Chapter 11. RHBA-2014:1353 - OpenShift Enterprise 2.1.7 Bug Fix and Enhancement Update

Broker

BZ#1140378

Previously, domain environment variables were not removed after an application failed during creation or if the operation was canceled by the user. As a result, attempting to later create an application with the same name would fail because MongoDB records still existed for the failed application. This bug fix updates the pending operations logic to remove domain environment variables and SSH keys for an application if the creation fails or is canceled.

BZ#1108762

The broker requires V8, which is provided by Red Hat Software Collections (SCL). While the Management Console and Administration Console plug-ins packaged this requirement, the openshift-origin-broker packages did not. As a result, if an administrator followed the manual installation instructions for a broker host, the broker service would fail to start unless either of the console plug-ins were also installed. This bug fix updates the openshift-origin-broker packages to require V8, and manual installations now work as intended.

BZ#1121243

It was previously possible for the oo-admin-chk and oo-admin repair tools to incorrectly notify administrators of stale domain SSH keys for users that had multiple domains. This was due to a logic error in how the list of keys was gathered for validation. This bug fix updates this logic, and the tools no longer report these false positives.

Cartridge

BZ#1106450

The control script "start" routine for the MySQL cartridge was configured to check the mysqld service 10 times, wait 1 second between each check, and time out if the service did not become available within that time. During periods of high I/O load, and under certain other conditions, the mysqld service might fail to start before all 10 checks have been performed. This caused the deployment or scale-up operation to fail. This bug fix increases the number of checks to 45 for a minimum timeout duration of 45 seconds. It also introduces the OPENSHIFT_MYSQL_START_TIMEOUT and OPENSHIFT_MYSQL_STOP_TIMEOUT environment variables which users can set using the client tools to specify the number of retries for the control "start" and "stop" routines, respectively. The MySQL cartridge is now more tolerant to high latency system conditions by default, and can be manipulated by the user to successfully deploy under a variety of system load and latency scenarios. After applying this fix, a cartridge upgrade is required.

BZ#1116819

Previously, the Jenkins client cartridge did not define a character encoding for builds. As a result, builds reported "InvalidByteSequenceError" failures in some cases, although the build was actually successful. This bug fix updates the Jenkins client cartridge to define a default character set of UTF-8, and the error message no longer appears during builds.

BZ#1112139

Due to an incompatibility between Red Hat Software Collections (SCL) packages, the mysql gem, and the Ruby cartridge, gem installations would fail in applications with Ruby and mysql-5.5 cartridges. This bug fix introduces a workaround to the Ruby cartridge that points bundler to the correct version of libmysqlclient until the MySQL 5.5 SCL is updated. After applying this fix, a cartridge upgrade is required.

BZ#1125430

Because JBoss EAP 6.1 no longer requires modules index files, the jboss-eap6-index package became obsolete. This bug fix removes the jboss-eap6-index package dependency from the JBoss EAP cartridge. Administrators can decide whether or not to remove this package from all node hosts. It is likely that the upgrade process to OpenShift Enterprise 2.2 will perform this step automatically.

Node

BZ#1131139

The oo-mco script is installed on node hosts during initial deployment due to an RPM configuration, however the SCL mco client executable is not installed. This caused attempts to run the oo-mco command on node hosts to fail with an unhelpful error message. This bug fix adds a more helpful error message to the oo-mco script for this scenario, and administrators attempting to run the oo-mco command on a node host are now informed that "/opt/rh/ruby193/root/usr/sbin/mco does not exist. mco client is likely not installed from scl."

BZ#1134034

Because node host administrators may want to have logging consolidated via Syslog while gear users may wish to have logs written to the application root directory, nodes now support log delivery to both Syslog and file output types at the same time. Administrators can set the 'outputType' in the /etc/openshift/logshifter.conf file on node hosts to 'multi', which results in logs being written to both 'file' and 'syslog'.

BZ#1135617

When using the apache-vhost front-end server plug-in, the systemu Ruby library attempted to create temporary files that were blocked by SELinux. This resulted in an AVC denial in the /var/log/audit/audit.log file, though no functionality was actually affected. Because the systemu library is redundant, this bug fix updates nodes to use the OpenShift library that provides the same utility in the rest of the node runtime. As a result, the AVC denials no longer occur.

BZ#1139359

Node platform log entries are built with attributes from the arguments of the MCollective call that caused them. In the case of the execute_parallel action, the arguments are structured differently. With application and gear context enabled for node platform logs, attributes were previously not logged for the execute_parallel actions, which include adding SSH keys and cartridges. This bug fix ensures that the arguments are now gathered from either argument structure, if present. As a result, all known actions with the relevant MCollective call arguments are logged with the appropriate context attributes.

Upgrade

BZ#1102138

The ose-upgrade tool coordinates running a number of scripts to perform an OpenShift Enterprise version upgrade. When running the scripts, the tool previously displayed script output only after completion, even though several steps of the upgrade can take a long time, for example while updating RPMs or waiting on another host to complete its steps. This meant that administrators had no method for verifying what the tool was doing or whether something had gone wrong while the scripts were running. This enhancement updates the tool to display ongoing output from the scripts while they are running, and administrators are now able to better check the status of the scripts being run.