Knife EC Backup

Build status Gem Version

Umbrella Project: Knife

Project State: Active

Issues Response Time Maximum: 14 days

Pull Request Response Time Maximum: 14 days

Description

knife-ec-backup can backup and restore the data in an Enterprise Chef Server installation, preserving the data in an intermediate, editable text format. It is similar to the knife download and knife upload commands and uses the same underlying libraries, but also includes workarounds for objects not yet supported by those tools and various Server API deficiencies. The long-run goal is to improve knife download, knife upload and the Chef Infra Server API and deprecate this tool.

Requirements

This knife plugin requires Chef Infra Client 11.8+.

Server Support

This plugin currently supports Enterprise Chef 11 and Chef Infra Server 12+. Support for the beta key rotation features is provided via the --with-keys-sql flag, but users of this feature should note that this may change once the Chef Infra Server supports an API-based export of the key data.

Installation

This gem is installed with chef-server-core 12.0.0 and newer.

For Private Chef 11 (or Enterprise Chef 11) you'll need to download and build locally to get the correct dependencies, either with git clone or by downloading the .zip file. Once unpacked, run:

/opt/opscode/embedded/bin/gem build knife-ec-backup.gemspec
/opt/opscode/embedded/bin/gem install knife-ec-backup*gem --no-ri --no-rdoc -V

Note on installing with existing development tools:

The latest versions of knife-ec-backup require gems with native extensions, thus you must install a standard build toolchain. To install knife-ec-backup without installing libpq development headers on your system, try the following:

/opt/opscode/embedded/bin/gem install knife-ec-backup -- --with-pg-config=/opt/opscode/embedded/postgresql/9.2/bin/pg_config

This uses the libpq headers that are included in the Chef Infra Server package installed in /opt/opscode.

Chef Workstation Install (Unsupported)

On systems other than the Chef Infra Server, installation of this gem is not tested or supported. However, if you attempt to do so you will need the postgresql libraries installed.

For example, on macOS:

brew install libpq
gem install knife-ec-backup -- --with-pg-config=/usr/local/Cellar/libpq/9.2/bin/pg_config

The current location of pg_config can be determined with brew info libpq.

Running tests

$ bundle install
$ bundle exec rspec spec/

If bundle install fails on the pg gem and the note above does not work for you, try:

$ brew install postgres (if not present)
$ ARCHFLAGS="-arch x86_64" gem install pg
$ bundle exec rspec spec/

Build from source

Clone the git repository and run the following from inside:

gem build knife-ec-backup.gemspec
gem install knife-ec-backup*gem

Configuration

Permissions

Note that most users in an EC installation lack the permissions to pull all of the data from all organizations and other users. This plugin REQUIRES THE PIVOTAL KEY AND WEBUI KEY from the Chef Infra Server. It is recommended that you run this from a frontend Enterprise Chef Infra Server, you can use --user and --key to pass the pivotal information along.

Subcommands

Common Options

The following options are supported across all subcommands:

  • --sql-host: The hostname of the Chef Infra Server's postgresql server. (default: localhost)

  • --sql-port: The postgresql listening port on the Chef Infra Server. (default: 5432)

  • --sql-db: The postgresql Chef Infra Server database name. (default: opscode_chef) Specify 'automate-cs-oc-erchef' when using Automate Chef Infra Server API

  • --sql-user: The username of postgresql user with access to the opscode_chef database. (default: autoconfigured from /etc/opscode/chef-server-running.json)

  • --sql-password: The password for the sql-user. (default: autoconfigured from /etc/opscode/chef-server-running.json)

  • --purge: Whether to sync deletions from backup source to restore destination. (default: false)

  • --dry-run: Report what actions would be taken without performing any. (default: false)

knife ec backup DEST_DIR (options)

Path: If you have chef-client installed as well, you may need to invoke this as /opt/opscode/embedded/bin/knife ec backup backup

Options

  • --concurrency THREAD_COUNT: The maximum number of concurrent requests to make to the Chef Server. (default: 10)

  • --webui-key: Used to set the path to the WebUI Key (default: /etc/opscode/webui_priv.pem) skip any auto-configured options (default: false)

  • --with-user-sql: Whether to backup/restore user data directly from the database. This requires access to the listening postgresql port on the Chef Server. This is required to correctly handle user passwords and to ensure user-specific association groups are not duplicated.

  • --with-key-sql: Whether to backup/restore key data directly from the database. This requires access to the listening postgresql port on the Chef Infra Server. This is required to correctly handle keys in Chef Infra Servers with multikey support. This option will only work on restore if it was also used during the backup.

  • --skip-useracl: Skip download/restore of the user ACLs. User ACLs are the permissions that actors have on other global users. These are not the ACLs that control what permissions users have on various Chef objects.

  • --skip-version-check: Skip Chef Infra Server version check. This will also skip any auto-configured options (default: false)

  • --only-org ORG: Only donwload/restore objects in the named organization. Global objects such as users will still be downloaded/restored.

Creates a repository of an entire Enterprise Chef / Private Chef server.

The format of the repository is based on the knife-essentials (knife download) format and looks like this:

users
  <name>.json
user_acls
  <name>.json
organizations
  <orgname>
    acls
      <type>
        <name>.json
    clients
      <name>.json
    containers
      <name>.json
    cookbooks
      <name>-<version>
    data_bags
      <bag name>
        <item name>
    environments
      <name>.json
    groups
      <name>.json
    nodes
      <name>.json
    roles
      <name>.json
    org.json
    members.json
    invitations.json

This compares very closely with the "knife download /" from an OSC server:

clients
  <name>.json
cookbooks
  <name>-<version>
data_bags
  <bag name>
    <item name>
environments
  <name>.json
nodes
  <name>.json
roles
  <name>.json
users
  <name>.json>

knife ec restore DEST_DIR (options)

Restores all data from the specified DEST_DIR to an Enterprise Chef / Private Chef server. DEST_DIR should be a backup directory created by knife ec backup

Options

  • --webui-key: Used to set the path to the WebUI Key (default: /etc/opscode/webui_priv.pem)

  • --overwrite-pivotal: Whether to overwrite pivotal's key. Once this is done, future requests will fail until you fix the private key (default: false)

  • --skip-users: Skip the restore of global users. This may cause organization uploading to fail if the necessary users do not exist on the Chef Server.

  • --concurrency THREAD_COUNT: The maximum number of concurrent requests to make to the Chef Server. (default: 10)

  • --skip-version-check: Skip Chef Infra Server version check. This will also skip any auto-configured options (default: false)

  • --[no-]skip-user-ids: Reuses user ids from the restore destination when updating existing users to avoid database conflicts (default: true)

  • --with-user-sql: Whether to backup/restore user data directly from the database. This requires access to the listening postgresql port on the Chef Server. This is required to correctly handle user passwords and to ensure user-specific association groups are not duplicated. This option will only work on restore if it was also used during the backup.

  • --with-key-sql: Whether to backup/restore key data directly from the database. This requires access to the listening postgresql port on the Chef Infra Server. This is required to correctly handle keys in Chef Infra Servers with multikey support. This option will only work on restore if it was also used during the backup.

  • --skip-useracl: Skip download/restore of the user ACLs. User ACLs are the permissions that actors have on other global users. These are not the ACLs that control what permissions users have on various Chef objects.

  • --only-org ORG: Only download/restore objects in the named organization. Global objects such as users will still be downloaded/restored.

knife ec key export [FILENAME]

Create a json representation of the users table from the Chef Infra Server database. If no argument is given, the name of the backup is key_dump.json.

Please note, most users should use knife ec backup with the --with-user-sql option rather than this command.

knife ec key import [FILENAME]

Import a json representation of the users table from FILENAME to the the Chef Infra Server database. If no argument is given, the filename is assumed to be key_dump.json.

Please note, most user should use knife ec restore with the --with-user-sql option rather than this command.

Known Bugs

  • knife-ec-backup cannot be installed in the embedded gemset of Chef Server 12. This will be resolved in a future Chef Infra Server release.

  • knife ec restore can fail to restore cookbooks, failing with an internal server error. A common cause of this problem is a concurrency bug in Chef Infra Server. Setting --concurrency 1 can often work around the issue.

  • knife ec restore can fail if the pool of pre-created organizations can not keep up with the newly created organizations. This can typically be resolved simply be restarting the restore. To avoid this error for backups with large number of organizations, try setting (in /etc/opscode/private-chef.rb):

    opscode_org_creator['ready_org_depth']
    

to the number of organizations in your backup and waiting for the pool to fill before running knife ec restore