HBASE-20387 ADDENDUM backport findHangingTests.py changes from master to branches-1.

This commit is contained in:
Sean Busbey 2018-08-16 23:55:28 -05:00
parent 18840e9510
commit 8716ac2568
1 changed files with 98 additions and 65 deletions

163
dev-support/flaky-tests/findHangingTests.py Normal file → Executable file
View File

@ -1,4 +1,4 @@
#!/usr/bin/python #!/usr/bin/env python
## ##
# Licensed to the Apache Software Foundation (ASF) under one # Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file # or more contributor license agreements. See the NOTICE file
@ -15,68 +15,101 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
##
# script to find hanging test from Jenkins build output
# usage: ./findHangingTests.py <url of Jenkins build console>
#
import urllib2
import sys
import string
if len(sys.argv) != 2 :
print "ERROR : Provide the jenkins job console URL as the only argument."
exit(1)
print "Fetching " + sys.argv[1]
response = urllib2.urlopen(sys.argv[1])
i = 0;
tests = {}
failed_tests = {}
summary = 0
host = False
patch = False
branch = False
while True:
n = response.readline()
if n == "" :
break
if not host and n.find("Building remotely on") >= 0:
host = True
print n.strip()
continue
if not patch and n.find("Testing patch for ") >= 0:
patch = True
print n.strip()
continue
if not branch and n.find("Testing patch on branch ") >= 0:
branch = True
print n.strip()
continue
if n.find("PATCH APPLICATION FAILED") >= 0:
print "PATCH APPLICATION FAILED"
sys.exit(1)
if summary == 0 and n.find("Running tests.") >= 0:
summary = summary + 1
continue
if summary == 1 and n.find("[INFO] Reactor Summary:") >= 0:
summary = summary + 1
continue
if summary == 2 and n.find("[INFO] Apache HBase ") >= 0:
sys.stdout.write(n)
continue
if n.find("org.apache.hadoop.hbase") < 0:
continue
test_name = string.strip(n[n.find("org.apache.hadoop.hbase"):len(n)])
if n.find("Running org.apache.hadoop.hbase") > -1 :
tests[test_name] = False
if n.find("Tests run:") > -1 :
if n.find("FAILURE") > -1 or n.find("ERROR") > -1:
failed_tests[test_name] = True
tests[test_name] = True
response.close()
print "Printing hanging tests" # pylint: disable=invalid-name
for key, value in tests.iteritems(): # To disable 'invalid constant name' warnings.
if value == False:
print "Hanging test : " + key """
print "Printing Failing tests" # Script to find hanging test from Jenkins build output
for key, value in failed_tests.iteritems(): # usage: ./findHangingTests.py <url of Jenkins build console>
print "Failing test : " + key """
import re
import sys
import requests
# If any of these strings appear in the console output, it's a build one should probably ignore
# for analyzing failed/hanging tests.
BAD_RUN_STRINGS = [
"Slave went offline during the build", # Machine went down, can't do anything about it.
"The forked VM terminated without properly saying goodbye", # JVM crashed.
]
def get_bad_tests(console_url):
"""
Returns [[all tests], [failed tests], [timeout tests], [hanging tests]] if successfully gets
the build information.
If there is error getting console text or if there are blacklisted strings in console text,
then returns None.
"""
response = requests.get(console_url)
if response.status_code != 200:
print "Error getting consoleText. Response = {} {}".format(
response.status_code, response.reason)
return
# All tests: All testcases which were run.
# Hanging test: A testcase which started but never finished.
# Failed test: Testcase which encountered any kind of failure. It can be failing atomic tests,
# timed out tests, etc
# Timeout test: A Testcase which encountered timeout. Naturally, all timeout tests will be
# included in failed tests.
all_tests_set = set()
hanging_tests_set = set()
failed_tests_set = set()
timeout_tests_set = set()
for line in response.content.splitlines():
result1 = re.findall("Running org.apache.hadoop.hbase.(.*)", line)
if len(result1) == 1:
test_case = result1[0]
if test_case in all_tests_set:
print ("ERROR! Multiple tests with same name '{}'. Might get wrong results "
"for this test.".format(test_case))
else:
hanging_tests_set.add(test_case)
all_tests_set.add(test_case)
result2 = re.findall("Tests run:.*?- in org.apache.hadoop.hbase.(.*)", line)
if len(result2) == 1:
test_case = result2[0]
if "FAILURE!" in line:
failed_tests_set.add(test_case)
if test_case not in hanging_tests_set:
print ("ERROR! No test '{}' found in hanging_tests. Might get wrong results "
"for this test. This may also happen if maven is set to retry failing "
"tests.".format(test_case))
else:
hanging_tests_set.remove(test_case)
result3 = re.match("^\\s+(\\w*).*\\sTestTimedOut", line)
if result3:
test_case = result3.group(1)
timeout_tests_set.add(test_case)
for bad_string in BAD_RUN_STRINGS:
if re.match(".*" + bad_string + ".*", line):
print "Bad string found in build:\n > {}".format(line)
print "Result > total tests: {:4} failed : {:4} timedout : {:4} hanging : {:4}".format(
len(all_tests_set), len(failed_tests_set), len(timeout_tests_set), len(hanging_tests_set))
return [all_tests_set, failed_tests_set, timeout_tests_set, hanging_tests_set]
if __name__ == "__main__":
if len(sys.argv) != 2:
print "ERROR : Provide the jenkins job console URL as the only argument."
sys.exit(1)
print "Fetching {}".format(sys.argv[1])
result = get_bad_tests(sys.argv[1])
if not result:
sys.exit(1)
[all_tests, failed_tests, timedout_tests, hanging_tests] = result
print "Found {} hanging tests:".format(len(hanging_tests))
for test in hanging_tests:
print test
print "\n"
print "Found {} failed tests of which {} timed out:".format(
len(failed_tests), len(timedout_tests))
for test in failed_tests:
print "{0} {1}".format(test, ("(Timed Out)" if test in timedout_tests else ""))
print ("\nA test may have had 0 or more atomic test failures before it timed out. So a "
"'Timed Out' test may have other errors too.")