My answers are in-line. On Tue, Mar 26, 2013 at 12:25 AM, Rudolf Streif <rstr...@linuxfoundation.org > wrote:
> I would appreciate if someone could explain to me why for > BaseEC2NodeDriver.list_locations() the name parameter is set to > availability_zone rather than availability_zone.name? > > def list_locations(self): > locations = [] > for index, availability_zone in \ > enumerate(self.ex_list_availability_zones()): > locations.append(EC2NodeLocation( > index, availability_zone, self.country, self, > availability_zone) > ) > return locations > I'm not sure about this one, it looks like a bug to me. I've pushed a commit with a fix to 0.12.x branch - http://svn.apache.org/viewvc?view=revision&revision=r1461400 Thanks for spotting and reporting this. > And the to make the confusion complete, at least for me, in __repr__ the > code outputs self.availability_zone.name for the availability_zone where I > would expect self.availability_zone. > > class EC2NodeLocation(NodeLocation): > def __init__(self, id, name, country, driver, availability_zone): > super(EC2NodeLocation, self).__init__(id, name, country, driver) > self.availability_zone = availability_zone > > def __repr__(self): > return (('<EC2NodeLocation: id=%s, name=%s, country=%s, ' > 'availability_zone=%s driver=%s>') > % (self.id, self.name, self.country, > self.availability_zone.name, self.driver.name)) > Also a bug which should now be fixed in 0.12.x branch. > It does not seem to make much sense to me how it is and it is actually > rather cumbersome to use. In my application I am creating a node using > default location since I cannot specify the location for create_node > (unless I am missing something). > > After that I want to create a volume for which I must specify the location > otherwise AWS EC2 will reject it (the code of BaseEC2NodeDriver in > incorrect in allowing location=None). I will look into this asap. I thought EC2 uses default availability zone if you don't explicitly specify one, but it looks like this might not be the case. > For that purpose I need to query my > node for its location with mynode.extra['availability']. However, I cannot > use that value for create_volume() because it expects a Location object. > > Now I have either manually create a Location object with a nested > availability_zone object for which I have to set the name attribute. > Alternatively, I have to use list_locations() and then iterate through the > list to find the one with the name I am looking for but not using the name > attribute but using availability_zone.name. > On a side note - we are slowly moving away from this (cumbersome) multiple classes per region + list_locations model in trunk. This should hopefully make the whole interface feel more natural and user friendly. > Thanks, > Rudi >
