You should make mysql start on startup, this will let nova services to start without fail. You can do this using update-rc.d, sudo update-rc.d mysql defaults check this for other services as well, like rabbitmq-server, etc. You can use parameters described in the link for setting number of tries to connect to database, maximum reties, retry interval, etc: http://docs.openstack.org/juno/config...

It seems that this is problem of OpenStack itself - it has no retry on opennning connection pool to db on startup after reboot. You can fix it manually by modifiying /usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py file:

• add the following lines to the beginning of the file

      import logging
      LOG = logging.getLogger(__name__)
      import time
  

• and replace the following

      return self._connection_cls(self,
                              self.pool.connect(),
                              close_with_result=close_with_result,
                              **kwargs)
  

  by

      ret = None
      try:
          ret= self._connection_cls(self,
                              self.pool.connect(),
                              close_with_result=close_with_result,
                              **kwargs)
      except:
          LOG.warn('Retrying to connect to the database, due to unavailable database! Sleeping 10 seconds before retry...')
          time.sleep( 10 )
          ret= self._connection_cls(self,
                              self.pool.connect(),
                              close_with_result=close_with_result,
                              **kwargs)
      return ret
  

Literally this will add 10 seconds delay to initial pool opening in cae of error and then reconnect.

In my case it was used for fixing (OperationalError) (2003, "Can't connect to MySQL server on 'jcloud' (111)") None None in logs of nova-cert and nova-scheduler after reboot.

I also had this problem. This solution http://stackblog.net/?p=14 worked fine for me.