OpenShift Container Platform 3.7 Release Notes

Many core features Google announced in June for Kubernetes 1.7 were the result of OpenShift engineering. Red Hat continues to influence the product in the areas of storage, networking, resource management, authentication and authorization, multi-tenancy, security, service deployments, templating, and controller functionality.

This feature is currently in Technology Preview and not for production workloads. CRI-O with builds will not yet work.

CRI-O v1.0 is a lightweight, native Kubernetes container runtime interface. By design, it provides only the runtime capabilities needed by the kubelet. CRI-O is designed to be part of Kubernetes and evolve in lock-step with the platform.

CRI-O brings:

To install and run CRI-O alongside docker, set the following in the [OSEv3:vars] section Ansible inventory file during cluster installation:

This setting pulls the openshift3/cri-o system container image from the Red Hat Registry by default. If you want to use an alternative CRI-O system container image from another registry, you can also override the default using the following variable:

When CRI-O use is enabled, it is installed alongside docker, which currently is required to perform build and push operations to the registry. Over time, temporary docker builds can accumulate on nodes. You can optionally set the following parameter to enable garbage collection. You must configure the node selector so that the garbage collector runs only on nodes where docker is configured for overlay2 storage.

For example, the above would ensure it is only run on nodes with the runtime: cri-o label. This can be helpful if you are running CRI-O only on some nodes, and others are only running docker.

See the upstream documentation for more information on CRI-O.

In a multi-tenant environment, you want to leverage administration controllers to help define rules that can help govern a cluster, should a tenant not set a toleration for placement.

The following is offered to administrators where the namespace setting will override the cluster setting:

While OpenShift Container Platform is a secured by default implementation of Kubernetes, there is now documentation on what security protocols and ciphers are used.

OpenShift Container Platform leverages Transport Layer Security (TLS) cipher suites, JSON Web Algorithms (JWA) crypto algorithms, and offers external libraries such as The Generic Security Service Application Program Interface (GSSAPI) and libgpgme.

Private and public key configurations and Crypto levels are now documented for OpenShift Container Platform.

Pods can no longer try to gain information from secrets, configuration maps, PV, PVC, or API objects from other nodes.

Node authorizer governs what APIs a kubelet can perform. Spanning read-, write-, and auth-related operations. In order for the admission controller to know the identity of the node to enforce the rules, nodes are provisioned with credentials that identify them with the user name system:node:<nodename> and group system:nodes.

These enforcements are in place by default on all new installations of OpenShift Container Platform 3.7. For upgrades from OpenShift Container Platform 3.6, they are not in place due to the system:nodes RBAC being granted from OCP 3.6. To turn the enforcements on, run:

With Advanced Auditing, administrators are now exposed to more information from the API call within the audit trail. This provides a deeper traceability of what is occurring across the cluster. We also capture all login events at the default logging level and modifications to role binds and SCC.

OpenShift Container Platform now has an audit policyFile or policyConfiguration where administrators can filter in on what they want to capture.

See Advanced Audit for more information.

The rolebinding and RBAC experience is now the same across all Kubernetes distributions.

Administrators do not have to do anything for this migration to occur. The upgrade process to OpenShift Container Platform 3.7 offers a seamless experience. Now, the user experience is consistent with upstream.

A role can be defined within a namespace with a Role, or cluster-wide with a ClusterRole.

A RoleBinding or ClusterRoleBinding binds a role to subjects. Subjects can be groups, users, or service accounts. A role binding grants the permissions defined in a role.

Administrators now have the ability to set different token timeouts for the different ways users connect to OpenShift Container Platform (for example, via the oc command line, from a GitHub authentication, or from the web console).

Administrators can edit oauthclients and set the accessTokenMaxAgeSeconds to a time value in seconds that meets their needs.

There are three possible OAuth client types:

See Other API Objects for more information.

flexVolumes allow users to integrate with new APIs easily by being able to mount in the items needed for integration. For example, the ability to bind mount in certain files without overwriting whole directories to integrate with Kerberos.

Administrators are now able to grant access to users to use specific flexVolume driver names. Previously, the only way administrators could restrict flexVolumes was by setting them as on or off.

Local storage persistent volumes is a feature currently in Technology Preview and not for production workloads.

Local persistent volumes (PVs) now offer the ability to allow tenants to request storage that is local to a node through the regular persistent volume claim (PVC) process without needing to know the node. Local storage is commonly used in data store applications.

The administrator needs to create the local storage on the nodes, mount them under directories, and then manually create the persistent volume (PV). Alternatively, they can use an external provisioner and feed it the node configuration via configMaps.

Example persistent volume named example-local-pv that some tenants can now claim:

See Configuring for Local Volume and Persistent Storage Using Local Volume for more information.

Tenant-driven storage snapshotting is currently in Technology Preview and not for production workloads.

Tenants now have the ability to leverage the underlying storage technology backing the persistent volume (PV) assigned to them to make a snapshot of their application data. Tenants can also now restore a given snapshot from the past to their current application.

An external provisioner is used to access the EBS, GCE pDisk, and HostPath, and Cinder snapshotting API. This Technology Preview feature has tested EBS and HostPath. The tenant must stop the pods and start them manually.

Public clouds are particular about not allowing storage to cross zones or regions, so tenants need an ability at times to specify a particular zone.

In OpenShift Container Platform 3.7, administrators can now leverage a zone’s definition within the StorageClass:

See Dynamic Provisioning and Creating Storage Classes for more information.

Container-native storage (CNS) on OpenShift Container Platform 3.7 now supports much higher persistent volume density (three times more) to support a large number of applications at scale. This is due to the introduction of brick-multiplexing support in GlusterFS.

Over 1,000 volumes in a 3-node cluster with 32 GB of RAM per node available to GlusterFS has been successfully tested. Also, 300 Block PVs are supported now on 3-node CNS.

Container-native storage (CNS) is now extended support iSCSI and S3 back end for OpenShift Container Platform. Heketi is enhanced to support persistent volume (PV) expansion, volume option, and HA.

Block device-based RWO implementation is added to CNS to improve the performance of ElasticSearch, PostgreSQL, and so on. With OpenShift Container Platform 3.7, Elastic and Cassandra are fully supported.

Container-native storage (CNS) now fully supports all OpenShift Container Platform infrastructure services: registry, logging, and metrics.

OpenShift Container Platform logging (with Elasticsearch) and OpenShift Container Platform metrics (with Cassandra) are fully supported on persistent volumes backed by CNS/CRS iSCSI block storage.

The OpenShift Container Platform registry is hosted on CNS/CRS by RWX persistent volumes, providing high availability and redundancy through Gluster architecture.

Logging and metrics were tested at scale with 1000+ pods.

OpenShift Container Platform 3.7 now includes an integrated and simplified installation of container-native storage (CNS) through the advanced installer. The advanced installer is enhanced for automated and integrated support for deployment of CNS including block provisioner, S3 provisioner, and files for correctly configured out-of-the-box OpenShift Container Platform and CNS. The CNS storage device details are added to the installer’s inventory file. The installer manages configuration and deployment of CNS, its dynamic provisioners, and other pertinent details.

There is now a supported interface to allow you to bind and mount in content from a running pod. FlexVolume is a script interface that runs on the kubelet and offers five main functions to help you mount in content such as device drivers, secrets, and certificates as bind mounts to the container from the host:

Updated guidance around Cluster Limits for OpenShift Container Platform 3.7 is now available.

The Tuned Profile Hierarchy is updated as of 3.7.

Guidance regarding use of Cluster Loader is now available with the release of OpenShift Container Platform 3.7. Cluster Loader is a tool that deploys large numbers of various objects to a cluster, which creates user-defined cluster objects. Build, configure, and run Cluster Loader to measure performance metrics of your OpenShift Container Platform deployment at various cluster states.

In OpenShift Container Platform 3.7, guidance about the benefits of using the Overlay Graph Driver with SELinux is now available.

Guidance on Providing Storage to an etcd Node Using PCI Passthrough with OpenStack is now available.

Network Policy is now fully supported in OpenShift Container Platform 3.7.

Network Policy is a specification of how groups of pods are allowed to communicate with each other and other network endpoints. It provides fine-grained network namespace isolation using labels and port specifications.

For example, the allow-to-red policy specifies all red pods in the same namespace as the NetworkPolicy object allow traffic from any pods in any namespace.

See Managing Networking for more information.

Cluster IP ranges are now more flexible by allowing multiple subnets for hosts. This provides the capability to allocate multiple, smaller IP address ranges for the cluster. This makes it easier to migrate from one allocated IP range to another.

There are multiple comma-delimited CIDRs in the configuration file. Each node is allocated only a single subnet from within any of the available ranges. You can not allocate different-sized host subnets, or use this to change the host subnet size. The clusterNetworkCIDRs can be different sizes, but must be equal to or larger than the host subnet size. It is not allowed to have some nodes use subnets that are not part of the clusterNetworkCIDRs. Nodes can allocate different-sized subnets by setting different hostSubnetLength values.

In regard to migration or edits, networks can be added to the list, CIDRs in the list may be re-ordered, and a CIDR can be removed from the list when there are no nodes that have an SDN allocation from that CIDR.

Example:

The HAProxy router can look for a cookie in a client request. Based on that cookie name and value, always route requests that have that cookie to the same pod instead of relying upon the client source IP, which can be obscured by an F5 doing load balancing.

A cookie with a unique name is used to handle session persistence.

Example scenario:

See Route-specific Annotations for more information.

HTTP Strict Transport Security (HSTS) ensures all communication between the server and client is encrypted and that all sent and received responses are delivered to and received from the authenticated server.

An HSTS policy is provided to the client via an HTTPS header (HSTS headers over HTTP are ignored) using an haproxy.router.openshift.io/hsts_header annotation to the route. When the Strict-Transport-Security response in the header is received by a client, it observes the policy until it is updated by another response from the host, or it times-out (max-age=0).

Example using reencrypt route:

As a cluster administrator, you can assign specific, static IP addresses to projects, so that traffic is externally easily recognizable. This is different from the default egress router, which is used to send traffic to specific destinations.

Recognizable IP traffic increases cluster security by ensuring the origin is visible. Once enabled, all outgoing external connections from the specified project will share the same, fixed source IP, meaning that any external resources can recognize the traffic.

Unlike the egress router, this is subject to EgressNetworkPolicy firewall rules.

See Managing Networking for more information.

You can gain external access to the OpenShift Container Platform registry using a re-encrypt route. This allows you to log in to the registry from outside the cluster using the route address and to tag and push images using the route host.

A public pull URL is provided for images versus being able to know the internal in-cluster IP or DNS of the service.

A new API field for the image stream with the public URL of the image was added, and a public URL is configured in the master-config.yaml file. The web console will understand this new field and generate the public pull specifications automatically to users (so users can just copy and paste the pull URL).

Example:

A resource is an endpoint in the Kubernetes API that stores a collection of API objects of a certain kind (for example, pod objects). A custom resource definition is a built-in API that enables the ability to plug in your own custom, managed object and application as if it were native to Kubernetes. Therefore, you can leverage Kubernetes cluster management, RBAC and authentication services, PI services, CLI, security, and so on, without having to know Kubernetes internals or modifying Kubernetes itself in any way.

Custom Resource Definitions (CRD) deprecates Third Party Resources in Kubernetes 1.7.

How it works:

See Extending the Kubernetes API with Custom Resources for more information.

There is now Kubernetes documentation on how API aggregation works in OpenShift Container Platform 3.7 and how other users can add third-party APIs:

Prometheus endpoint logic was added to upstream components so that monitoring and health indicators can be added around deployment configurations.

Starting in OpenShift Container Platform 3.7, the use of the etcd3 v3 data model is required.

OpenShift Container Platform gains performance improvements with the v3 data model. In order to upgrade the data model, an embedded etcd configuration option in no longer allowed. Embedded is not co-located. Migration scripts will convert the v3 data model and allow you to move an embedded etcd to an external etcd either on the same host or a different host than the masters. In addition, there is a new scale up ability for etcd clusters.

See Migrating Embedded etcd to External etcd for more information.

The installer has been enhanced to allow administrators to install specific components. By breaking up the roles and playbooks, there is better targeting of ad hoc administration tasks.

When you run the installer, OpenShift Container Platform now reports back at the end what phases you have gone through.

If the installation fails during a phase, you will be notified on the screen along with the errors from the Ansible run. Once you resolve the issue, rather than run the entire installation over again, you can pick up from the failed phase. This results in an increased level of control during installations and results in time savings.

With OpenShift Container Platform 3.7, there is added control over whether or not your cluster automatically upgrades all the content provided during cluster upgrades.

Edit the openshift_install_examples variable in the hosted file or set it as a variable in the installer.

Setting openshift_install_examples to false will cause the installer to not upgrade the imagestream and templates. True is the default behavior.

Red Hat CloudForms Management Engine (CFME) 4.6 is now fully supported running on OpenShift Container Platform 3.7 as a set of containers.

CFME is an available API endpoint on all OpenShift Container Platform clusters that choose to use it. More cluster administrators are now able to leverage CFME and begin experiencing the insight and automations available to them in OpenShift Container Platform.

To install CFME 4.6:

To configure CFME 4.6 to consume the OpenShift Container Platform installation it is running on:

You can also automate the configuration of the provider to point to multiple OpenShift clusters:

More health checks are now available for administrators to run after installations and upgrades. Administrators need the ability to run tests periodically to help determine the health of the framework components within the cluster. OpenShift Container Platform 3.7 offers test functionality via Ansible playbooks that can be run and output can be sent as file-based output.

Alternatively, they are included in the health playbook:

To capture the output:

Fluentd captures standard error and standard out from the running containers on the node. With this change, fluentd collects all the errors and events coming from the docker daemon running on the node and sends it to Elasticsearch (ES).

Enable this via the OpenShift Container Platform installer:

The collected information is in operation indices of ES and only cluster administrators have visual access. The event message includes action, pod name, image name, and user time-stamp.

The eventrouter pod scrapes the events from kubernetes API and and outputs to STDOUT. The fluentd plug-in transforms the log message and sends it to Elasticsearch (ES).

Enable openshift_logging_install_eventrouter by setting it to true. It is off by default. Eventrouter is deployed to the default namespace. Collected information is in operation indices of ES and only cluster administrators have visual access.

See the design documentation for more information.

This allows OpenShift Container Platform administrators the ability to share saved Kibana searches, visualizations, and dashboards.

When openshift_logging_elasticsearch_kibana_index_mode is set to shared_ops, one admin user can create queries and visualizations for other admin users. Other users can not see those same queries and visualizations.

When openshift_logging_elasticsearch_kibana_index_mode is set to unique, users can only see saved queries and visualizations they created. This is the default behavior.

See Aggregating Container Logs for more information.

ES_Copy was replaced with the secure_formard plug-in for fluentd to send logs from fluentd to external fluentd (that can then ingest into ES). ES_COPY is removed from the installer and the documentation.

When openshift_installer is run for logging to upgrade to 3.7, the installer now checks for ES_COPY in the inventory and fails the upgrade with:

See Aggregating Container Logs for more information.

By default, Elasticsearch (ES) deployed with OpenShift aggregated logging is not accessible from outside the logging cluster. This enables a route for external access to ES for those tools that want to access its data.

You now have direct access to ES using only your OpenShift token and have the ability to provide the external ES and ES Ops hostnames when creating the server certificate (similar to Kibana). Ansible tasks now simplify route deployment.

The metrics and logging deployers bare now replaced with playbook2image for oc cluster up so that openshift-ansible is used to install logging and metrics:

Check metrics and pod status:

OpenShift Container Platform operators deploy Prometheus (currently in Technology Preview and not for production workloads) on a OpenShift Container Platform cluster, collect Kubernetes and infrastructure metrics, and get alerts. Operators can see and query metrics and alerts on the Prometheus web dashboard, or bring their own Grafana and hook it up to Prometheus.

See Prometheus on OpenShift for more information.

Hawkular OpenShift Agent (HOSA) remains in Technology Preview and not for production workloads. It is packaged and can now be installed with the openshift_metrics_install_hawkular_agent option in the installer by setting it to true.

See Enabling Cluster Metrics for more information.

Users can seamlessly configure, deploy, and scale AWS services like Amazon RDS, Amazon Aurora, Amazon Athena, Amazon Route 53, and AWS Elastic Load Balancing directly within the OpenShift Container Platform console. For installation instructions, see AWS Service Broker and Getting Started Guide.

Clients can now easily invoke a server API instead of relying on client logic.

OpenShift Container Platform now includes:

CLI plug-ins are currently in Technology Preview and not for production workloads.

Usually called plug-ins or binary extensions, this feature allows you to extend the default set of oc commands available and, therefore, allows you to perform new tasks.

See Extending the CLI for information on how to install and write extensions for the CLI.

In OpenShift Container Platform 3.7, Chaining Builds is a better approach for producing runtime-only application images, and fully replaces the Extended Builds feature.

Benefits of Chaining Builds include:

OpenShift Container Platform uses the following default configuration for eviction-hard.

See Handling Out of Resource Errors for more information.

In OpenShift Container Platform 3.7, Open Service Broker API is implemented, enabling users to leverage Ansible for provisioning and managing services from the Service Catalog. This is a standardized approach for delivering simple to complex multi-container OpenShift services via Ansible. It works in conjunction with Ansible Playbook Bundle (APB) for lightweight application definition. APBs can be used to deliver and orchestrate on-platform services, but could also be used to provision and orchestrate off-platform services (from cloud providers, IaaS, and so on).

OpenShift Ansible Broker supports production workloads and multiple service plans. There is now secure connectivity between Service Catalog and Service Broker.

You can interact with the Service Catalog to provision and manage services while the details of the broker remain largely hidden.

Ansible Playbook Bundles (APBs) are short-lived, lightweight container image consisting of:

Developer tooling is included, providing a guided approach to APB creation. There is also support for the test playbook, allowing for functional testing of the service.) Two new APBs are introduced for MariaDB (SCL) and MySQL DB (SCL).

When a user provisions an application from the Service Catalog, the Ansible Service Broker will download the associated APB image from the registry and run it.

Developing APBs can be done in one of two ways: Creating the APB container image manually using standardized container creation tooling, or with APB tooling that Red Hat will deliver, which provides a guided approach to creation.

The OpenShift Template Broker exposes templates through a Open Service Broker API to the Service Catalog.

The Template Broker matches the lifecycles of provision, deprovision, bind, and unbind with existing templates. No changes are required to templates, unless you expose bind. Your application will get injected with configuration details.

OpenShift Container Platform 3.7 provides a better initial user experience with the Service Catalog. This includes:

The new user interface is designed to really streamline the getting started process, in addition to incorporating the new Service Catalog items. It shows the existing content (for example, builder images and templates) as well as catalog items (if the catalog is enabled).

OpenShift Container Platform 3.7 provides a simple way to quickly get what you want The new Search Catalog user interface is designed to make it much easier to find items in a number of ways, making it even faster to find the items you are wanting to deploy.