I'm able to reproduce this in the end-to-end tests. Filed https://issues.apache.org/jira/browse/AURORA-1409
We're pretty serious about keeping the end-to-end tests green, so this will be ~top priority. -=Bill On Wed, Jul 22, 2015 at 1:18 PM, meghdoot bhattacharya < [email protected]> wrote: > Kevin what is the fix for this?Not sure what is causing this but we are > noticing this as well and the odd thing is some of us has 192.X.X.X > registered and but we are finding cases where 127.0.1.1 is there as well. > Thx > From: Kevin Sweeney <[email protected]> > To: [email protected] > Sent: Monday, July 6, 2015 6:23 PM > Subject: Re: http auth and aurora tutorial > > I just reproduced this as well - the issue appears to be that the scheduler > is registering in ZooKeeper with an IP address of 127.0.1.1, but the netrc > file is configured for 192.168.33.7. > > vagrant@aurora:~$ PEX_INTERPRETER=1 aurora > Python 2.7.6 (default, Jun 22 2015, 17:58:13) > [GCC 4.8.2] on linux2 > Type "help", "copyright", "credits" or "license" for more information. > (InteractiveConsole) > >>> from kazoo.client import KazooClient > >>> from twitter.common.zookeeper.serverset import get_serverset_hosts, > ServiceInstance > >>> client = KazooClient() > >>> client.start() > >>> hosts = get_serverset_hosts('/aurora/scheduler', client) > >>> [ServiceInstance.to_dict(host) for host in hosts] > [{'status': 'ALIVE', 'additionalEndpoints': {u'http': {'host': > u'*127.0.1.1*', > 'port': 8081}}, 'serviceEndpoint': {'host': u'*127.0.1.1*', 'port': 8081}}] > > > > > > > On Mon, Jul 6, 2015 at 10:43 AM, Bill Farner <[email protected]> wrote: > > > I got as far as creating a job following the tutorial [1] and didn't > > encounter any issues. > > > > The client is configured for basic auth, just in a non-obvious way with a > > .netrc file [2]. This is certainly confusing when the clusters.json > > plainly suggests that there is no auth happening [3]. This is the result > > of our move to push authentication down from the application layer, which > > is what the client-side clusters.json entry is for. > > > > Hopefully this clear things up, and if you're encountering any errors you > > think might be related please report back! > > > > [1] > > > > > https://github.com/apache/aurora/blob/14e7b84f4303968029c3803e9b096908f3499d57/docs/tutorial.md#creating-the-job > > [2] > > > > > https://github.com/apache/aurora/blob/be75c36f6ce0b2ca6cc7a910b64e2bb9ca40953d/examples/vagrant/provision-dev-cluster.sh#L92-L99 > > [3] > > > > > https://github.com/apache/aurora/blob/056c35b169874023760edca4fcc222c1ad21d8b1/examples/vagrant/clusters.json#L5 > > > > -=Bill > > > > On Sun, Jul 5, 2015 at 6:01 PM, Mauricio Garavaglia < > > [email protected]> wrote: > > > > > Hi, > > > the vagrant setup seems to enable http_authentication_mechanism=BASIC > in > > > /examples/vagrant/upstart/aurora-scheduler.conf but, AFAIK, the aurora > > CLI > > > client doesn't support authentication yet. This affects the tutorial > > since > > > this is not considered and is not pretty obvious why it doesn't work. > > Would > > > it make sense to get rid of the authentication in the upstart script? > > > > > > > > >
