[Sqlalchemy-tickets] [sqlalchemy] #2570: Problem while handling server restart with postgresql 8.4.12

[sqlalchemy <mi...@zz...>]

#2570: Problem while handling server restart with postgresql 8.4.12
-----------------------------------------+--------------------
      Reporter:  abdelrahman             |       Type:  defect
        Status:  new                     |   Priority:  medium
     Milestone:                          |  Component:  access
      Severity:  no triage selected yet  |   Keywords:
Progress State:  awaiting triage         |
-----------------------------------------+--------------------
 I have a system where I am upgrading my postgresql from version 8.4.3 to
 8.4.12 while my sqlachamy version is 0.6.4 which i also update to version
 0.7.8.
 I had a problem that the sqlachamy was not able to handle the server
 restart anymore where any requests from my session if the restart fails
 with:
 {{{
 (InterfaceError) connection already closed None None
 }}}

 I had a debugged the problem for a bit and it seems that the exception msg
 raised by the postgresql server when its shutdown is changes from 8.4.3 to
 8.4.12, where in 8.4.3 it was
 {{{

 OperationalError('terminating connection due to administrator
 command\nserver closed the connection unexpectedly\n\tThis probably means
 the server terminated abnormally\n\tbefore or while processing the
 request.\n',)

 }}}

 while the msg in 8.4.12 is
 {{{


 OperationalError('terminating connection due to administrator
 command\nFATAL:  terminating connection due to administrator command\n',)

 }}}

 in the psycopg2.py in both the latest stable version and the development
 version of sqlalchamy, the method check if the connection is lost is:
 {{{
 def is_disconnect(self, e, connection, cursor):
         if isinstance(e, self.dbapi.OperationalError):
             # these error messages from libpq: interfaces/libpq/fe-misc.c.
             # TODO: these are sent through gettext in libpq and we can't
             # check within other locales - consider using
 connection.closed
             return 'closed the connection' in str(e) or \
                     'connection not open' in str(e) or \
                     'could not receive data from server' in str(e)
         elif isinstance(e, self.dbapi.InterfaceError):
             # psycopg2 client errors, psycopg2/conenction.h,
 psycopg2/cursor.h
             return 'connection already closed' in str(e) or \
                     'cursor already closed' in str(e)
         elif isinstance(e, self.dbapi.ProgrammingError):
             # not sure where this path is originally from, it may
             # be obsolete.   It really says "losed", not "closed".
             return "losed the connection unexpectedly" in str(e)
         else:
             return False
 }}}

 Notice that when the OperationalError is raised(which is the one raised on
 server shutdown), the exception message is parsed to match of the expected
 error msgs which does not match in postgresql 8.4.12

-- 
Ticket URL: <http://www.sqlalchemy.org/trac/ticket/2570>
sqlalchemy <http://www.sqlalchemy.org/>
The Database Toolkit for Python