Chapter 5. RHBA-2015:1463 - OpenShift Enterprise 2.2.6 Bug Fix and Enhancement Update

Cartridge

BZ#1212583

The Passenger rubygem previously prespawned on the local 127.0.0.1 IP address. As a result, an error was reported at startup in the Ruby log. This bug fix updates Passenger to connect in prespawn to the proper gear IP address, and as a result prespawn functions properly and no error is reported in the logs at startup.

Copy to Clipboard Toggle word wrap

Logging

BZ#1222953

If a blank line was inserted between options in the logshifter configuration file, any options after the first blank line were not parsed. This bug fix updates logshifter to allow for blank lines when reading configuration, and as a result configuration is parsed when blank lines are present.

Copy to Clipboard Toggle word wrap

Broker

BZ#1218049

Previously, when a node containing a head gear for a scaled application was lost and the gear could not be recovered, running the oo-admin-repair tool still attempted to recover the gear. The tool then reported an error and the gear was not recovered, but changes were made to the broker's database. As a result, running the tool again reported all gears had been recovered, even though the head gear still did not exist. This bug fix updates oo-admin-repair to distinguish applications with lost head gears from regular lost gears from scaled applications. The tool now offers to delete any applications with lost head gears, informing the administrator to first re-create such applications from source or recent backups and move any existing alias before proceeding.

Copy to Clipboard Toggle word wrap

BZ#1217588

When moving multiple gears with the oo-admin-move command, the command would previously fail if a single gear could not be found by the provided UUID. Additionally, failures and source server identities were not reported properly in JSON when the --json option was specified. This bug fix updates the command to report an error individually for each gear unable to be found. The rest of the valid gears are still moved properly. JSON output also now includes source server identities, and failures while the --json option is specified are reported in JSON.

Copy to Clipboard Toggle word wrap

BZ#1221786

When implementing a custom billing provider, the usage_account_id field must be set for tracked users. Previously, there was no tool to set the usage_account_id field, and the database had to be modified manually to set this field. With this enhancement, administrators can now use the oo-admin-ctl-user command with the --setusageaccountid option to set the usage_account_id field for users without direct modifications to the database.

Copy to Clipboard Toggle word wrap

BZ#1217587

Previously, all oo-admin-ctl-district commands required that the district name or UUID was specified. This made performing actions on multiple node hosts using oo-admin-ctl-district impossible if they were not in the same district. Additionally, finding the district a node host belonged to was difficult, requiring the administrator to sift through all district node hosts. This bug fix updates oo-admin-ctl-district to allow the district not to be specified for numerous actions; the district can still be specified manually with --uuid or --name options. As a result, the oo-admin-ctl-district command now searches for the correct district if no district is specified, and administrators can additionally perform actions on numerous node hosts across multiple districts.

Copy to Clipboard Toggle word wrap

BZ#1207300

In some cases, it is necessary to use cartridge IDs when activating or deactivating cartridges. As there was no easy way to obtain the cartridge IDs without directly accessing the broker's database, this enhancement updates the 'oo-admin-ctl-cartridge -c list' command to enable cartridge IDs to be displayed by using the --all option.

Copy to Clipboard Toggle word wrap

BZ#1191238

The oo-admin-ctl-app command did not allow specifying a namespace when performing actions on applications. If two applications owned by the same user under two different domains had the same name, it was impossible to perform actions on one of the applications using oo-admin-ctl-app. This bug fix updates oo-admin-ctl-app to allow the namespace of an application to be specified for actions. As a result, any application can be manipulated using oo-admin-ctl-app, even if another application exists under a different domain for the same user.

Copy to Clipboard Toggle word wrap

BZ#1212020

In the /etc/openshift/routing-daemon.conf file, if the MONITOR_TYPE parameter for the F5 routing daemon listener is set to 'https', an 'https-ecv' type monitor was created. When the monitor was to be deleted, a type of 'https' was expected when the type was actually 'https-ecv', returning a 400 error. This bug fix updates the routing daemon such that when 'https' is specified as the MONITOR_TYPE, it expects the monitor to be deleted to have a type of 'https-ecv'. As a result, monitors created with a configured MONITOR_TYPE of 'https' or any other type can be successfully deleted.

Copy to Clipboard Toggle word wrap

BZ#1145344

Previously, administrators could not manage domain membership and could only instruct users to perform domain membership actions themselves using the OpenShift client tools. This enhancement updates the oo-admin-ctl-domain command so that administrators can now use the add_member, remove_member, update_member, and list_members subcommands to manage domain membership for users directly.

Copy to Clipboard Toggle word wrap

BZ#1094454

When attempting to use an alternative cache store for the broker or console, the Gemfile must be modified. The Gemfile is not marked as a configuration file in the respective RPMs, and any changes can be overwritten on updates. This enhancement enables additional gems to be loaded into the broker or console Rails environments by utilizing the ADDITIONAL_GEMS parameter in the /etc/openshift/broker.conf or /etc/openshift/console.conf file, respectively. The broker or console service must be restarted for the change to take effect. As a result, additional gems can be loaded into the broker or console environments without the need to modify Gemfiles and without losing this configuration change upon update or upgrade.

Copy to Clipboard Toggle word wrap

BZ#1210489

A feature was previously added, but never fully implemented, to allow an application to be made highly-available (HA) at creation. When an application is made HA and MANAGE_HA_DNS=true in the /etc/openshift/broker.conf file, the HA routing DNS is created twice. Then, it is created again whenever another HAProxy instance is added. This works without issue when the nsupdate DNS plug-in is used because nsupdate has no problem updating an existing record to the same existing record. However, when other DNS plug-ins, such as a plug-in that makes HTTP requests to manage DNS, attempted to add the same entry twice, an error could be returned. Since the feature to allow an application to be made HA on creation was never fully implemented, the addition of a routing DNS during gear creation has been removed and disabled. As a result, the routing DNS is only registered once, when the application is made HA through the appropriate API call workflow.

Copy to Clipboard Toggle word wrap

Node

BZ#1216223

Administrators who wish to maintain a particular version of JBoss EAP can subscribe to version-specific channels (e.g., 6.4) using RHN Classic or Red Hat Subscription Management (RHSM). OpenShift Enterprise must support this configuration, and part of that includes validating that the correct channel subscription and priorities are set. This is automated by running the oo-admin-yum-validator tool, thus oo-admin-yum-validator must be updated to handle all available JBoss EAP channels typical for OpenShift Enterprise installations. This enhancement updates oo-admin-yum-validator, which can now automatically detect appropriate roles for systems that have particular JBoss EAP channel subscriptions enabled, and can automatically configure JBoss EAP 6.4 channel subscriptions when the --role=node-eap-6.4 argument is specified on the command line.

Copy to Clipboard Toggle word wrap

BZ#1216220

The oo-snapshot command, which is a wrapper for '/usr/bin/gear snapshot', reports an unhelpful error when used outside of a gear environment. This bug fix updates /usr/bin/gear to report an error stating it should only be run inside of a gear when necessary environment variables are not defined. As a result, oo-snapshot reports the same error when run outside of a gear.

Copy to Clipboard Toggle word wrap

BZ#1202705

When a message came through rsyslog from a gear with a new UID, the mmopenshift plug-in set the metadata keys and values in cache for each environment variable specified in the template. If an environment variable was not present at this time, the key would not be created and stored in the metadata cache. This bug fix updates the mmopenshift plug-in so that on every message, if a key expected to be present in the metadata cache is not present, the plug-in checks for the environment variable. If the environment variable exists, the plug-in sets the metadata key to the value of the environment variable. Otherwise, it checks again on the next message. As a result, new gears from a scaled application may report a NULL namespace a few times until the OPENSHIFT_NAMESPACE environment variable is defined. Additionally, any other environment variables in the template will have the same checks. Note that all environment variables specified in the template are expected to be present for all applications. If an environment variable is specified that does not exist for an application, a slight performance degradation may be seen as the environment variable will be checked for on every message.

Copy to Clipboard Toggle word wrap

BZ#1136425

An organization's standardized deployments may enforce password aging for users by default. This made it possible for users to be unable to SSH to their gears after the password expired for the gear users. This bug fix updates nodes to allow administrators to set the DISABLE_PASSWORD_AGING parameter in the /etc/openshift/node.conf file on each node to disable password aging when creating gear users.

Copy to Clipboard Toggle word wrap

BZ#1214832

Previously when a bonded NIC was configured as the external interface for a node host, the network service was not started when the openshift-tc service was started. This was due to bonded NICs depending on the network service to be started, which would not start until after the openshift-tc service. This bug fix modifies the priority of the openshift-tc service on nodes to start after the network service by default. As a result, the openshift-tc service is able to start successfully on boot when a bonded interface is configured as the external interface; the network service is started first, initializing the bonded interface.

Copy to Clipboard Toggle word wrap

Upgrade

BZ#1172055

The openshift-origin-cartridge-php upgrade script previously did not properly upgrade existing cartridges with a version greater than 0.0.17; the Composer directory was not created, which broke Composer functionality. This bug fix updates the upgrade script to perform Composer migration steps on any cartridge with a version less than that used in OpenShift Enterprise release 2.2.6 (i.e., version 0.0.26.1). As a result, PHP cartridges existing before Composer features were added are able to use Composer after an upgrade.

Copy to Clipboard Toggle word wrap