Reland Wait cloudtail termination in goma module

scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py

Index: scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py
diff --git a/scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py b/scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py
index 79f20dc1edd7c1ee81c9d20086e9b363e61c5bb5..46907313b749f7f4972dff7f485186ec2412b35d 100644
--- a/scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py
+++ b/scripts/slave/recipe_modules/goma/resources/cloudtail_utils.py
@@ -9,6 +9,7 @@ import os
 import signal
 import subprocess
 import sys
+import time
 
 from slave import goma_utils
 
@@ -24,6 +25,64 @@ def start_cloudtail(args):
   with open(args.pid_file, 'w') as f:
     f.write(str(proc.pid))
 
+def is_running_posix(pid):

[shinyak, 2016/10/11 06:47:02]

Two blank lines between top-level definitions

https://google.github.io/styleguide/pyguide.html?showone=Blank_Lines#Blank_Lines

[tikuta, 2016/10/11 08:09:00]

Done.

+  """Return True if process of pid is running.
+
+  Args:
+    pid(int): pid of process which this function checks
+              whether it is running or not.
+
+  Returns:
+    bool: True if process of pid is running.
+
+  Raises:
+    OSError if something happens in os.kill(pid, 0)
+  """
+
+  try:
+    os.kill(pid, 0)
+  except OSError as e:
+    if e.errno == errno.ESRCH or e.errno == errno.EPERM:
+      return False
+    os.kill(pid, signal.SIGKILL)

[shinyak, 2016/10/11 06:47:02]

Oh, is_running_posix(pid) sends SIGKILL? I feel it's strange...
I believe it's parent function job.

[tikuta, 2016/10/11 08:09:00]

Done.

+    print('killed process %d due to OSError %s' % (pid, e))
+    raise e
+  return True
+
+def wait_termination(pid):
+  """Send SIGINT to pid and wait termination of pid.
+
+  Args:
+    pid(int): pid of process which this function waits termination.
+
+  Raises:
+    OSError: is_running_posix, os.waitpid and os.kill may throw OSError.
+  """
+
+  try:
+    os.kill(pid, signal.SIGINT)
+  except:
+    os.kill(pid, signal.SIGKILL)
+    raise
+
+  if os.name == 'nt':
+    try:
+      os.waitpid(pid, 0)
+    except OSError as e:
+      if e.errno == errno.ECHILD:
+        print('ignore errno.ECHILD %s' % e)
+        return
+      raise e
+  else:
+    for _ in xrange(10):
+      if not is_running_posix(pid):
+        break
+      time.sleep(1)
+
+    if is_running_posix(pid):
+      os.kill(pid, signal.SIGKILL)
+      print('killed process %d running more than 10 seconds' % pid)
+
 def main():

[shinyak, 2016/10/11 06:47:02]

Two blank lines between top-level definitions

[tikuta, 2016/10/11 08:09:00]

Done.

   parser = argparse.ArgumentParser(
       description='cloudtail utility for goma recipe module.')
@@ -52,7 +111,7 @@ def main():
     with open(args.killed_pid_file) as f:
       # cloudtail flushes log and terminates
       # within 5 seconds when it recieves SIGINT.

[shinyak, 2016/10/11 06:47:02]

5 seconds. Right?

[tikuta, 2016/10/11 08:09:00]

From
https://codereview.chromium.org/2375843005/#msg28

-      os.kill(int(f.read()), signal.SIGINT)
+      wait_termination(int(f.read()))
 
 if '__main__' == __name__:
   sys.exit(main())