aurora git commit: Tear down the observer in case of on unhandled errors

[se...@apache.org]

Repository: aurora
Updated Branches:
  refs/heads/master b099e2f0f -> d56f8c644


Tear down the observer in case of on unhandled errors

I was not able to manually trigger the root cause of AURORA-1801 by altering the Mesos filesystem layout. I have therefore adopted the general teardown idea.

Example output (using a hardcoded throw):

```
Bottle v0.11.6 server starting up (using CherryPyServer())...
Listening on http://192.168.33.7:1338/
Hit Ctrl-C to quit.

E1106 23:03:36.722500 8699 exceptional.py:41] Unhandled error in thread Thread-1 [TID=8705]. Tearing down.
Traceback (most recent call last):
  File "apache/thermos/common/exceptional.py", line 37, in _excepting_run
    self.__real_run(*args, **kw)
  File "apache/thermos/observer/task_observer.py", line 135, in run
    self._detector.refresh()
  File "apache/thermos/observer/detector.py", line 74, in refresh
    self._refresh_detectors()
  File "apache/thermos/observer/detector.py", line 58, in _refresh_detectors
    new_paths = set(self._path_detector.get_paths())
  File "apache/aurora/executor/common/path_detector.py", line 35, in get_paths
    return list(set(path for path in iterate() if os.path.exists(path)))
  File "apache/aurora/executor/common/path_detector.py", line 35, in <genexpr>
    return list(set(path for path in iterate() if os.path.exists(path)))
  File "apache/aurora/executor/common/path_detector.py", line 34, in iterate
    raise RuntimeError("Fail on purpose...")
RuntimeError: Fail on purpose...
I1106 23:03:42.513900 8728 static_assets.py:34] detecting assets...
I1106 23:03:42.541809 8728 static_assets.py:38]   detected asset: observer.js
I1106 23:03:42.542799 8728 static_assets.py:38]   detected asset: bootstrap.css
I1106 23:03:42.543728 8728 static_assets.py:38]   detected asset: jquery.pailer.js
I1106 23:03:42.544576 8728 static_assets.py:38]   detected asset: jquery.js
I1106 23:03:42.548482 8728 static_assets.py:38]   detected asset: favicon.ico
Bottle v0.11.6 server starting up (using CherryPyServer())...
Listening on http://192.168.33.7:1338/
Hit Ctrl-C to quit.
```

Bugs closed: AURORA-1801

Reviewed at https://reviews.apache.org/r/53519/


Project: http://git-wip-us.apache.org/repos/asf/aurora/repo
Commit: http://git-wip-us.apache.org/repos/asf/aurora/commit/d56f8c64
Tree: http://git-wip-us.apache.org/repos/asf/aurora/tree/d56f8c64
Diff: http://git-wip-us.apache.org/repos/asf/aurora/diff/d56f8c64

Branch: refs/heads/master
Commit: d56f8c64466a94a990db3308a3130d3fce0584af
Parents: b099e2f
Author: Stephan Erb <se...@apache.org>
Authored: Thu Nov 24 16:37:51 2016 +0100
Committer: Stephan Erb <se...@apache.org>
Committed: Thu Nov 24 16:37:51 2016 +0100

----------------------------------------------------------------------
 .../apache/aurora/tools/thermos_observer.py     | 24 ++++++++++++++++----
 1 file changed, 20 insertions(+), 4 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/aurora/blob/d56f8c64/src/main/python/apache/aurora/tools/thermos_observer.py
----------------------------------------------------------------------
diff --git a/src/main/python/apache/aurora/tools/thermos_observer.py b/src/main/python/apache/aurora/tools/thermos_observer.py
index 0a07df7..4bba019 100644
--- a/src/main/python/apache/aurora/tools/thermos_observer.py
+++ b/src/main/python/apache/aurora/tools/thermos_observer.py
@@ -14,9 +14,12 @@
 
 """A Mesos-customized entry point to the thermos_observer webserver."""
 
+import sys
+import thread
+import threading
 import time
 
-from twitter.common import app
+from twitter.common import app, log
 from twitter.common.exceptions import ExceptionalThread
 from twitter.common.log.options import LogOptions
 from twitter.common.quantity import Amount, Time
@@ -69,17 +72,30 @@ def initialize(options):
   return TaskObserver(path_detector, interval=polling_interval)
 
 
+def handle_error(exc_type, value, traceback):
+  """ Tear down the observer in case of unhandled errors.
+
+  By using ExceptionalThread throughout the observer we have ensured that sys.excepthook will
+  be called for every unhandled exception, even for those not originating in the main thread.
+  """
+  log.error("An unhandled error occured. Tearing down.", exc_info=(exc_type, value, traceback))
+  # TODO: In Python 3.4 we will be able to use threading.main_thread()
+  if not isinstance(threading.current_thread(), threading._MainThread):
+    thread.interrupt_main()
+
+
 def main(_, options):
   observer = initialize(options)
   observer.start()
   root_server = configure_server(observer)
 
-  thread = ExceptionalThread(target=lambda: root_server.run(options.ip, options.port, 'cherrypy'))
-  thread.daemon = True
-  thread.start()
+  server = ExceptionalThread(target=lambda: root_server.run(options.ip, options.port, 'cherrypy'))
+  server.daemon = True
+  server.start()
 
   sleep_forever()
 
 
+sys.excepthook = handle_error
 LogOptions.set_stderr_log_level('google:INFO')
 app.main()