juju fails to bootstrap local provider with vivid and upstart

Bug #1450092 reported by Stuart Bishop
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
juju-core
Won't Fix
High
Unassigned
1.23
Won't Fix
High
Unassigned
1.24
Won't Fix
High
Unassigned

Bug Description

Upstart is still available as an option for people experiencing bugs with systemd under Vivid. Unfortunately, juju 1.23 can not bootstrap the local provider under Vivid if the user has booted with upstart as the init system.

stub@aargh:~/charms/cassandra/spike$ juju bootstrap
Bootstrapping environment "local"
Starting new instance for initial state server
Building tools to upload (1.23.2.1-vivid-amd64)
Installing Juju agent on bootstrap instance
....
2015-04-29 15:07:47 INFO juju.provider.local environprovider.go:42 opening environment "local"
2015-04-29 15:07:48 INFO juju.cmd supercommand.go:436 command finished
Failed to get D-Bus connection: Operation not permitted
ERROR failed to bootstrap environment: exit status 1

Curtis Hovey (sinzui)
Changed in juju-core:
status: New → Triaged
importance: Undecided → High
tags: added: upstart vivid
Curtis Hovey (sinzui)
Changed in juju-core:
milestone: none → 1.25.0
Revision history for this message
Aaron Bentley (abentley) wrote :

I can reproduce this.

Curtis Hovey (sinzui)
summary: - juju 1.23 fails to bootstrap with upstart
+ juju 1.23 fails to bootstrap with vivid and upstart
Curtis Hovey (sinzui)
description: updated
tags: added: systemd
Changed in juju-core:
status: Triaged → In Progress
assignee: nobody → Eric Snow (ericsnowcurrently)
Revision history for this message
Eric Snow (ericsnowcurrently) wrote : Re: juju 1.23 fails to bootstrap with vivid and upstart

The current behavior is expected, even if not helpful. For cloudinit (see environs/cloudinit/cloudinit.go) we decide the init system based on the juju version of the to-be-provisioned instance. In the case of vivid, that translates to systemd.

When bootstrapping, cloudinit is used to set up the local host as the state instance. In this case, combined with vivid (or later) running upstart, the selection of systemd is incorrect. Early on we made the decision to handle this case by providing a feature flag: "legacy-upstart". So using that flag should allow bootstrap to succeed.

However, the feature flag will cause all vivid (and later) containers to also try to use upstart, which will thus cause problems when deploying services. This is less of an issue since there aren't any vivid charms in the charm store. However, it could be a problem for personal namespace charms or local charms that support vivid.

I'll look at finding a better solution to the cloudinit situation.

Revision history for this message
Eric Snow (ericsnowcurrently) wrote :
Changed in juju-core:
status: In Progress → Fix Committed
Curtis Hovey (sinzui)
Changed in juju-core:
status: Fix Committed → In Progress
Changed in juju-core:
status: In Progress → Fix Committed
Changed in juju-core:
status: Fix Committed → In Progress
Revision history for this message
Eric Snow (ericsnowcurrently) wrote :

The change I landed for this patch broke Juju as did the changes I landed to fix the break:

lp:1457031
lp:1457011
lp:1457218

In the interest of getting CI unblocked I am reverting the changes for now:

1.23: https://github.com/juju/juju/pull/2398
1.24: https://github.com/juju/juju/pull/2396
master: https://github.com/juju/juju/pull/2397

Thus the patch to resolve this bug will need to un-revert those merges as well as land the following:

1.23: https://github.com/juju/juju/pull/2368
1.24: https://github.com/juju/juju/pull/2399
master: https://github.com/juju/juju/pull/2373

Revision history for this message
Eric Snow (ericsnowcurrently) wrote :

This shouldn't be a wide-spread problem. It affects folks bootstrapping local provider on vivid (or later) while using upstart. It will also affect folks deploying charms to instances running vivid/upstart, which is extremely unlikely given that official charms only support LTS and that official cloud images for vivid run systemd.

With that in mind, I've changed the status for 1.23 to "won't fix".

Revision history for this message
Ian Booth (wallyworld) wrote :

Also dropping off the 1.24 milestone - can look at fixing for 1.25

Revision history for this message
Eric Snow (ericsnowcurrently) wrote : Re: juju fails to bootstrap with vivid and upstart

Given priorities, I'm tabling my work on this bug.

summary: - juju 1.23 fails to bootstrap with vivid and upstart
+ juju fails to bootstrap with vivid and upstart
Changed in juju-core:
status: In Progress → Triaged
assignee: Eric Snow (ericsnowcurrently) → nobody
summary: - juju fails to bootstrap with vivid and upstart
+ juju fails to bootstrap local provider with vivid and upstart
Curtis Hovey (sinzui)
Changed in juju-core:
status: Triaged → Won't Fix
milestone: 1.25.0 → none
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Bug attachments

Remote bug watches

Bug watches keep track of this bug in other bug trackers.