[dockerctl] Race condition in check_ready function for container 'nailgun'

When container 'nailgun' started with dockerctl, it incorrectly detects that the container is ready, i.e. Puppet successfully applied inside the container.

Steps to reproduce:
1. Install Fuel Admin node using ISO 7.0
2. Install fuel-octane from the source code:

    yum install -y git python-pip python-paramiko
    git clone https://github.com/openstack/fuel-octane
    cd fuel-octane && git checkout stable/7.0 && pip install --no-deps -e .

3. Create backup of Fuel Admin node configuration

    octane fuel-backup --to /tmp/backup.tar.gz

4. Copy backup.tar.gz file to external location.
5. Install Fuel Admin node using ISO 8.0
6. Install fuel-octane from the source code:

    yum install -y git python-pip python-paramiko
    git clone https://github.com/openstack/fuel-octane
    cd fuel-octane && git checkout stable/8.0 && pip install --no-deps -e .

7. Copy backup file to 8.0 Admin node's /tmp dir
8. Restore backup:

    octane fuel-restore -v --debug --from /tmp/backup.tar.gz -p <admin_pass>

Expected result:

Restore is successfully finished

Actual result:

Restore failed at openstack fixture upload phase due to ProgrammaticError from SQL server:

  2016-02-15 12:37:16 ERROR octane.util.subprocess fuel[22974] stderr: 500 Server Error: Internal Server Error ((psycopg2.ProgrammingError) column releases.components_metadata does not exist
  2016-02-15 12:37:16 ERROR octane.util.subprocess fuel[22974] stderr: LINE 1: ..._metadata AS releases_vmware_attributes_metadata, releases.c...
  2016-02-15 12:37:16 ERROR octane.util.subprocess fuel[22974] stderr: ^
  2016-02-15 12:37:16 ERROR octane.util.subprocess fuel[22974] stderr: [SQL: 'SELECT releases.id AS releases_id, releases.name AS releases_name, releases.version AS releases_version, releases.can_update_from_versions AS releases_can_update_from_versions, releases.description AS releases_description, releases.operating_system AS releases_operating_system, releases.state AS releases_state, releases.networks_metadata AS releases_networks_metadata, releases.attributes_me$adata AS releases_attributes_metadata, releases.volumes_metadata AS releases_volumes_metadata, releases.modes_metadata AS releases_modes_metadata, releases.r$les_metadata AS releases_roles_metadata, releases.network_roles_metadata AS releases_network_roles_metadata, releases.wizard_metadata AS releases_wizard_meta$ata, releases.deployment_tasks AS releases_deployment_tasks, releases.vmware_attributes_metadata AS releases_vmware_attributes_metadata, releases.components_$etadata AS releases_components_metadata, releases.modes AS releases_modes, releases.extensions AS releases_extensions \nFROM releases'])```

Root cause:

Command 'dockerctl start nailgun' starts the 'nailgun' container and immediately reports it's readiness. However, at the moment of check, puppet is not yet started in the container, not finished it's work.

Thus, the database sync is still not done when the request is sent, and SQL server replies with ProgrammaticError, since the column in question is not yet created in the DB schema.

Regression was introduced with the following change that fixed bug in systemd support in dockerctl: https://github.com/openstack/fuel-library/commit/c3fc592bbf235985e17eca20b92ae0c1185aa7c8#diff-1ce68b99aae786dd6a4fb77c5c52dd67R139

Workaround:

Insert delay in dockerctl when docker-nailgun.service unit checked.

Solution:

Add check for the nailgun API response in dockerctl functions.