Commit 9124f8eb authored by Sebastien Robin's avatar Sebastien Robin Committed by Eteri

Task Distribution: give a chance for test suite to finish when testnodes are missing

parent 0c2a5546
......@@ -475,6 +475,12 @@ class TestTaskDistribution(ERP5TypeTestCase):
checkTestResultLine([('testBar', 'started'), ('testFoo', 'stopped')])
def test_07_reportTaskFailure(self):
"""
When all test nodes report failures, we should mark the test result as
failed. If we do not do so, test node would always pickup same repository
revision and might fail with same failure forever (for example, a slapos
build issue).
"""
test_result_path, revision = self._createTestResult(node_title="Node0")
next_test_result_path, revision = self._createTestResult(node_title="Node1")
self.assertEqual(test_result_path, next_test_result_path)
......@@ -493,6 +499,48 @@ class TestTaskDistribution(ERP5TypeTestCase):
self.assertEqual("failed", test_result.getSimulationState())
checkNodeState("failed", "failed")
def test_07b_reportTaskFailureWithRunningTest(self):
"""
Similar to above test. Though, sometimes there is failure reported only because
runTestSuite reached timeout. This happens when not enough testnode are working
on a very long test suite. So code investigate if tests looked working fine, and
it might try to not cancel test result if there is chance that tests could be
continued.
For example :
- testnode0 start test suite Foo with revision r0 which would take 6 hours (other
testnodes are busy)
- after 4 hours, runTestSuite reach timeout of 4 hours (value set in test nodes).
thus it report a failure. We do not cancel the test result since everything went
fine up to know
- after some time testnode0 come back to run test suite Foo, revision r0, and
just do the 2 remaining hours. Test Suite can go up to the end even if we have
timeout smaller than total time for test suite.
"""
now = DateTime()
try:
self.pinDateTime(now - 1.0/24*2)
test_result_path, revision = self._createTestResult(node_title="Node0",
test_list=['testFoo', 'testBar'])
test_result = self.getPortalObject().unrestrictedTraverse(test_result_path)
self.assertEqual("started", test_result.getSimulationState())
node, = test_result.objectValues(portal_type="Test Result Node",
sort_on=[("title", "ascending")])
self.assertEqual("started", node.getSimulationState())
line_url, test = self.tool.startUnitTest(test_result_path)
# We have a failure but with recent activities on tests
self.pinDateTime(now - 1.0/24*1.5)
self.tool.reportTaskFailure(test_result_path, {}, "Node0")
self.assertEqual("failed", node.getSimulationState())
self.assertEqual("started", test_result.getSimulationState())
# We have a failure but with no recent activities on tests
self.pinDateTime(now)
self.tool.reportTaskFailure(test_result_path, {}, "Node0")
self.assertEqual("failed", node.getSimulationState())
self.assertEqual("failed", test_result.getSimulationState())
finally:
self.unpinDateTime()
def test_08_checkWeCanNotCreateTwoTestResultInParallel(self):
"""
To avoid duplicates of test result when several testnodes works on the
......
......@@ -27,6 +27,7 @@
##############################################################################
import random
from DateTime import DateTime
from AccessControl import ClassSecurityInfo
from Products.ERP5Type import Permissions, PropertySheet, Constraint, interfaces
from Products.ERP5Type.Tool.BaseTool import BaseTool
......@@ -269,6 +270,15 @@ class TaskDistributionTool(BaseTool):
for node in test_result.objectValues(portal_type='Test Result Node'):
if node.getSimulationState() != 'failed':
break
else:
# now check if we had recent work on test line, if so, this means
# we might just add timeout due to too much tests to execute for too
# little nodes. In that case we would like to continue the work later
recent_time = DateTime() - 1.0/24
for test_result_line in test_result.objectValues(
portal_type="Test Result Line"):
if test_result_line.getModificationDate() > recent_time:
break
else:
if test_result.getSimulationState() not in ('failed', 'cancelled'):
test_result.fail()
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment