juju bootstrap manual/127.0.0.1 on a unit of a different model causes that unit to lose agent
Bug #1750386 reported by
Ryan Beisner
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Canonical Juju |
Triaged
|
Low
|
Unassigned |
Bug Description
Running `juju bootstrap manual/127.0.0.1` on a unit of a different model causes that unit to lose agent, apparently irrecoverably.
When a user tries to bootstrap manual locally, juju does give an error ("ERROR failed to bootstrap model: machine is already provisioned"), but at that point the damage is already done. Juju unit agents are clobbered in some way, and the agents go lost forever.
If a user then tries again to bootstrap, it is successful, presumably because the juju unit agent was already killed by the first attempt.
The end result appears to be that the user can turn the unit into a controller, and that unit will never take place in its original model again.
When done on the openstack provider:
On the top level model:
$ juju deploy ubuntu
On ubuntu/0:
$ juju bootstrap manual/127.0.0.1 test
Creating Juju controller "test" on manual
Looking for packaged Juju agent version 2.3.3 for amd64
ERROR failed to bootstrap model: machine is already provisioned
Back at the top level: serverstack 2.2.6.3 unsupported
Model Controller Cloud/Region Version SLA
icey icey-serverstack serverstack/
App Version Status Scale Charm Store Rev OS Notes
ubuntu 16.04 active 0/1 ubuntu jujucharms 12 ubuntu
Unit Workload Agent Machine Public address Ports Message
ubuntu/0 unknown lost 1 10.5.0.42 agent lost, see 'juju show-status-log ubuntu/0'
Machine State DNS Inst id Series AZ Message 3860-4aad- b685-2e4085d440 24 xenial nova ACTIVE
1 down 10.5.0.42 83e48847-
Relation provider Requirer Interface Type Message