juju 1.17.2 client doesn't like juju 1.17.3 .jenv files

[James Page]

ubuntu@james-page-bastion:~/openstack-bundles⟫ juju-deployer -c ssl-everywhere.yaml -d precise-havana
2014-02-24 13:59:42 [DEBUG] deployer.cli: Using runtime GoEnvironment
2014-02-24 13:59:42 [INFO] deployer.cli: Starting deployment of precise-havana
2014-02-24 13:59:42 [DEBUG] deployer.env: Connected to environment
2014-02-24 13:59:42 [DEBUG] deployer.import: Getting charms...
2014-02-24 13:59:43 [DEBUG] deployer.deploy: Resolving configuration
2014-02-24 13:59:43 [INFO] deployer.import: Deploying services...
2014-02-24 13:59:43 [INFO] deployer.import: Deploying service ceilometer using local:precise/ceilometer
2014-02-24 13:59:43 [ERROR] deployer.env: Error deploying service 'ceilometer'
2014-02-24 13:59:43 [ERROR] deployer.env: Command (juju deploy --config /tmp/tmpuHxZ9p --constraints mem=1G --repository=. local:precise/ceilometer ceilometer) Output:

 ERROR empty image-stream in environment configuration

2014-02-24 13:59:43 [INFO] deployer.cli: Deployment stopped. run time: 1.19

The OpenStack catalog has end points for juju-tools and image data which worked fine with 1.17.2.

ProblemType: Bug
DistroRelease: Ubuntu 14.04
Package: juju-core 1.17.2-0ubuntu4
ProcVersionSignature: User Name 3.13.0-3.18-generic 3.13.0-rc8
Uname: Linux 3.13.0-3-generic x86_64
ApportVersion: 2.13.2-0ubuntu5
Architecture: amd64
Date: Mon Feb 24 14:00:03 2014
Dependencies:
 gcc-4.9-base 4.9-20140110-0ubuntu1
 libc6 2.18-0ubuntu7
 libgcc1 1:4.9-20140110-0ubuntu1
 multiarch-support 2.18-0ubuntu7
Ec2AMI: ami-0000000c
Ec2AMIManifest: FIXME
Ec2AvailabilityZone: serverstack-az-1
Ec2InstanceType: m1.small
Ec2Kernel: aki-00000002
Ec2Ramdisk: ari-00000002
ProcEnviron:
 TERM=screen-256color-bce
 PATH=(custom, no user)
 XDG_RUNTIME_DIR=<set>
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: juju-core
UpgradeStatus: Upgraded to trusty on 2013-12-04 (82 days ago)

[James Page (james-page)]

I note that I'm still using the 1.17.2 client - upgrading that now to see if that helps.

[James Page (james-page)]

OK _ so with a 1.17.3 client, I was able to bootstrap and deploy to 1.17.3 in this environment; so it looks like the issue is 1.17.2 client with a 1.17.3 environment.

[Curtis Hovey (sinzui)]

streams.canonical.com is out of date. I think updating it will reduce this issue to High. Juju CI is broken with the testing of version 1.17.4 because the tools-metadata-url is not honored for aws deployments.

[Dimiter Naydenov (dimitern)]

How is this a regression in juju-core when the issue seems to be the slower updating of streams.canonical.com after the release? This is bound to happen at any time we do a release and update streams.c.c some time after.

[James Page (james-page)]

I don't think this has anything todo with streams.canonical.com; juju-tools and product-streams are provided in the keystone catalog for the openstack deployment where I see this issue so streams.canonical.com should not come into it?

[Andrew Wilkins (axwalk)]

This only affects 1.17.2 and 1.17.3 interaction, so we can just close it.

There was a change introduced in 1.17.2 to add an "image-stream" config attribute, which was previously specific to the Azure provider. The config default (omit) was invalid, though, causing the bug lp:1277636. We fixed this in 1.17.3 by changing the default to match what it was for Azure (empty string). If you try to mix 1.17.2 and 1.17.3, you'll get this complaint. 1.17.3 has set image-stream to "", and the 1.17.2 client doesn't like that.

[John A Meinel (jameinel)]

The specific bug doesn't have to do with streams or any such thing. As Andrew mentioned, it is just that we changed the default values for an item in the environment configuration, and 1.17.2 didn't like the content being the empty string.

The way to "fix" that is to not use 1.17.2. So I think we're ok.

[John A Meinel (jameinel)]

Marked "Won't Fix" because it is technically a bug in 1.17.2's interpretation of a newer config, 1.16 didn't have that field so it isn't a stable release compatibility problem.