xxq250
/
homework-jianmu
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Crash_gen tool is now tracking TDengine service exit status

This commit is contained in:
Steven Li 2020-08-01 23:35:36 +00:00
parent c978075497
commit a53aafeef4
1 changed files with 33 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
tests/pytest/crash_gen.py
View File

@ -2294,6 +2294,12 @@ class ServiceManagerThread:
self._thread.daemon = True # thread dies with the program self._thread.daemon = True # thread dies with the program
self._thread.start() self._thread.start()
self._thread2 = threading.Thread(
target=self.svcErrorReader,
args=(self._tdeSubProcess.getStdErr(), self._ipcQueue))
self._thread2.daemon = True # thread dies with the program
self._thread2.start()
# wait for service to start # wait for service to start
for i in range(0, 10): for i in range(0, 10):
time.sleep(1.0) time.sleep(1.0)
@ -2322,12 +2328,12 @@ class ServiceManagerThread:
raise RuntimeError("sub process object missing") raise RuntimeError("sub process object missing")
self._status = MainExec.STATUS_STOPPING self._status = MainExec.STATUS_STOPPING
self._tdeSubProcess.stop() retCode = self._tdeSubProcess.stop()
print("Attempted to stop sub process, got return code: {}".format(retCode))
if self._tdeSubProcess.isRunning(): # still running if self._tdeSubProcess.isRunning(): # still running
print( print("FAILED to stop sub process, it is still running... pid = {}".format(
"FAILED to stop sub process, it is still running... pid = {}".format( self._tdeSubProcess.getPid()))
self._tdeSubProcess.pid))
else: else:
self._tdeSubProcess = None # not running any more self._tdeSubProcess = None # not running any more
self.join() # stop the thread, change the status, etc. self.join() # stop the thread, change the status, etc.
@ -2343,6 +2349,9 @@ class ServiceManagerThread:
self._thread.join() self._thread.join()
self._thread = None self._thread = None
self._status = MainExec.STATUS_STOPPED self._status = MainExec.STATUS_STOPPED
# STD ERR thread
self._thread2.join()
self._thread2 = None
else: else:
print("Joining empty thread, doing nothing") print("Joining empty thread, doing nothing")
@ -2423,6 +2432,10 @@ class ServiceManagerThread:
print("\nNo more output from IO thread managing TDengine service") print("\nNo more output from IO thread managing TDengine service")
out.close() out.close()
def svcErrorReader(self, err: IO, queue):
for line in iter(err.readline, b''):
print("\nTD Svc STDERR: {}".format(line))
class TdeSubProcess: class TdeSubProcess:
def __init__(self): def __init__(self):
@ -2431,9 +2444,15 @@ class TdeSubProcess:
def getStdOut(self): def getStdOut(self):
return self.subProcess.stdout return self.subProcess.stdout
def getStdErr(self):
return self.subProcess.stderr
def isRunning(self): def isRunning(self):
return self.subProcess is not None return self.subProcess is not None
def getPid(self):
return self.subProcess.pid
def getBuildPath(self): def getBuildPath(self):
selfPath = os.path.dirname(os.path.realpath(__file__)) selfPath = os.path.dirname(os.path.realpath(__file__))
if ("community" in selfPath): if ("community" in selfPath):
@ -2469,18 +2488,18 @@ class TdeSubProcess:
os.rename(logPath, logPathSaved) os.rename(logPath, logPathSaved)
# os.mkdir(logPath) # recreate, no need actually, TDengine will auto-create with proper perms # os.mkdir(logPath) # recreate, no need actually, TDengine will auto-create with proper perms
svcCmd = [taosdPath, '-c', cfgPath] svcCmd = [taosdPath, '-c', cfgPath]
svcCmdSingle = "{} -c {}".format(taosdPath, cfgPath) # svcCmdSingle = "{} -c {}".format(taosdPath, cfgPath)
# svcCmd = ['vmstat', '1'] # svcCmd = ['vmstat', '1']
if self.subProcess: # already there if self.subProcess: # already there
raise RuntimeError("Corrupt process state") raise RuntimeError("Corrupt process state")
# print("Starting service: {}".format(svcCmd)) # print("Starting service: {}".format(svcCmd))
self.subProcess = subprocess.Popen( self.subProcess = subprocess.Popen(
# svcCmd, shell=False, svcCmd, shell=False,
svcCmdSingle, shell=True, # capture core dump? # svcCmdSingle, shell=True, # capture core dump?
stdout=subprocess.PIPE, stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
# bufsize=1, # not supported in binary mode # bufsize=1, # not supported in binary mode
close_fds=ON_POSIX close_fds=ON_POSIX
) # had text=True, which interferred with reading EOF ) # had text=True, which interferred with reading EOF
@ -2488,9 +2507,9 @@ class TdeSubProcess:
def stop(self): def stop(self):
if not self.subProcess: if not self.subProcess:
print("Sub process already stopped") print("Sub process already stopped")
return return -1
retCode = self.subProcess.poll() retCode = self.subProcess.poll() # contains real sub process return code
if retCode: # valid return code, process ended if retCode: # valid return code, process ended
self.subProcess = None self.subProcess = None
else: # process still alive, let's interrupt it else: # process still alive, let's interrupt it
@ -2501,11 +2520,15 @@ class TdeSubProcess:
self.subProcess.send_signal(signal.SIGINT) self.subProcess.send_signal(signal.SIGINT)
try: try:
self.subProcess.wait(10) self.subProcess.wait(10)
retCode = self.subProcess.returncode
except subprocess.TimeoutExpired as err: except subprocess.TimeoutExpired as err:
print("Time out waiting for TDengine service process to exit") print("Time out waiting for TDengine service process to exit")
retCode = -3
else: else:
print("TDengine service process terminated successfully from SIG_INT") print("TDengine service process terminated successfully from SIG_INT")
retCode = -4
self.subProcess = None self.subProcess = None
return retCode
class ThreadStacks: # stack info for all threads class ThreadStacks: # stack info for all threads
def __init__(self): def __init__(self):