Add HttpsMonitor

infra_libs/ts_mon/common/pb_to_popo.py

+# Copyright 2015 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+from google.protobuf.descriptor import FieldDescriptor as fd
+
+def convert(pb):
+  """Convert protobuf to plain-old-python-object"""
+  obj = {}
+  for field, value in pb.ListFields():
+    if field.label == fd.LABEL_REPEATED:
+      obj[field.name] = list(_FD_TO_JSON[field.type](v) for v in value)

[Sergey Berezin, 2016/07/01 22:05:21]

nit: just to be on the safe side, I'd use _FD_TO_JSON.get(field.type, unicode).
If a proto changes and gets a new type, at least the monitoring code won't crash
the application.

[tnn, 2016/07/01 22:31:39]

If a proto changes, crashing an application is better for uncovering a problem
(epsecially when we do canarying first). Otherwise it would silently sending
errorneous metrics.

However, I don't have a strong feeling about this. Please let me know what you
think.

[Sergey Berezin, 2016/07/06 17:18:39]

For application code a crash may be preferred in some cases. But this is
monitoring code, and a bug in it shouldn't crash all the applications in the
fleet. I'd rather add excessive fault tolerance, log errors, and possibly
corrupt monitoring data, but at least the app itself should continue to work
normally and serve its users.

[tnn, 2016/07/06 19:28:06]

Fair enough. I'll use unicode and log something here. Thanks!

+    else:
+      obj[field.name] = _FD_TO_JSON[field.type](value)
+  return obj
+
+
+_FD_TO_JSON  = {
+  fd.TYPE_BOOL: bool,
+  fd.TYPE_DOUBLE: float,
+  fd.TYPE_ENUM: int,
+  fd.TYPE_FIXED32: float,
+  fd.TYPE_FIXED64: float,
+  fd.TYPE_FLOAT: float,
+  fd.TYPE_INT32: int,
+  fd.TYPE_INT64: long,
+  fd.TYPE_SFIXED32: float,
+  fd.TYPE_SFIXED64: float,
+  fd.TYPE_SINT32: int,
+  fd.TYPE_SINT64: long,
+  fd.TYPE_STRING: unicode,
+  fd.TYPE_UINT32: int,
+  fd.TYPE_UINT64: long,
+  fd.TYPE_MESSAGE: convert
+}