'container failed to start' with local provider
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
juju-core |
Won't Fix
|
Low
|
Unassigned |
Bug Description
juju-core 1.21-beta2-
When using juju local provider on a fresh utopic install, bootstrap succeeds but adding a machine fails. juju status returns:
environment: local
machines:
"0":
agent-state: started
agent-version: 1.21-beta2.1
dns-name: localhost
instance-id: localhost
series: utopic
state-
"1":
agent-
instance-id: pending
series: trusty
services: {}
/var/lib/
lxc-start 1416342599.055 ERROR lxc_conf - conf.c:
lxc-start 1416342599.055 ERROR lxc_conf - conf.c:
(full log attached)
$ ls -l /home/liam/
ls: cannot access /home/liam/
ls: cannot access /usr/lib/
$ ls -l /home/liam/
total 0
no longer affects: | juju-core/1.21 |
tags: | added: deploy doc local-provider |
Changed in juju-core: | |
importance: | High → Medium |
Changed in juju-core: | |
importance: | Medium → Low |
Changed in juju-core: | |
status: | Triaged → Won't Fix |
The issue is that there was a template container plus config generated by Juju 1.20 or earlier. This template container is used (when lxc-clone is true) to make container creattion fast. But the template config is out of date with 1.21, because it contains a bind mount to a tools directory that no longer exists in 1.21.
So, an upgrade to 1.21 should check for existing lxc templates and do something - whether that's warning the user and advising them to delete the template, or attempting to parse and fix the lxc config, I'm not sure.