race: concurrent charm deployments corrupts deployments

Still an issue

lucky(~/src/launchpad.net/juju-core) % juju deploy -n7 wordpress &
[1] 12085
lucky(~/src/launchpad.net/juju-core) % juju deploy -n7 wordpress wp-1 &
[2] 12089
lucky(~/src/launchpad.net/juju-core) % error: cannot add charm: cannot add charm "cs:precise/wordpress-9": E11000 duplicate key error index: juju.charms.$_id_ dup key: { : "cs:precise/wordpress-9" }

This is still an issue at rev 827

lucky(~/src/launchpad.net/juju-core) % juju deploy -n10 mysql a & juju deploy -n10 mysql b &
[1] 3537
[2] 3538
lucky(~/src/launchpad.net/juju-core) % error: cannot add charm: cannot add charm "cs:precise/mysql-14": E11000 duplicate key error index: juju.charms.$_id_ dup key: { : "cs:precise/mysql-14" }

[1]- Exit 1 juju deploy -n10 mysql a

Dimitern, assigning this to you because I suspect you've fixed it. When you get back from holiday, please verify and update as appropriate.

Nope, it's not fixed - I can reproduce it on latest trunk (r2259) using this script:

#!/bin/bash

juju switch local
juju bootstrap --upload-tools
juju deploy -n7 wordpress &
juju deploy -n7 wordpress wp1 &
juju status

I'm looking into it now.

I'm ready to propose the branch with the fix. Running the script above now produces this:

local -> local
[sudo] password for dimitern:
Logging to /home/dimitern/.juju/local/log/cloud-init-output.log on remote host
Fetching tools: wget --no-verbose -O $bin/tools.tar.gz 'http://10.0.3.1:8040/tools/releases/juju-1.17.1.1-saucy-amd64.tgz'
Starting MongoDB server (juju-db-dimitern-local)
Bootstrapping Juju machine agent
Starting Juju machine agent (juju-agent-dimitern-local)
environment: local
machines:
  "0":
    agent-state: pending
    dns-name: localhost
    instance-id: localhost
    series: saucy
services: {}
Added charm "cs:precise/wordpress-21" to the environment.
ERROR charm "cs:precise/wordpress-21" already uploaded

We catch the race before it happens and report it to the user.