Beaker

What’s New in Beaker 0.13?

The principal feature added in Beaker 0.13 is the initial implementation of elements of the proposal-enhanced-user-groups design proposal.

Migrating to Beaker 0.13

There are no changes in this release that should impact task or job compatibility.

The new Group jobs feature is intended to completely replace the existing behaviour where group members are automatically granted increased access to all jobs submitted by fellow group members. Accordingly, the legacy implicit sharing behaviour is now deprecated and will be removed in the next Beaker release. This phased removal should allow users to switch to using explicitly shared group jobs before the implicit sharing is removed.

Note

This interim backwards compatibility with the old system does create a loophole where it is possible to create a new group, add another user to it and then have additional access to their jobs. The current release will track any such behaviour in the group and job activity logs, until the removal of the old implicit job sharing model in the next release prevents it entirely.

Group jobs

Beaker 0.13 allows job submitters to specify a group owner for a job. This is done via the group attribute on a job XML’s <job> element, or by passing the --job-group option to a client workflow command. All members of the associated group will have the same permissions to view and modify a job as the original submitter.

In addition, the public SSH keys for all group members will be added to any systems provisioned for a group job, and group owners may optionally set a shared root password for the group that will be used on those systems.

The owning group (if any) for each job is given on the job details page, as well as in the global and individual job lists. The field is also made available as a new option when filtering the job lists.

The list in “My Jobs” has been updated to include both jobs submitted by the user and those that have been submitted on behalf of any groups they belong to. The new filtering options can be used to limit the list back to just those jobs the user has submitted directly.

(Contributed by Raymond Mancy, Amit Saha and Dan Callaghan, primarily in #%s908183, #%s908186, #%s952980, #%s961192, #%s961194, and #%s961580.)

More flexible user groups

Historically, creating and populating user groups in Beaker has been limited to the administrators of a Beaker instance. With the introduction of the group jobs feature, it is desirable that users have more direct control over the groups defined in Beaker.

Creating user groups

Beaker users can now create user groups without admin privileges. A new sub-command bkr group-create is added to create a group from the command line. New groups can also be defined through the Groups and My Groups pages in the web UI.

(Contributed by Amit Saha in #%s908172.)

Modifying user groups

Beaker users can modify group details (group name, display name) and change the group membership (add/remove group members) of the groups they own. Users will get an email notification when their membership changes (added/removed).

Existing group owners can also grant and revoke ownership privileges, allowing a group to have multiple co-owners.

A new sub-command bkr group-modify has been added to modify the group details from the command line. Groups details can also be accessed and updated through the Groups and My Groups pages in the web UI.

(Contributed by Amit Saha and Dan Callaghan, primarily in #%s908174, #%s908176, #%s952978, #%s961248 and #%s970493.)

LDAP groups

In addition to allowing all users to create new groups, Beaker 0.13 allows Beaker admins to define new groups that are automatically populated from an external LDAP directory.

These groups are automatically refreshed daily from LDAP, but admins can also force an immediate update by running beaker-refresh-ldap on the main Beaker server.

(Contributed by Dan Callaghan in #%s908173.)

Notable enhancements

Improved documentation for task development

A new section on Writing tasks in the Beaker documentation covers how to create new tasks, details of the metadata that can be defined for tasks, details of the environment variables automatically defined by Beaker when running a task, and the command line tools that are automatically provided for use by tasks during execution.

The standard tasks provided along with Beaker have also been documented.

The task related documentation has also been updated to ensure it is accurate for current versions of Beaker.

(Contributed by Dan Callaghan, Amit Saha and Nick Coghlan in #%s872421, #%s921346, and #%s960317.)

Command line tool for listing group members

A new beaker command, bkr group-members is now available for listing a group’s members. This command returns the username, email for each member and also identifies whether the member is a group owner or not.

(Contributed by Amit Saha.)

Chrony is enabled when appropriate

The chrony clock synchronization daemon is now installed and enabled by default for recipes running on supported versions of Fedora and other related distros where the ntpd daemon is no longer available. In addition, the harness is now configured to wait for clock synchronization before it starts.

If you want to opt out of this behaviour (for example, if the presence of the chrony package interferes with your testing) you can pass ks_meta="no_clock_sync" in your job XML.

(Contributed by Bill Peck and Dan Callaghan in #%s901670.)

Power script for Hyper-V guests

A new power script was added for controlling Hyper-V guests. This allows you to add predefined Hyper-V guests as systems in Beaker, similar to the existing virsh power script.

(Contributed by Ladislav Jozsa in #%s884558.)

optional-debuginfo repos are imported

beaker-distro-import now correctly imports the optional-debuginfo repository (where it is available). To add this repo to already imported distros, the distros will have to be re-imported.

(Contributed by Raymond Mancy in #%s952963.)

System filtering using inventory date and status

A new XML element <last_inventoried> has been added to enable system filtering using the date on which the system was last inventoried. For example, to specify that your job should be run on a system inventoried after 2013-01-02, you should add the following in your job XML:

<hostRequires>
    <system> <last_inventoried op="&gt;" value="2013-01-02" /> </system>
</hostRequires>

Besides the above utility, this enhancement also allows you to use the bkr command line tool to list systems based on their last inventoried status or date using the --xml-filter option.

(Contributed by Amit Saha in #%s949777.)

System search using inventory date and status

It is now possible to search for systems from the Beaker Web UI (See System searching) based on the date when they were last inventoried.

This can also be used to check the inventory status of the systems. Searching for systems using the “is” operator with a blank date value returns all uninventoried systems, where as using the “is not” operator will return all inventoried systems.

(Contributed by Amit Saha in #%s949777.)

Bug fixes and minor enhancements

A number of other smaller fixes and enhancements have been included in this release.

Client tools

  • #%s929202: beaker-wizard package detection is once again supported
  • #%s929190: beaker-wizard is now compatible with more recent versions of the python-bugzilla support library.
  • #%s952486: rhts-test-env no longer depends on beakerlib (when appropriate, Beaker now installs beakerlib explicitly on provisioned systems).

Provisioning and task execution

  • #%s886875: Kernel options of the form key=value1 key=value2 are once more correctly added to a recipe instead of being passed through as key=value2.
  • #%s929116: Non-ASCII characters are now supported in job and recipe whiteboards.
  • #%s903416: The handling of kickstarts for Red Hat Storage has been updated so that the variable RHS specific elements are retrieved automatically from the distro tree.

Alternative harnesses

  • #%s961300: Uploading an empty log file is now permitted.
  • #%s962253: Uploading a log file to a finished task now returns a 409 Conflict response instead of triggering an internal server error.
  • #%s962254: Reporting a result for a finished task now returns a 409 Conflict response instead of triggering an internal server error.
  • A specific 411 Length Required response is now returned for a missing Content-Length header when uploading a log file, rather than a general 400 Bad Request.

Miscellaneous changes

  • #%s970512: Systems can once again be removed from a group directly from the group’s page
  • #%s955868: XML filtering based on a system’s Added data now gives sensible results.
  • #%s963700: The URL scheme, hostname and port used to serve uploaded log files are now configurable.
  • #%s962582: All Beaker components now require version 1.0 or later of the python-requests library.
  • #%s957577: bkr is now a properly stateless Python namespace package.
  • #%s951985: beaker-import now handles failing to import a distro tree with a % character in the error message
  • #%s903902: Eliminated a race condition when marking systems broken.
  • #%s952929: Eliminated a race condition in the application log initialisation.
  • #%s962901: Eliminated a race condition when archiving console logs.
  • #%s880855: Eliminated a race condition when creating per-recipe repositories.
  • #%s950895: Mitigated the negative effects of interference between processing of queued recipes and completion of recipes.

Maintenance updates

The following fixes have been included in Beaker 0.13 maintenance updates.

Beaker 0.13.1

Bug fixes:

  • #%s974382: Distro trees for Red Hat Enterprise Linux 4 can once again be imported into Beaker.

Beaker 0.13.2

Minor features:

  • #%s973092: Setting “grubport=” in the kickstart metadata is now supported
  • #%s973595: btrfs volumes can now be created when installing with recent anaconda versions
  • #%s973893: The Administrator’s Guide now covers how to upgrade an existing Beaker installation to a new maintenance or feature release
  • #%s972417: Beaker workflow commands now default the number of clients and servers to zero, allowing the options to be used independently

Bug fixes:

  • #%s952587: Some steps in the job scheduler have been serialised to eliminate scheduling anomalies seen with the previous approach.
  • #%s974352: XML-RPC retries on lab controllers are now logged correctly
  • #%s974319: Lab controller requests left over from a previous network failure are now purged without aborting new recipes running on affected systems
  • #%s972397: Sorting certain combinations of data grid columns no longer triggers an internal server error
  • #%s972411: Submitting malformed CSV to CSV import no longer triggers an internal server error
  • #%s972412: Submitting invalid UTF-8 characters in job XML no longer triggers an internal server error
  • #%s957011: RHEL 6 Kickstarts are once again generated correctly when provisioning systems in manual mode
  • #%s979999: The link to the Relax NG schema from the docs home page has been fixed.