1

We've been trying to install Openstack via MAAS / JuJu / Landscape. We get to 90% complete Installing region1 in cloud1 but get a node that sticks on Add juju machine for Node3 as in progress with the rest for that node waiting.

MAAS shows all the other nodes deployed except this one (shows failed deployment). What logs and where are they located can we look at to see why this is failing? The node that's failing is the same hardware as three others that are deployed. All help appreciated.

MAAS Version 1.8.0+bzr4001-0ubuntu2 (trusty1) Any other data that's helpful tell me what to send.

  • From your controller, you could try "juju debug-log" and see if there are any errors associated with that node. In the MAAS console you can try to identify the node in question and look to see if it's been issued a static IP address. You can then try "ssh Ubuntu@nodename.domain" from your controller and attempt to log in remotely. If successful you can review what's in /var/log/maas to see what's there, for example, the logs associated with "cloud-init". – Michael Blankenship Sep 22 '15 at 21:04
  • from the maas controller - juju debug-log gives me an "ERROR environment is not bootstrapped". When I try to ssh to the node in question by IP I get "Read from socket failed: Connection reset by peer". It responds to console requests but have no way to login from console (no password). – dwoods Sep 23 '15 at 13:46
  • From maas - I powered it off. Then had to start it manually. Both "power on" and "deploy" give me an error "1 node cannot be deployed, To proceed, update your selection". When powered on manually, it PXE boots, runs whatever script and then shuts itself off. – dwoods Sep 23 '15 at 13:48
  • Unfortunately the bootstrap didn't work so attempting to run any of the other "juju ..." commands won't know about the environment. At the end of a commissioning step it pushes the necessary keypair information so that an ssh attempt can be successful. Until then, the ssh will fail due to missing keys. – Michael Blankenship Sep 23 '15 at 20:37
  • Agree. We did an openstack-install -u and then started over from that point. We also erased the drive of the node in question and did a deploy and release on it. Then performed another openstack-install and then did an install from Landcape. That got us to 98% but Generate Juju Metadata has now failed. Did a download logs, but theres a ton to go through and just haven't sifted through everything to see why its failing now. Our guess is that we have to abort again and take it back to the openstack-install point and then have another run at it. – dwoods Sep 25 '15 at 13:26
0

If the MAAS node shows "failed deployment" then Landscape should have aborted the cloud (I assume there are not enough nodes for there to be spares).

There are logs in /var/log/maas/ that are worth watching; do a couple of commissioning and deployment runs on that node to see if it behaves itself. Also, check if the node will retry PXE or if it just goes to local disk, for example.

  • Lanscape didn't abort. I have the option to abort, but it didn't. The Node was hung at "In progress Add Juju machine for 'Node3-SM-1.maas'" The node was responding to console access but as stated above couldn't log in due to no password and couldn't ssh in. Almost as if the node and the controller were both waiting for the other. I'm going to abort the installation so that I can commission and deploy that node to see what happens. – dwoods Sep 23 '15 at 13:52

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.