in some scenarios with mongo not running stop_hook will fail and preclude progress

[Samuele Pedroni]

when attaching a volume for example

[Nick Moffitt (nick-moffitt)]

I provisioned units, knowing that they would not be useful until I could get the volumes up and in the volume-map setting. Unfortunately once I did get these new units sorted out for persistent volumes, mongodb was down and all hooks complained about the absence of /var/lib/mongodb/mongod.lock

In the end I had to work around this by manually starting up badly-configured mongodb services on all units and doing a resolved --retry. Ultimately it shouldn't care if mongodb was down already, at this point.

[Xav Paice (xavpaice)]

  @hooks.hook('stop')
  def stop_hook():
      juju_log("stop_hook")
      try:
          retVal = service('stop', 'mongodb')
          os.remove('/var/lib/mongodb/mongod.lock')
          # FIXME Need to check if this is still needed
      except Exception, e:
          juju_log(str(e))
          retVal = False
      finally:
          juju_log("stop_hook returns: %s" % retVal)
          return(retVal)

In the stop hook, although removal of the mongod.lock is caught in a try/except, if the remove fails for any reason (including it not being there to remove), the hook itself will return a fail.