juju gives up on bootstrapping with 'bootstrap instance started but did not change to Deployed state'

Also, I can't destroy the environment after this happens:

jenkins@juju-oil-machine-15:~$ juju destroy-environment --force maas
WARNING! this command will destroy the "maas" environment (type: maas)
This includes all machines, services, data and other resources.

Continue [y/N]? y
ERROR cannot release nodes: gomaasapi: got error back from server: 500 INTERNAL SERVER ERROR ('NoneType' object has no attribute 'has_perm')

Another failure after 4m22s - weird.

jenkins@juju-oil-machine-15:~$ time juju bootstrap -e ci-oil-slave8
Bootstrapping environment "ci-oil-slave8"
Starting new instance for initial state server
Launching instance
WARNING no architecture was specified, acquiring an arbitrary node
WARNING no architecture was specified, acquiring an arbitrary node
 - /MAAS/api/1.0/nodes/node-94afcdc8-aea0-11e3-9074-00163efc5068/
ERROR failed to bootstrap environment: bootstrap instance started but did not change to Deployed state: instance "/MAAS/api/1.0/nodes/node-94afcdc8-aea0-11e3-9074-00163efc5068/" is started but not deployed

real 4m22.135s
user 0m1.576s
sys 0m0.112s

FWIW, I got the same error with juju 1.22-beta1-trusty-amd64 but couldn't reproduce with 1.20.13-trusty-amd64

This error was introduced 2015-01-14 apparently by https://github.com/juju/juju/commit/899f14b4 which introduces the use of env.waitForNodeDeployment(result.Instance.Id()) this will wait, if I got it right, at the most 50 checks of up to 5 seconds which ammounts to a little less than 5 mins.
So if the actual deployment takes up to 7 mins this is not going to be succesful
It would seem that with maas API older than 1.7 it will succeed as per the description:
"""
When starting bootstrap node, wait for node to be reported as Deployed.
The API used is new in MAAS 1.7. When used with older versions of MAAS, a 400 is returned. In these cases, the error is ignored and the node is assumed to be deployed.
"""

I will see how cant it be fixed.

The timeout value used when waiting for the node to transition to deployed is now read from the "bootstrap-timeout" value in the environment configuration. The default for MAAS is 1800s.

We need to port Ian's fix from https://github.com/juju/juju/pull/1503 to 1.23 (trunk) as well - I'm having the same issue when trying to deploy any maas node without using the fast installer. I'll ask James to put it on his list.

This bug is most likely the reason why http://juju-ci.vapour.ws:8080/job/maas-1_8-upgrade-trusty-amd64/65/console and similar CI jobs occasionally fail with 1.21.1 (the fix is not backported even in 1.21.3).

Fix for trunk set to land - https://github.com/juju/juju/pull/1700