New built-in admission controller replaces PodSecurityPolicy ◦ PSP slated for removal in 1.25 ◦ CIS guidelines still call for using PSPs ◦ OpenShift’s SCCs are unaffected Major Themes and Features • API deprecation ◦ Affects many popular APIs (beta→stable) ◦ Marked as deprecated for many releases, finally removed • CSI for Windows nodes is GA CRI-O 1.22 Kubernetes 1.22 OpenShift 4.9 Blog: https://cloud.redhat.com/blog/whats-new-in-kubernetes-v1.22 5 Kubernetes 1.22
DEV • OpenShift Builds v2 & Buildpacks TP • Shared Resource CSI Driver GA • Tekton Hub on OpenShift • Unprivileged builds • Image build cache • Manual approval in pipelines • Global Operators Model & new Operator API • Operator Maturity increase via SDK • Serverless Functions Orchestration • Stateful Functions HOST • Cost mgmt integration to Subs Watch, ACM • Detailed Quota Usage in cluster manager • ROSA/OSD: AWS Dedicated instances • Dev Preview of App Studio, a hosted developer experience • OpenShift Serverless Functions IDE Experience • OpenShift Dev CLI (odo onboarding & more) 2022+ • OpenShift Serverless Functions GA • Encryption of inflight data natively in Serverless • Integration of Knative(Serverless) with KEDA • OpenShift Serverless Kafka Broker • Operator SDK for Java (Tech Preview) • File-based operator catalog management • MetalLB BGP support • Azure Stack Hub (IPI) • Alibaba, Nutanix & IBM Cloud (UPI/IPI) • OpenShift on ARM (AWS and Bare Metal) • SRO manages third party special devices • Additional capabilities for Windows containers: health management, 3rd party network plugins such as Calico • ZeroTouchProvisioning and Central infrastructure management GA (ACM) • ExternalDNS Support • NetFlow/sFlow/IPFIX Collector • Introduce Gateway API • ROSA/OSD: FedRAMP High on AWS GovCloud • ROSA/OSD: Terraform provider • ROSA/OSD/ARO: GPU Support • ARO: Upgrades through cluster manager • Cost management understands IBM Cloud IaaS H1 2022 HOSTED PLATFORM Q4 2021 APP APP DEV • OpenShift Pipelines - Tekton Triggers GA • Tekton Hub integration in Pipeline builder • Automated retrieval of RHEL entitlement • Mount Secret and ConfigMap in BuildConfigs • Export Application (Dev Preview) • ROSA: cluster manager UI for ROSA provisioning • ROSA/OSD: Cluster Hibernation • ARO: Azure Portal UI for ARO provisioning • Cost: Improved models for distribution of costs HOST PLATFORM • OpenShift Serverless cold start improvements • Dynamic Plugins for the OCP Console • MetalLB Support (L2) • Azure Stack Hub (UPI) • RHEL 8 Server Compute/Infra Nodes • AWS support for China Regions • Single Node OpenShift (UPI) • Custom audit profiles by group • OpenShift API compatibility level discovery tools • API for Custom Route Name and Certificates • Operator Metering end of life • GA of BYOH Windows host for Windows Containers + additional supported platforms • Disable case-sensitivity for case-insensitive IdPs • Multi Service-Serving-Certificates for Headless StatefulSet • Service Mesh federation between meshes/clusters • ZeroTouchProvisioning and Central infrastructure management Tech Preview (ACM) • Azure China • Utilize cgroups v2 • Expand cloud providers for OpenShift on ARM • Enable user namespaces • Windows Containers: CSI proxy, improved monitoring/logging & more platforms supported • Gateway API / Ingress Controller support • Network Topology and Analysis Tooling • SmartNIC Integrations • eBPF Support • Network Policy v2 & OVN no-overlay option • BGP Advertised Services (FRR) • Service Mesh on VMs • SigStore style image signature verification • Disconnected mirroring simplification
• External software interacting with a cluster may use deprecated APIs. • To prevent breakage, an admin will acknowledge external software has been updated prior to cluster upgrade • This “ack” is a boolean on a ConfigMap • We expect to use this functionality for similar changes of this magnitude in the future Affected APIs • CRD (beta→stable) • CertificateSigningRequest (beta→stable) • Mutating/ValidatingAdmissionWebhook (b→s) • Full list and more details Operators • Change affects Operators that still use a beta CRD • Partners and layered products have been audited and notified of updates they require • Operators installed in 4.8 that do not have a compatible 4.9 release will block cluster upgrade API Removal and Upgrade Behavior apiVersion: v1 kind: ConfigMap metadata: name: admin-acks namespace: openshift-cluster-version data: ack-4.8-kube-122-api-removals-in-4.9: "true" PM: Tony Wu 8
cases for Bare Metal • Does not have a workload runtime dependency on a central control plane • Bootstrap In Place - no additional bootstrap node needed • Upgrade support • Deployment via openshift-install (GA) • Deployment via RHACM (ZTP/CIM) /Assisted installer (TP) • OLM available to install Operators • 8 cores 32GB mem minimal requirements • ~2 cores 16GB platform footprint (vanilla OCP) Single Node OpenShift PM: Moran Goldboim Consistent application platform from the datacenter to the edge https://www.youtube.com/watch?v=QFf0yVAHQKc 9
Marc Curry, Deepthi Dharwar • MetalLB has two modes to announce reachability information for load balancer IP addresses: ◦ Layer 2 (4.9) ◦ BGP (4.10) • Two components: ◦ Controller - One per cluster ◦ Speaker - Per Node (DaemonSet) • L2 mode: ARP (IPv4) or NDP (IPv6) announces location of a LB’d IP address from the Speaker, then relies on Service load balancing within the cluster • BGP mode: Traffic can target multiple nodes – routers can perform load balancing across the cluster using ECMP apiVersion: v1 kind: Service metadata: name: nginx spec: ports: - name: http port: 80 protocol: TCP targetPort: 80 selector: app: nginx type: LoadBalancer
• Tetkon Triggers GA • Auto-pruning configurations per namespace • Pipeline as code ◦ Private Git repository support ◦ Hosted BitBucket support • Granular observability and metrics configurations • CRD introduced for customizing Tekton configs • (Dev Console) Search and install Tasks from TektonHub in the Pipeline builder • (Dev Console) Repository list views for pipeline as code OpenShift Pipelines PM: Siamak Sadeghianfar 11
• OpenShift GitOps 1.3 • User groups and kube-admin support when log into Argo CD with OpenShift credentials • ApplicationSet integration with RHACM for cluster lookup • kustomize 4 support • External cert manager support for TLS configs in Argo CD • Router sharding for Argo CD • (Dev Console) Application deployment environment details 12
& Updates ❖ Update to Knative 0.24 ❖ Security: Encryption of Inflight Data with Service Mesh ❖ Custom Domain Mapping through DevConsole ❖ Visualization: New Monitoring Dashboards ➢ CPU, Memory, Network Usage ➢ Scaling Debugging ➢ User workload monitoring through Knative Queue Proxy ❖ Support for emptyDir ➢ Share files between sidecar and the main container ❖ Functions Tech Preview: ➢ Node, Quarkus, Python, Go, SpringBoot, TypeScriptNew, RustNew ➢ Access to data stored in secrets and config maps ➢ Available on MacOS , RHEL, Windows with Docker and/or Podman PM: Naina Singh OPENSHIFT OpenShift Serverless SERVING EVENTING Red Hat Enterprise Linux CoreOS Physical Virtual Private cloud Public cloud Applications/Functions Events F FUNCTIONS* Edge
Builds OpenShift • Tech preview in 4.9 • Insights Operator pulls RHEL entitlements for OpenShift clusters • Simple Content Access (SCA) must be enabled for customer’s Red Hat account (by the customer) • Entitlements stored as Secret named etc-pki-entitlement in the openshift-config-managed namespace • Entitlements rotated and refreshed regularly • Admin responsible to distribute entitlement secret to namespaces • Mount entitlement secret into Pods and Tekton for entitled builds • Mount entitlement and other credential secrets (or configmaps) in BuildConfigs for entitled builds 14 Insights Operator openshift-config-managed etc-pki-entitlement cloud.redhat.com (OCM) PM: Siamak Sadeghianfar
Manager: Gaurav Singh 15 Multi-component/microservice deployments before New option: Multi-component/microservice deployments with 4.9 OpenShift Global Pull Secret: - quay.io/openshift-release-dev - quay.io/foo/ - quay.io/bar/ - quay.io/baz/image Image Pull Secret, e.g. quay.io/foo/image:v1 Image Pull Secret, e.g. quay.io/bar/image:v1 Image Pull Secret, e.g. quay.io/baz/image:v1 Multiple Secrets containing different registry credentials per Deployment / image Simplified registry credential management using a single Secret containing different logins, even for the same registry
Mobrem & Serena Chechile Nichols Focus on the projects you care about • Hide system projects with the default projects toggle for privileged users • Star favorite projects • Quick project search User Preference • Set your individual preferences for the console experience • Changes will be autosaved 17
Mobrem Troubleshoot Node-level problems directly from the Console • View Node logs in the Console just like you can with your Pods • Ability to filter journald by unit Clean Operator uninstall • Cleanly remove all workloads, applications & resources managed by operator when uninstalling the operator itself Break down cluster utilization by node type • Defaults to Master & Work node types • Additional node types will auto appear in the list once created 18
Full Stack Automation (IPI) Pre-existing Infrastructure (UPI) Bare Metal Product Manager(s): Marcos Entenza (AWS, Azure, Azure Stack Hub, GCP), Maria Bracho (VMware), Peter Lauterbach (RHV & OCP Virtualization), Anita Tragler (OSP), Ramon Acedo Rodriguez (BM), & Duncan Hardie (IBM Z & Power) IBM Power Systems 23 Bare Metal NEW Azure Stack Hub
Hub Installing a cluster on user-provisioned infrastructure • Azure’s solution to run applications in an on-premises environment and deliver Azure services in your data centre. • Allows an OpenShift cluster to be deployed to an user-provisioned infrastructure on Azure Stack Hub. • Option to use the provided Azure Resource Manager (ARM) templates to assist with the installation. Generally Available Product Manager: Marcos Entenza 24 apiVersion: v1 baseDomain: example.com controlPlane: name: master replicas: 3 compute: - name: worker platform: {} replicas: 0 metadata: name: ash-cluster networking: clusterNetwork: - cidr: 10.128.0.0/14 hostPrefix: 23 machineNetwork: - cidr: 10.0.0.0/16 networkType: OpenShiftSDN serviceNetwork: - 172.30.0.0/16 platform: azure: armEndpoint: azurestack_arm_endpoint baseDomainResourceGroupName: resource_group region: azure_stack_local_region resourceGroupName: existing_resource_group outboundType: Loadbalancer cloudName: AzureStackCloud pullSecret: '{"auths": ...}' fips: false sshKey: ssh-ed25519 AAAA... Azure Stack Hub
and infra nodes Support of Red Hat Enterprise Linux 8 • RHEL 8 machines can be added to any UPI or IPI deployed cluster in day-2. • OCP 4.9 starts with RHEL 8.4. • Adding RHEL 7 machines to OCP is deprecated and support for RHEL7 workers will be removed in OCP 4.10 • RHEL 7 compute machines cannot be upgraded to RHEL 8, new RHEL 8 compute machines must be deployed. Product Manager: Marcos Entenza 25
Product Manager: Marcos Entenza Increased subnet size within the machine CIDR • When installing OpenShift on Azure, subnets are created as large as possible within the machine CIDR • Allows the cluster to use a machine CIDR appropriately sized to accommodate the number of nodes in the cluster. OpenShift 4.8 or below OpenShift 4.9 networking: clusterNetwork: - cidr: 10.128.0.0/14 hostPrefix: 23 machineNetwork: - cidr: 10.0.0.0/16 networkType: OpenShiftSDN serviceNetwork: - 172.30.0.0/16 networking: clusterNetwork: - cidr: 10.128.0.0/14 hostPrefix: 23 machineNetwork: - cidr: 10.0.0.0/16 networkType: OpenShiftSDN serviceNetwork: - 172.30.0.0/16 8190 hosts per subnet 32766 hosts per subnet 26
Deploy OpenShift to AWS China regions • New China Regions Beijing (cn-north-1) and Ningxia (cn-northwest-1) are now available to deploy OpenShift Container Platform. • An Internet Content Provider (ICP) license is required to use these Regions. • RHCOS images must be manually uploaded to your AWS account. • The AWS Region and the accompanying custom AMI must be manually included in the installation configuration file. Generally Available Product Manager: Marcos Entenza 27 apiVersion: v1 baseDomain: example.com ... ... ... platform: aws: region: cn-north-1 userTags: adminContact: mak costCenter: 7536 subnets: - subnet-1 - subnet-2 - subnet-3 amiID: ami-96c6f8f7 serviceEndpoints: - name: ec2 url: https://vpce-id.ec2.cn-north-1.vpce.amazonaws.com.cn hostedZone: Z3URY6TWQ91KVV fips: false sshKey: ssh-ed25519 AAAA... publish: Internal pullSecret: '{"auths": ...}'
Management 2.4 (Infrastructure Operator) Provides a separate interfaces for: -Infra-Admin (IT) - to manage on-prem compute across different datacenters/locations -Cluster creator (Dev/Ops) - to consume allocated compute resources for clusters creation • Fully integrated with ACM • Consisted UXD with Assisted installer (SAAS) • Integrated preflight checks, monitoring and eventing • K8S native API • Any type of OpenShift deployment (SNO, RWN, Compact..) for Bare metal and platform agnostic 29
Cloud® Bare Metal platform: baremetal: apiVIP: <api-ip> ingressVIP: <wildcard-ip> provisioningNetworkInterface: <NIC1> provisioningNetworkCIDR: <CIDR> hosts: - name: openshift-master-0 role: master bmc: address: ipmi://10.196.130.145?privilegelevel=OPERATOR username: root password: <password> bootMACAddress: 00:e0:ed:6a:ca:b4 rootDeviceHints: deviceName: "/dev/sda" - name: <openshift-worker-0> role: worker bmc: address: ipmi://10.196.130.146?privilegelevel=OPERATOR username: <user> password: <password> bootMACAddress: <NIC1-mac-address> rootDeviceHints: deviceName: "/dev/sda" Developers who require their apps to run on physical nodes can now provision and dispose of OpenShift clusters on bare metal from IBM Cloud® using the IPI bare metal installation workflow to install OpenShift on IBM Cloud®‘s physical nodes Note that this isn’t using a cloud provider for IBM Cloud®, it uses the regular bare metal IPI workflow with servers from IBM Cloud® IBM Cloud server BMC PM: Ramon Acedo Rodriguez 30 IBM Cloud server BMC $ ibmcloud sl hardware create --hostname <SERVERNAME> \ --domain <DOMAIN> \ --size <SIZE> \ --os <OS-TYPE> \ --datacenter <DC-NAME> \ --port-speed <SPEED> \ --billing <BILLING> Create your IBM Cloud physical servers with your account Use the standard bare metal IPI workflow to provision your cluster on IBM Cloud
cluster with remote worker nodes $ oc edit provisioning [...] spec: [...] virtualMediaViaExternalNetwork: true Add remote worker nodes any time over virtual media You can now Install your cluster locally with DHCP/PXE as usual and later add remote worker nodes outside of your provisioning network The Bare Metal Operator will map the installation image as virtual media on the remote node’s BMC with the advantage of not needing L2-adjacency (as PXE-booting) PM: Ramon Acedo Rodriguez 31 Simply ask your provisioning custom resource (CR) to use the external network for virtual media
evenly across nodes HighNodeUtilization Pack as many pods as possible on to as few nodes NoScoring Quickest scheduling cycle by disabling all score plugins Customize default out of box behaviour of openshift scheduler with Scheduling Profiles *Note: in OSP 4.7 customer can use both policy API and profiles but going forward policy API will be depreciated to profiles Extension points Scheduling profile Scheduling plugin Extension points Scheduling plugin Add more Scheduling plugins Pre Build Profile Build your own Profile Scheduling profile : Openshift-scheduler can have only one profile Scheduling plugin : Implements one or more extension points Extension point : Plugins that define the scheduling logic 33
for OpenShift Cluster Components now allows for any level of flexibility in customers environments. The current <name>.apps.<cluster>.<domain> can be customized for the OAuth server and the the OCP console. • The OAuth server route can be customized using the ingress config route configuration API. A custom hostname and a TLS certificate can be set using the spec.componentRoutes part of the configuration.Set the custom hostname and optionally configure the serving certificate and key. Custom Route Name and Certs for certain cluster components PM: Anand Chandramohan Component Custom Route supported? OAuth Yes (from 4.9) Console Yes (from 4.8) Downloads Yes (from 4.8) Monitoring (AlertManager, Prometheus, Grafana, Thanos) No Image Registry No apiVersion: config.openshift.io/v1 kind: Ingress metadata: name: cluster spec: componentRoutes: - name: oauth-openshift namespace: openshift-authentication hostname: <custom-hostname> servingCertKeyPairSecret: name: <secret-name>
policy (introduced in OpenShift 4.6) Control the amount of information that is logged to the API audit logs by choosing the audit log policy profile to use. • Default: Logs only metadata for read and write requests; does not log request bodies except for OAuth access token requests. This is the default policy. • WriteRequestBodies: In addition to logging metadata for all requests, logs request bodies for every write request to the API servers (create, update, patch). This profile has more resource overhead than the Default profile. • AllRequestBodies: In addition to logging metadata for all requests, logs request bodies for every read and write request to the API servers (get, list, create, update, patch). This profile has the most resource overhead. apiVersion: config.openshift.io/v1 kind: APIServer metadata: ... spec: audit: profile: WriteRequestBodies The Default audit log policy in OCP 4.8 logged request bodies for OAuth access token creation (login) and deletion (logout) requests. Previously, deletion request bodies were not logged. Improved customization of Audit Config 35 PM: Anand Chandramohan
audit log policy that defines custom rules (new in 4.9). You can specify multiple groups and define which profile to use for that group. These custom rules take precedence over the top-level profile field. The custom rules are evaluated from top to bottom, and the first that matches is applied apiVersion: config.openshift.io/v1 kind: APIServer metadata: ... spec: audit: customRules: - group: system:authenticated:oauth profile: WriteRequestBodies - group: system:authenticated profile: AllRequestBodies profile: Default • Add one or more groups and specify the profile to use for that group. These custom rules take precedence over the top-level profile field. The custom rules are evaluated from top to bottom, and the first that matches is applied. • Set audit log profile to Default, WriteRequestBodies, AllRequestBodies, or None. If you do not set this top-level audit.profile field, it defaults to the Default profile. Configuring the Audit log policy with custom rules 36 PM: Anand Chandramohan
logging for OpenShift Container Platform. When you disable audit logging, even OAuth access token requests and OAuth authorize token requests are not logged. kind: APIServer metadata: ... spec: audit: profile: None • You can also disable audit logging only for specific groups by specifying custom rules in the spec.audit.customRules field. Disable Audit Logging 37 PM: Anand Chandramohan
38 TLS security profiles provide a way for servers to regulate which ciphers a client can use when connecting to the server. This ensures that OpenShift Container Platform components use cryptographic libraries that do not allow known insecure protocols, ciphers, or algorithms. apiVersion: config.openshift.io/v1 kind: APIServer metadata: name: cluster spec: tlsSecurityProfile: type: Custom custom: ciphers: - ECDHE-ECDSA-CHACHA20-POLY1305 - ECDHE-RSA-CHACHA20-POLY1305 - ECDHE-RSA-AES128-GCM-SHA256 - ECDHE-ECDSA-AES128-GCM-SHA256 minTLSVersion: VersionTLS11 Name: cluster Namespace: ... API Version: operator.openshift.io/v1 Kind: Etcd ... Spec: Log Level: Normal Management State: Managed Observed Config: Serving Info: Cipher Suites: TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256 TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384 TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256 TLS_ECDHE_RSA_WITH_CHACHA20_POLY1305_SHA256 Min TLS Version: VersionTLS12 Add the spec.tlsSecurityProfile field: Verify that the TLS security profile is set in the etcd CR: PM: Anand Chandramohan
39 etcd certificates are used for encrypted communication between etcd member peers, as well as encrypted client traffic. The following certificates are generated and used by etcd and other processes that communicate with etcd: Peer certificates: Used for communication between etcd members. Client certificates: Used for encrypted server-client communication. Client certificates are currently used by the API server only, and no other service should connect to etcd directly except for the proxy. Client secrets (etcd-client, etcd-metric-client, etcd-metric-signer, and etcd-signer) are added to the openshift-config, openshift-monitoring, and openshift-kube-apiserver namespaces. Server certificates: Used by the etcd server for authenticating client requests. Metric certificates: All metric consumers connect to proxy with metric-client certificates. etcd certificates are signed by the etcd-signer; they come from a certificate authority (CA) that is generated by the bootstrap process. The CA certificates are valid for 10 years. The peer, client, and server certificates are valid for three years. These certificates are only managed by the system and are automatically rotated. PM: Anand Chandramohan
40 This feature enables an automated mechanism which provides defragmentation as the result of observations from the cluster. The goal is to provide a controller which manages automation of etcd defragmentation based on observable thresholds. A defrag controller has been added to the cluster-etcd-operator that will check the endpoints of the etcd cluster every 10 minutes for observable signs of fragmentation. The criteria for defragmentation is defined as: • Cluster must be observed as HighlyAvailableTopologyMode. • The etcd cluster must report all members as healthy. • minDefragBytes this is a minimum database size in bytes before defragmentation should be considered. 100 MB. • maxFragmentedPercentage this is a percentage of the store that is fragmented and would be reclaimed after defragmentation. 45%. Benefits • Large scale clusters, that previously felt the pain of lack of fragmentation (resource bloat, OOM events, down time) can now can now benefit from the defrag controller that manages the maintenance of etcd backend • Heavy churn workloads do not negatively impact cluster performance. • Cluster operands utilize only the necessary resources to provide service PM: Anand Chandramohan
OVN Clusters 42 PM: Marc Curry, Deepthi Dharwar • Ability to assign one or more source egress IPs to a project or pod, for filtering at the edge of the cluster • Label a set of nodes as “egress nodes” to route EgressIP-related traffic to external endpoints • EgressIP traffic (non-regular) is SNAT-ed to the defined egress IPs on the egress nodes • Multiple egress nodes can be specified to avoid “choke points” • Traffic is LB’d between participant nodes (using OVS’s dp_hash)
Curry, Deepthi Dharwar Support for Network Adapters in Fast Data Path list on OpenShift • Support Matrix • Going forward any fast datapath network adapter that RHEL supports will be supported on OpenShift. Hardware Enablement SR-IOV for Single Node OpenShift • Support for running SR-IOV operator on a far less resource -constrained hardware • Onboard realtime and low latency workloads with high performance networking support. Hardware Enablement Support for DPDK & RDMA with SR-IOV • Better throughput and performance. • Moved from Tech Preview to GA DPDK
Deepthi Dharwar Allow Setting mTLS Through the Ingress Operator Support • Support client-TLS which enables router to verify client certificates. • Admin must provide CA certificate to the router. Support TLS 1.3 for OpenShift 4.x Ingress: • Supports faster TLS handshake • Simpler, secure cipher suites • Better performance and stronger security. Ingress Updates HAProxy timeout Variables Customization a. clientTimeout/serverTimeout b. clientFinTimeout/serverFinTimeout c. tunnelTimeout d. tlslnspectDelay Set as part of Ingress controller spec under tuning Options. Global Options to Enforce HTTP Strict Transport Security [HSTS] • Manual per-route annotations to enable HSTS is sub-optimal. • Allow cluster admins to enforce this policy globally with ease and flexibility. Ingress Updates
is GA PM: Franck Baudin Pod with CAP_NET_RAW capability vfr-foo net1: 10.0.0.100/24 eth0: 10.0.0.55/24 vfr-bar net2: 10.0.0.100/24 OVN-K network FOO network BAR Problem statement: How to connect a pod to several networks with overlapping CIDRs? Solution: Create multiple routing and forwarding domains within the pod with Linux VRFs. ▸ VRF CNI run on multus secondary interfaces, as long as they are kernel bound, netdevice only no DPDK: SR-IOV CNI and MACVLAN CNI
Key Features & Updates • Update to Istio 1.9 • New Feature: Service Mesh Federation ◦ Securely connect service meshes across OCP clusters without the Kubernetes API Server ◦ Share services between meshes on a strict “need to know” basis ◦ Manage traffic with remote services as if they were local services • ServiceMeshExtensions API becomes GA ◦ Extend service mesh API using Envoy’s WebAssembly Extensions • Service Mesh 2.1 Release Date: November 2021 Service A Service B Service C Service D Control Plane (Istiod) Control Plane (Istiod) Gateway Gateway OCP Cluster OCP Cluster Service Mesh Service Mesh Federated Service Meshes PM: Jamie Longmuir
Octavia as Service Type Load Balancer (without Kuryr) • OCP 4.7 introduced External Load Balancer [UPI] • OCP 4.9 and OSP 16.1 - Octavia Amphora as Service Load Balancer (L4-L7) with IPI installer ◦ HA-proxy/IPVS based load balancer ◦ HTTP/HTTPS, other TCP (non-http) ◦ Amphora VM per Tenant (project) or OCP cluster ◦ Amphora supports Active/Standby HA ◦ UDP [OCP 4.10 new external Cloud Provider] ◦ SCTP [ OSP 17.x] • Octavia OVN as L4 Load Balancer [OCP 4.9 Tech Preview] ◦ Relies on k8s health monitoring, no member health monitoring [OSP 17.x] ◦ TCP ports (non-http), UDP [OCP 4.10], SCTP ◦ No extra hop (latency) needed since LB service is distributed with OVN • How to set this up? Docs VM Worker1 podA podB 10.2.1.2/24 ToR Tenant Access network: 11.1.1.0/24 eth0 VM Worker2 podD podC eth0 10.2.1.3/24 Octavia (Amphora) External LB 11.1.1.2/24 11.1.1.3/24 GW: 11.1.1.1/24 VIP 10.1.1.2/24 10.1.1.3/24 10.1.2.2/24 10.1.2.3/24 PM: Anita Tragler
Router Sharding Goal - Use Octavia Load balancer service with router sharding in order to isolate DMZ external traffic from internal API traffic OCP 4.9 adds support for Ingress Operator router sharding with Octavia Load balancer • With OCP 4.9, Apps can leverage Router Sharding with Octavia LB • Separate LB service for DMZ and internal API traffic using Labels, namespaces or service mesh • Separate Load balancer for infrastructure internal API and separate LB service for workloads How to set this up? Docs LB PM: Anita Tragler 48
and serverless PM: Peter Lauterbach Public Cloud Support • AWS Bare-metal (Tech Preview) - Consistent environment with on-premise VM workloads to support on-demand scaling and cloud migration Enhanced Data Protection • Crash-consistent online VM snapshots • Improved Data Protection w/ upstream Velero plugin for VM backups (Tech Preview) SAP HANA for test and non-production • SAP HANA enablement for testing and non-production deployments ◦ production certification in a future release Enhanced Security and Performance • Additional modes to boot UEFI guests, High performance workloads with vNUMA • VM workloads in a FIPS compliant OpenShift cluster Operational Enhancements • Hybrid workload with container and VMs in the same service mesh • VM workflow management, easily configure Windows guests with sysprep 50
Miguel Pérez Colino Migration Toolkit for Virtualization 2.1 • Easy to use UI • Mass migration of VMs from VMware and RHV to OpenShift • Added Red Hat Virtualization as supported source provider (Cold Migration only) • Validation service (Tech Preview): Includes SR-IOV cards and Opaque networks that are configured. • Hooks: Automated tasks to be performed pre and post migration • Must-Gather: specific add-ons created to help debug issues during migrations 51
We are proud to announce the General Availability of Bring Your Own Host Support for Windows nodes to Red Hat OpenShift. • With this offering you will be able to onboard your custom Windows nodes (aka pets) into an OpenShift cluster. • We recognize customers have dedicated Windows server instances in their data centers that they regularly update, patch and manage. Often these instances run on vSphere, or Bare Metal platforms. • It is essential to take advantage of these servers to run containerized workloads so their computing power can be harnessed in a hybrid cloud world. • Enabling the Bring Your Own Host Support for these Windows servers can help customers lift and shift their on-premises workloads to a cloud native world. Bring Your Own Hosts (BYOH) for Windows Nodes
for Windows Nodes PM: Anand Chandramohan 53 Windows traditional .NET framework containers Windows application Linux containers .NET core containers Windows containers Linux containers Windows virtual machine Red Hat OpenShift Virtualization Red Hat Enterprise Linux CoreOS Red Hat Enterprise Linux CoreOS Microsoft Windows Mixed Windows and Linux workloads .NET core containers Windows traditional .NET framework containers Windows containers Microsoft Windows .NET core containers Machine API Managed Infrastructure BYOH instance • BYOH instance and Linux worker nodes on the cluster have to be on the same network • The platform type for BYOH must match that set for the OCP cluster
FIPS Compliance Updates & Upgrades Must Gather Disconnected Mode Now you can run the OpenShift sandboxed containers operator on a FIPS enabled cluster without worrying about tainting its state. Our Operator, and Kata Containers are FIPS Validated. You can now seamlessly upgrade a cluster, as well as the operator and its artifacts (Kata Containers + QEMU extensions). An initial version of must-gather will be available in this release. This will help automate data-collection for you to get a better support experience. 54 Our operator now works in disconnected mode. PM: Adel Zaalouk
Gallen 55 Hardware Accelerator Special Resource Operator (SRO) • Orchestrator to manages the deployment of software stacks for hardware accelerators • SRO uses recipes to enable the out-of-tree driver and manage the driver life cycle • Day 2 operations: ◦ Building and loading a kernel module ◦ Deploying the driver ◦ Deploying one device plugin ◦ Monitoring stack • Red Hat third-party support and certification policies. • Tech Preview in OpenShift 4.9 Hardware Accelerator Driver Toolkit (DTK) • The Driver Toolkit is a container image to be used as a base image for driver containers. • The DTK contains tools and the kernel packages required to build or install kernel modules • Usable for partner builds or local builds • Reduce cluster entitlement requirements • Tech Preview in OpenShift 4.9 Hardware Accelerator OpenShift worker node (RHEL CoreOS) Driver Daemonset CRI-O Plugin Daemonset Device Plugin Daemonset Feature Discovery Daemonset Node Exporter Daemonset Special Resource Operator (SRO) kubelet Node Feature Discovery (NFD) CRI-O Driver Toolkit
Developer Preview of Arm is here ◦ AWS only for now ◦ Get your customers to try it out - feedback is needed • IBM Power and IBM Z Features ◦ Enhancement for developers using odo with support for Service Binding Operator ◦ Now supported to use Helm to deploy applications ◦ Multiple Network Interfaces - support “backdated” to 4.6 ◦ OpenShift Pipelines 1.6 • New for IBM Z only ◦ Support for zVM 7.1 • New for IBM Power only ◦ Support for Power 10 56
handles proxy settings in the pods for managed Operands • Helper functions for reading proxy info so it can be passed down to Operands pods. • Easier to build “proxy-aware” Operators for proxied cluster environment. Uses UBI and other downstream images by default • Base image (v4.y) is guaranteed with compatibility fixes in two OCP releases (4.y and 4.y+1). • Easier create and maintain Operator projects in a Red Hat supported way. Bundle validate: WARN on k8s removed APIs • Easily see and be aware of those removed k8s APIs in the bundled manifests. • Get handy guidance on how to migrate per k8s upstream doc. PM: Tony Wu
Kubernetes/OCP-version specific operator catalogs and automatically switch during cluster updates, e.g. "quay.io/org/catalog:v{kube_major_version}.{kube_minor_version}" Support for “large” operator bundles Bundles with lots of metadata (for example large CRD manifests) are now compressed to stay below the 1MB etcd limit OpenShift Operator release compatibility OpenShift release compatibility can be denoted via operator metadata, initially blocks cluster upgrades metadata: annotations: operators.coreos.com/maxOpenShiftVersion: "4.8" Reduced resource usage / Better troubleshooting OLM catalog pods now use significantly less RAM. More status information in OperatorGroup and Subscription API, covering most install and update error scenarios. PM: Daniel Messer Operator Lifecycle Management Enhancements
for disconnected OpenShift cluster installations PM: Daniel Messer ▸ We prefer customers to run Quay on top of OCP ▸ But: disconnected clusters need a registry to store OCP release images and Operators before OCP can be installed ▸ Solution: tailored version of Quay helping customers to get a registry up and running quickly, mirroring is carried out via oc ▸ Local all-in-one Quay instance on RHEL 8 ▸ Released as part of OpenShift, post 4.9 GA, included in every OCP subscription 61
deployments PM: Daniel Messer Numerous stability and configuration enhancements: ▸ All deployed components are now auto-scaled and have at minimum two replicas and anti-affinity rules ▸ Separate status updates for unhealthy component of a Quay deployment, separate health checks ▸ Support for TLS-encrypted connections to external databases ▸ Direct updates from Quay 3.3 to 3.6 are supported ▸ OpenShift-based TLS certificate management for Routes 62
and organization of registry content PM: Daniel Messer ▸ Audience: Quay user / OpenShift administrator ▸ Use Cases: ･ Mirror content of multiple upstream registries into a single Quay* organization ･ Organize images into “subfolders” inside a single Quay organization ▸ Benefit: Eases skopeo mass mirroring, OpenShift Operator catalog mirroring ▸ Caveat: no hierarchical permission management Regular container image reference: quay.local/organization/repository:tag Nested container image references: quay.local/organization/collection/repository:tag quay.local/organization/folder/v1/repository:tag quay.local/ocp/v4/redhat-pipelines/operator:v4.9 quay.local/ocp/v4/redhat-pipelines/tekton:v4.9 * available in Quay 3.6 past OCP 4.9 GA, quay.io will get this towards the end of 2021 63
CSI PM: Gregory Charot • CSI Operators - plugable, built-in upgrade, could include new functionality ◦ Azure Stack Hub (GA) ◦ AWS EBS (GA) ◦ AWS EFS (Tech Preview) ◦ vSphere enhancements (Tech Preview) • CSI Migration - allow easy move from using existing intree drivers to new CSI drivers ◦ GCE Disk (Tech Preview) ◦ Azure Disk (Tech Preview) • Prepare for vSphere CSI transition ◦ CSI Driver will be the only option in 4.11 ◦ New CSI Driver requires hardware version 15 ▪ And version 6.7u3 and later ◦ Get your customers ready to upgrade ◦ h/w version 15 will be default starting 4.9 (but h/w 13 is still supported) CSI Operators Operator target Migration Driver OpenStack Cinder Tech Preview Tech Preview AWS EBS Tech Preview GA AWS EFS n/a Tech Preview GCE Disk Tech Preview GA Azure Disk Tech Preview Tech Preview Azure Stack Hub n/a GA vSphere - Tech preview 65
Storage to OpenShift Data Foundation • Disaster Recovery & Security ◦ Regional DR with ACM (Tech Preview) ◦ PV encryption with a service account • Extended Control Plane with IBM Flashsystem • Multicloud Object Gateway - Namespace replication • Managed service on ROSA - Early trial Out of the box support Block, File, Object Platforms AWS/Azure Google Cloud (Tech Preview) ARO - Self managed OCS IBM ROKS & Satellite - Managed OCS (GA) RHV OSP (Tech Preview) Bare metal/IBM Z/Power VMWare Thin/Thick IPI/UPI Deployment modes Disconnected environment and Proxied environments PM:Eran Tamir OpenShift Data Foundation updates
Innovations Red Hat Advanced Cluster Security 1 Enhanced protections for the Kubernetes API server allow teams to detect and alert on actions against their organizations most sensitive secrets and configmaps. 3 Shorten feedback loops by allowing teams to target workflows for security alert distribution with namespace annotations. 2 6 Accelerates security use case adoption in the cloud with certified testing and support for ROSA and ARO 5 Help organizations improve cybersecurity gap analysis and incident response prioritization by aligning security policies & alerts with the MITRE ATT&CK Framework Platform Support Advanced Security Uses Self-service workflows 4 Enable self-service security among application delivery organizations at scale with scoped access control annotations and labels Enhance OpenShift security with DeploymentConfig configuration checks for CI security testing 68
What’s new in RHACM 2.4 69 Product Managers: Jeff Brent, Scott Berens, Bradd Weidenbenner, Christian Stark, Sho Weimer Better Together • RH Advanced Cluster Security, formerly known as Stackrox, deployed using ACM policy • Support for OpenShift GitOps (ArgoCD) ApplicationSet • Drive notifications from GRC Compliance into AlertManager and other incident management tools • Observe cluster health metrics for non-OCP: EKS, GKE, AKS, IKS • Service Level Objectives (SLO) can be defined on the Grafana dashboard Red Hat Advanced Cluster Management brings together the portfolio including Ansible, OpenShift GitOps, Red Hat Advanced Cluster Security, across cloud vendors and all from a single pane of glass.
Scott Berens, Bradd Weidenbenner, Christian Stark, Sho Weimer Manage OpenShift Everywhere • Cluster Lifecycle Support Enhancements - Microsoft Azure Government • FIPS-Compliance • RHACM Hub deployed on IBM Power and Z • Centralized Infrastructure Management (CIM) for Bare Metal Deployments - Tech Preview • Advanced image registry configurations for managed clusters in public clouds Meeting the needs of customers in Public Sector (NAPS), whether on premise with FIPS compliance or in the cloud with Microsoft Azure Gov. Advanced Cluster Management for Kubernetes What’s new in RHACM 2.4
Scott Berens, Bradd Weidenbenner, Christian Stark, Sho Weimer Manage At the Edge • 1K Management scale with IPv6 Dual Stack support • Zero Touch Provisioning - Tech Preview • Single Node OpenShift management (SNO) • Hub-side Policy Templating • PolicyGenerator simplifies distribution of Kubernetes resource objects to managed clusters • Policy UX enhancements for fleet compliance At Red Hat, we see edge computing as an opportunity to extend the open hybrid cloud all the way to the data sources and end users. Edge is a strategy to deliver insights and experiences at the moment they’re needed. Advanced Cluster Management for Kubernetes What’s new in RHACM 2.4
Berens, Bradd Weidenbenner, Christian Stark, Sho Weimer Business Continuity • RHACM Hub backup and restore - Tech Preview • Leverage ODF (aka OCS) and RHACM for stateful workloads - Tech Preview • Persistent Volumes replication using VolSync (Scribe) - Tech Preview Red Hat customers expect a centralized management platform without the need for additional tooling to support their disaster recovery scenarios. Data Center 2 ACM-Hub ManagedCluster 2 PASSIVE NAMESPACE PVs RESOURCES RESOURCES RESOURCES PVs PVs • ODF/Volsync - Asynchronous Data Replication Data Center 1 ManagedCluster 1 NAMESPACE ACTIVE PVs RESOURCES RESOURCES RESOURCES PVs PVs Region 1 Region 2 • Operator for Backup & Restore of Hub ACM-Hub backup S3 • Restore and reattachment of new Hub Advanced Cluster Management for Kubernetes What’s new in RHACM 2.4
distribution on memory • It is now possible to select CPU or Memory Column management • You can hide/show columns for infrastructure and supplementary (to increase clarity) CSV exports with labels • Labels are now included in CSV exports Source pause • Now you can pause a source to avoid receiving error messages if the source is not updating Overall service reliability improvements PM: Sergio Ocón-Cárdenas 73
graduation PM: Franck Baudin apiVersion: performance.openshift.io/v2 kind: PerformanceProfile spec: [...] numa: topologyPolicy: "single-numa-node" Problem statement • High performance applications (typically running DPDK) requires to have CPUs, Devices/NICs and memory on the same NUMA node • Until OpenShift 4.9, Topology manager solely aligns CPUs and Devices/NICs Solution: Enhance the Kubelet Memory manager to permit Topology manager to align also memory and huge pages on the same NUMA node as the CPUs and Devices/NICs • Applies to Guaranteed QoS class containers only • Only enforced for single-numa-node and restricted topology manager policies CPU Socket0 RAM CPU Socket1 RAM PCI PCI
(CSR) GMC - Grandmaster Clock BC - Boundary Clock OC - Ordinary Clock (GMC) NIC RU RU RU 76 • OpenShift Node as a Boundary Clock • [O-RAN Approved] Low-latency, Node-local Event Bus w/ Ordinary Clock PTP Events and sidecar image for easy CNF (DU) consumption Red Hat OpenShift / Red Hat CoreOS DU Workload RH Provided Event Bus Sidecar Red Hat PTP SW Stack (PTP Operator, ptp4l, phys2sys, …) PTP Events AMQ Interconnect (Event Bus) PTP Events System Clock OpenShift Node as an Ordinary Clock currently supported in OpenShift 4.8 Far Edge Hardware Platform PMs: Robert Love, Michal Zaspera, Franck Baudin
PM: Shannon Wilber Enhanced capabilities to improve working with the OpenShift Console Monitoring Experience: • Support for Prometheus 2.29.2 and Thanos 0.22.0 • Enhancements to Alert Manager Rules, Cluster Monitoring Operator and refined triggering conditions ◦ Additional options to set Alerts on Cube States ◦ Improvements to detect more quickly when disk space is running low ◦ Expanded Alert rules for Kube Client errors with Thanos queries • Monitoring for User-Defined Projects • Remote write storage for Prometheus Metrics New Kube State Metrics & Alertmanager Functionality 78 Note: You can now disable the default Grafana dashboard deployment using a configuration option. https://github.com/openshift/cluster-monitoring-operator /pull/1241
OpenShift Available with OpenShift Logging 5.2.x • (Preview) New Added Logging Support and Flexibility (Fluentd to Vector): ◦ Option to replace fluentd with Vector as the primary collector for Logging: ▪ Provides a smooth upgrade path from the previous GA version of fluentd to Vector ▪ New Compatible API designed to extend requests to the Vector collector for expanded functionality • Support for assembling multi-line stacktrace log messages ◦ Now have the ability to assemble log messages part of a stacktrace to store it as one single log record instead of multiple. ◦ Users can log stacktraces as JSON • More flexibility to provide simple log exploration. ◦ New API experience inside the OpenShift console ◦ Ability to display contextualized logs inside an individual alert details page • New Support and Capabilities for Loki: ◦ Loki operator that is capable of providing an on-cluster solutions ▪ Ability to install, update, and manage a cluster with an alternative, scalable and performing log store 79 PM: Shannon Wilber
OpenShift Insights Advisor for OpenShift PM: Radek Vokál • Insights Advisor continues delivering proactive support to all connected users • Air-gapped Insights documented • Immediate email notifications for new Insights Events configurable on console.redhat.com • New recommendations focused on storage and network configuration, misconfiguration in user management • Smaller insights footprint with conditional data gathering 80 https://console.redhat.com/openshift https://console.redhat.com/settings/notifications/openshift
leading provider of enterprise open source software solutions. Award-winning support, training, and consulting services make Red Hat a trusted adviser to the Fortune 500. Thank you
redhatlivestreaming
cmusudakeisuke
estie
isaoshimizu
shigashiyama
sansantech
ken5scal
tubone24
yoshidashingo
linyows
4su_para
oracle4engineer
soudai
geoffreycrofte
morganepeng
colly
orderedlist
jnunemaker
keathley
irinanazarova
moore
cassininazir
chriscoyier
thoeni
