will be dumped here
mkdir_p(TestContext.results_dir(self.test_context, self.test_index))
start_time = -1
stop_time = -1
test_status = PASS
summary = ""
data = None
try:
# Instantiate test
self.test = self.test_context.cls(self.test_context)
self.log(logging.DEBUG, "Checking if there are enough nodes...")
min_cluster_spec = self.test.min_cluster_spec()
# Check test resource
msg = self.cluster.all().nodes.attempt_remove_spec(min_cluster_spec)
if len(msg) > 0:
raise RuntimeError("There are not enough nodes available in the cluster to run this test. " + msg)
# Run the test unit
start_time = time.time()
self.setup_test()
data = self.run_test()
test_status = PASS
self.log(logging.INFO, "PASS")
except BaseException as e:
err_trace = str(e) + "\n" + traceback.format_exc(limit=16)
self.log(logging.INFO, "FAIL: " + err_trace)
test_status = FAIL
summary += err_trace
finally:
self.teardown_test(teardown_services=not self.session_context.no_teardown, test_status=test_status)
stop_time = time.time()
if hasattr(self, "services"):
service_errors = self.test_context.services.errors()
if service_errors:
summary += "\n\n" + service_errors
result = TestResult(
self.test_context,
self.test_index,
self.session_context,
test_status,
summary,
data,
start_time,
stop_time)
self.log(logging.INFO, "Summary: %s" % str(result.summary))
self.log(logging.INFO, "Data: %s" % str(result.data))
result.report()
# Tell the server we are finished
self._do_safely(lambda: self.send(self.message.finished(result=result)), "Problem sending FINISHED message:")
# Release test_context resources only after creating the result and finishing logging activity
# The Sender object uses the same logger, so we postpone closing until after the finished message is sent
self.test_context.close()
self.test_context = None
self.test = None
def run(self):
self.log(logging.INFO, "Loading test %s" % str(self.test_metadata))
self.test_context = self._collect_test_context(**self.test_metadata)
self.test_context.test_index = self.test_index
self.send(self.message.running())
if self.test_context.ignore:
# Skip running this test, but keep track of the fact that we ignored it
result = TestResult(self.test_context,
self.test_index,
self.session_context,
test_status=IGNORE,
start_time=time.time(),
stop_time=time.time())
result.report()
# Tell the server we are finished
self.send(self.message.finished(result=result))
return
# Results from this test, as well as logs will be dumped here
mkdir_p(TestContext.results_dir(self.test_context, self.test_index))
start_time = -1
stop_time = -1
test_status = PASS
summary = ""
data = None
try:
# Instantiate test
self.test = self.test_context.cls(self.test_context)
self.log(logging.DEBUG, "Checking if there are enough nodes...")
for (operating_system,
node_count) in self.test.min_cluster_size().iteritems():
if node_count > self.cluster.num_nodes_for_operating_system(
operating_system):
raise RuntimeError(
"There are not enough nodes available in the cluster to run this test. "
"Cluster size for %s: %d, Need at least: %d. Services currently registered: %s"
% (operating_system,
self.cluster.num_nodes_for_operating_system(
operating_system), node_count,
self.test_context.services))
# Run the test unit
start_time = time.time()
self.setup_test()
data = self.run_test()
test_status = PASS
self.log(logging.INFO, "PASS")
except BaseException as e:
err_trace = str(e.message) + "\n" + traceback.format_exc(limit=16)
self.log(logging.INFO, "FAIL: " + err_trace)
test_status = FAIL
summary += err_trace
finally:
self.teardown_test(
teardown_services=not self.session_context.no_teardown,
test_status=test_status)
stop_time = time.time()
result = TestResult(self.test_context, self.test_index,
self.session_context, test_status, summary,
data, start_time, stop_time)
self.log(logging.INFO, "Summary: %s" % str(result.summary))
self.log(logging.INFO, "Data: %s" % str(result.data))
result.report()
# Tell the server we are finished
self._do_safely(
lambda: self.send(self.message.finished(result=result)),
"Problem sending FINISHED message:")
# Release test_context resources only after creating the result and finishing logging activity
# The Sender object uses the same logger, so we postpone closing until after the finished message is sent
self.test_context.close()
self.test_context = None
self.test = None
def run_all_tests(self):
self.receiver.start()
self.results.start_time = time.time()
# Report tests which cannot be run
if len(self.scheduler.unschedulable) > 0:
self._log(logging.ERROR,
"There are %d tests which cannot be run due to insufficient cluster resources" %
len(self.scheduler.unschedulable))
for tc in self.scheduler.unschedulable:
msg = "Test %s requires more resources than are available in the whole cluster. " % tc.test_id
msg += self.cluster.all().nodes.attempt_remove_spec(tc.expected_cluster_spec)
self._log(logging.ERROR, msg)
result = TestResult(
tc,
self.test_counter,
self.session_context,
test_status=FAIL,
summary=msg,
start_time=time.time(),
stop_time=time.time())
self.results.append(result)
result.report()
self.test_counter += 1
# Run the tests!
self._log(logging.INFO, "starting test run with session id %s..." % self.session_context.session_id)
self._log(logging.INFO, "running %d tests..." % len(self.scheduler))
while self._ready_to_trigger_more_tests or self._expect_client_requests:
try:
while self._ready_to_trigger_more_tests:
next_test_context = self.scheduler.next()
self._preallocate_subcluster(next_test_context)
self._run_single_test(next_test_context)
if self._expect_client_requests:
try:
event = self.receiver.recv()
self._handle(event)
except Exception as e:
err_str = "Exception receiving message: %s: %s" % (str(type(e)), str(e))
err_str += "\n" + traceback.format_exc(limit=16)
self._log(logging.ERROR, err_str)
# All processes are on the same machine, so treat communication failure as a fatal error
raise
except KeyboardInterrupt:
# If SIGINT is received, stop triggering new tests, and let the currently running tests finish
self._log(logging.INFO,
"Received KeyboardInterrupt. Now waiting for currently running tests to finish...")
self.stop_testing = True
for proc in self._client_procs.values():
proc.join()
self.receiver.close()
return self.results
def run(self):
self.log(logging.INFO, "Loading test %s" % str(self.test_metadata))
self.test_context = self._collect_test_context(**self.test_metadata)
self.test_context.test_index = self.test_index
self.send(self.message.running())
if self.test_context.ignore:
# Skip running this test, but keep track of the fact that we ignored it
result = TestResult(self.test_context,
self.test_index,
self.session_context,
test_status=IGNORE,
start_time=time.time(),
stop_time=time.time())
result.report()
# Tell the server we are finished
self.send(self.message.finished(result=result))
return
# Results from this test, as well as logs will be dumped here
mkdir_p(TestContext.results_dir(self.test_context, self.test_index))
start_time = -1
stop_time = -1
test_status = PASS
summary = ""
data = None
try:
# Instantiate test
self.test = self.test_context.cls(self.test_context)
self.log(logging.DEBUG, "Checking if there are enough nodes...")
min_cluster_spec = self.test.min_cluster_spec()
os_to_num_nodes = {}
for node_spec in min_cluster_spec:
if not os_to_num_nodes.get(node_spec.operating_system):
os_to_num_nodes[node_spec.operating_system] = 1
else:
os_to_num_nodes[node_spec.operating_system] = os_to_num_nodes[node_spec.operating_system] + 1
for (operating_system, node_count) in os_to_num_nodes.iteritems():
num_avail = len(list(self.cluster.all().nodes.elements(operating_system=operating_system)))
if node_count > num_avail:
raise RuntimeError(
"There are not enough nodes available in the cluster to run this test. "
"Cluster size for %s: %d, Need at least: %d. Services currently registered: %s" %
(operating_system, num_avail, node_count, self.test_context.services))
# Run the test unit
start_time = time.time()
self.setup_test()
data = self.run_test()
test_status = PASS
self.log(logging.INFO, "PASS")
except BaseException as e:
err_trace = str(e.message) + "\n" + traceback.format_exc(limit=16)
self.log(logging.INFO, "FAIL: " + err_trace)
test_status = FAIL
summary += err_trace
finally:
self.teardown_test(teardown_services=not self.session_context.no_teardown, test_status=test_status)
stop_time = time.time()
service_errors = self.test_context.services.errors()
if service_errors:
summary += "\n\n" + service_errors
result = TestResult(
self.test_context,
self.test_index,
self.session_context,
test_status,
summary,
data,
start_time,
stop_time)
self.log(logging.INFO, "Summary: %s" % str(result.summary))
self.log(logging.INFO, "Data: %s" % str(result.data))
result.report()
# Tell the server we are finished
self._do_safely(lambda: self.send(self.message.finished(result=result)), "Problem sending FINISHED message:")
# Release test_context resources only after creating the result and finishing logging activity
# The Sender object uses the same logger, so we postpone closing until after the finished message is sent
self.test_context.close()
self.test_context = None
self.test = None
def run(self):
self.log(logging.INFO, "Loading test %s" % str(self.test_metadata))
self.test_context = self._collect_test_context(**self.test_metadata)
self.test_context.test_index = self.test_index
self.send(self.message.running())
if self.test_context.ignore:
# Skip running this test, but keep track of the fact that we ignored it
result = TestResult(self.test_context,
self.test_index,
self.session_context,
test_status=IGNORE,
start_time=time.time(),
stop_time=time.time())
result.report()
# Tell the server we are finished
self.send(self.message.finished(result=result))
return
start_time = -1
stop_time = -1
test_status = FAIL
summary = []
data = None
self.all_services = ServiceRegistry()
num_runs = 0
try:
while test_status == FAIL and num_runs < self.deflake_num:
num_runs += 1
self.log(logging.INFO, "on run {}/{}".format(num_runs, self.deflake_num))
start_time = time.time()
test_status, summary, data = self._do_run(num_runs)
if test_status == PASS and num_runs > 1:
test_status = FLAKY
msg = str(test_status.to_json())
if summary:
msg += ": {}".format(summary)
if num_runs != self.deflake_num:
msg += "\n" + "~" * max(len(line) for line in summary.split('\n'))
self.log(logging.INFO, msg)
finally:
stop_time = time.time()
test_status, summary = self._check_cluster_utilization(test_status, summary)
if num_runs > 1:
# for reporting purposes report all services
self.test_context.services = self.all_services
# for flaky tests, we report the start and end time of the successful run, and not the whole run period
result = TestResult(
self.test_context,
self.test_index,
self.session_context,
test_status,
summary,
data,
start_time,
stop_time)
self.log(logging.INFO, "Data: %s" % str(result.data))
result.report()
# Tell the server we are finished
self._do_safely(lambda: self.send(self.message.finished(result=result)),
"Problem sending FINISHED message for " + str(self.test_metadata) + ":\n")
# Release test_context resources only after creating the result and finishing logging activity
# The Sender object uses the same logger, so we postpone closing until after the finished message is sent
self.test_context.close()
self.all_services = None
self.test_context = None
self.test = None
