Fail gracefully when the db doesn't speak unicode

When sqlalchemy/mysql doesn't contain the charset=utf8 or use_unicode=1
parameters (and the default mysql connection charset is not utf8),
sqlalchemy will connect to mysql in ascii mode; In ascii mode, it will
attempt to reencode all input data to latin-1 regardless of the
database/table collation setting. This catches that error and turns it
into an invalid parameter.

(The error message associated with this is something like
UnicodeEncodeError: 'latin-1' codec can't encode character u'\u2026' in
position 30: ordinal not in range(256))

This fixes bug 944034 and is related to bug 898808

Change-Id: I082b7568ef9e9d2104e13aa58d207535ef201bd3

nova/exception.py

@@ -88,6 +88,8 @@ def wrap_db_error(f):
     def _wrap(*args, **kwargs):
         try:
             return f(*args, **kwargs)
+        except UnicodeEncodeError:
+            raise InvalidUnicodeParameter()
         except Exception, e:
             LOG.exception(_('DB exception wrapped.'))
             raise DBError(e)
@@ -275,6 +277,11 @@ class InvalidRPCConnectionReuse(Invalid):
     message = _("Invalid reuse of an RPC connection.")
 
 
+class InvalidUnicodeParameter(Invalid):
+    message = _("Invalid Parameter: "
+                "Unicode is not supported by the current database.")
+
+
 # Cannot be templated as the error syntax varies.
 # msg needs to be constructed when raised.
 class InvalidParameterValue(Invalid):