conjure-up isn't properly handling a failed bootstrap

[battlemidget]

In some cases where a bootstrap doesn't complete fully some of the configs for juju may be in a inconsistent state which in turn enables conjure-up to fail when attempting to login to the controller.

[mikemccracken]

How does the bootstrap fail without causing a Conjure-up error that stops everything from continuing?

…

-Mike

On Mon, Jan 30, 2017 at 6:24 PM Adam Stokes ***@***.***> wrote: In some cases where a bootstrap doesn't complete fully some of the configs for juju may be in a inconsistent state which in turn enables conjure-up to fail when attempting to login to the controller. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#641>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABr6qnVJB1t9Fc_Ap2JCHQk9Dd1Fih_Qks5rXpttgaJpZM4LyKxU> .

[battlemidget]

This is the current error seen if the accounts.yaml file is either missing or incomplete: http://pastebin.com/2HwvdqPn

I think the controllers.yaml file exists but a matching account for that controller does not. Still waiting on the contents of this particular users's juju files.

[mikemccracken]

ok, yeah that's a confusing error, it'd probably be helpful to see the stdout and stderr from the bootstrap. That should be in ~/.cache/conjure-up/$SPELLNAME/$MODELNAME-bootstrap.out and .err

…

On Mon, Jan 30, 2017 at 7:50 PM, Adam Stokes ***@***.***> wrote: This is the current error seen if the accounts.yaml file is either missing or incomplete: http://pastebin.com/2HwvdqPn I think the controllers.yaml file exists but a matching account for that controller does not. Still waiting on the contents of this particular users's juju files. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#641 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABr6qvEc7WM41E3F7_boQd5cLXbS8_Ogks5rXq98gaJpZM4LyKxU> .

[mc0e]

I'm about to try deploying kubernetes core to localhost a third time. In each of the attempts so far the process has been incomplete, and it's at least unclear how to restart it.

In the first attempt, one of the etcd processes and one of the workers had still not come up after 3 hours, and I terminated conjure-up (ctrl-c). It was far from clear how I might try to recover the situation, so in the end I used lxc stop <id> and lxc delete <id> to remove everything, and then since conjure-up wouldn't run, I deleted the .local/share/juju directory. No-doubt there's a better way somehow, but the "It's all so easy" sales pitch is already looking dubious.

I tried again. It takes a long time to run, so I left it running while I went off to bed. Next day I saw that it had gotten to a point where it appeared about to install and configure kubectl, but that hadn't happened (waiting for user input?) and the network link had had a glitch sometime over night.

What I'd like to see is conjure-up having the ability on a subsequent invocation to recognise that it has an incomplete build operation, and ask if it should pick up where it left off. Failing that it should at least assist with tearing down the failed installation.

After my first attempt, I thought that maybe conjure-up would find the remains of the previous run, but that just led to me having many lxd containers running by the time I decided to just wipe them out and start over.

[battlemidget]

@mc0e while we address your other points in the code can you give us some additional information on your environment? Are you running conjure-up inside a VM? If so, what hardware specs? If not, what is the hardware specs on the system you're running conjure-up? If it's a laptop does it meet the recommended requirements from http://conjure-up.io/docs/en/users/#hardware-requirements?

[battlemidget]

Also for this particular issue this is the error that isn't being surfaced properly:

ERROR enabling HTTPS listener: cannot listen on https socket: listen tcp [::]:8443: socket: address family not supported by protocol

So it's giving us a false error thinking it has something to do with missing juju files.

[battlemidget]

https://bugs.launchpad.net/juju/+bug/1633362

[mc0e]

I'm running in VM with 8GB RAM and 200GB SSD. Less than those hardware requirements you point me to. That might make sense of the varied points at which the process fails.

Would I expect 8GB of RAM to prevent me even installing, or would it just mean there's not all that much room to do things in it once it's up? At this point I'm primarily interested in exploring how it works.

[battlemidget]

@mikemccracken this is the bug I was wanting to address with properly handling a failed bootstrap and providing the user with the errorview and a correct traceback.

[battlemidget]

This is handled with 2403933