R10.0 Release Notes

About this task

StarlingX is a fully integrated edge cloud software stack that provides everything needed to deploy an edge cloud on one, two, or up to 100 servers.

This section describes the new capabilities, Known Limitations and Workarounds, Defects fixed and deprecated information in StarlingX 9.0 Release.

ISO image

The pre-built ISO (Debian) for StarlingX Release 9.0 is located at the StarlingX mirror repo:

https://mirror.starlingx.windriver.com/mirror/starlingx/release/9.0.0/debian/monolithic/outputs/iso/

Source Code for StarlingX Release 9.0

The source code for StarlingX Release 9.0 is available on the r/stx.9.0 branch in the StarlingX repositories.

Deployment

To deploy StarlingX Release 9.0, see Consuming StarlingX.

For detailed installation instructions, see StarlingX 9.0 Installation Guides.

New Features and Enhancements

The sections below provide a detailed list of new features and links to the associated user guides (if applicable).

Kubernetes up-version

In StarlingX 9.0, the Kubernetes version that is supported is in the range of v1.24 to v1.27.

Platform Application Components Revision

The following applications have been updated to a new version in StarlingX Release 9.0. All platform application up-versions are updated to remain current and address security vulnerabilities in older versions.

  • app-sriov-fec-operator: 2.7.1

  • cert-manager: 1.11.1

  • metric-server: 1.0.18

  • nginx-ingress-controller: 1.9.3

  • oidc-dex: 2.37.0

  • vault: 1.14.8

  • portieris: 0.13.10

  • istio: 1.19.4

  • kiali: 1.75.0

FluxCD Maintenance

FluxCD helm-controller is upgraded from v0.27.0 to v0.35.0 and is compatible with Helm version up to v3.12.1 and Kubernetes v1.27.3.

FluxCD source-controller is upgraded from v0.32.1 to v1.0.1 and is compatible with Helm version up to v3.12.1 and Kubernetes v1.27.3.

Helm Maintenance

Helm has been upgraded to v3.12.2 in StarlingX Release 9.0.

Support for Silicom TimeSync Server Adaptor

The Silicom network adaptor provides local time sync support via a local GNSS module which is based on the Intel Columbiaville device.

  • cvl-4.10 Silicom driver bundle
    • ice driver: 1.10.1.2

    • i40e driver: 2.21.12

    • iavf driver: 4.6.1

    Note

    cvl-4.10 is only recommended if the Silicom STS2 card is used.

Kubernetes Upgrade Optimization - AIO-Simplex

Configure Kubernetes Multi-Version Upgrade Cloud Orchestration for AIO-SX

You can configure Kubernetes multi-version upgrade orchestration strategy using the sw-manager command. This feature is enabled from StarlingX release 8.0 and is supported only for the AIO-SX system.

See: Configure Kubernetes Multi-Version Upgrade Cloud Orchestration for AIO-SX

Manual Kubernetes Multi-Version Upgrade in AIO-SX

AIO-SX now supports multi-version Kubernetes upgrades. In this model, Kubernetes is upgraded by two or more versions after disabling applications and then applications are enabled again. This is faster than upgrading Kubernetes one version at a time. Also, the upgrade can be aborted and reverted to the original version. This feature is supported only for AIO-SX.

See: Manual Kubernetes Multi-Version Upgrade in AIO-SX

Platform Admin Network Introduction

The newly introduced admin network is an optional network that is used to monitor and control internal StarlingX between the subclouds and system controllers in a Distributed Cloud environment. This function is performed by the management network in the absence of an admin network. However, the admin network is more easily reconfigured to handle subnet and IP address network parameter changes after initial configuration.

In deployment configurations, static routes from the management or admin interface of subclouds controller nodes to the system controller’s management subnet must be present. This ensures that the subcloud comes online after deployment.

Note

The admin network is optional. The default management network will be used if it is not present.

You can manage an optional admin network on a subcloud for IP connectivity to the system controller management network where the IP addresses of the admin network can be changed.

See:

L3 Firewalls for all StarlingX Platform Interfaces

StarlingX incorporates default firewall rules for the platform networks (OAM, management, cluster-host, pxeboot, admin, and storage). You can configure additional Kubernetes Network Policies to augment or override the default rules.

See:

app-sriov-fec-operator upgrade to FEC operator 2.7.1

A new version of the FEC Operator v2.7.1 (for all Intel hardware accelerators) is supported to include igb_uio along with making the accelerator resource names configurable and enabling accelerator device configuration using igb_uio driver when secure boot is enabled in the BIOS.

Note

FEC operator is now running on the StarlingX platform core.

See: Configure Intel Wireless FEC Accelerators using SR-IOV FEC operator

Redundant System Clock Synchronization

The phc2sys application can be configured to accept multiple source clock inputs. The quality of these sources are compared to user-defined priority values and the best available source is selected to set the system time.

The quality of the configured sources is continuously monitored by phc2sys application and will select a new best source if the current source degrades or if another source becomes higher quality.

See: Redundant System Clock Synchronization.

Configure Intel E810 NICs using Intel Ethernet Operator

You can install and use Intel Ethernet operator to orchestrate and manage the configuration and capabilities provided by Intel E810 Series network interface cards (NICs).

See: Configure Intel E810 NICs using Intel Ethernet Operator.

AppArmor Support

AppArmor is a Mandatory Access Control (MAC) system built on Linux’s LSM (Linux Security Modules) interface. In practice, the kernel queries AppArmor before each system call to know whether the process is authorized to do the given operation. Through this mechanism, AppArmor confines programs to a limited set of resources.

AppArmor helps administrators in running a more secure kubernetes deployment by restricting what operations containers/pods are allowed, and/or provide better auditing through system logs. The access needed by a container/pod is configured through profiles tuned to allow access such as Linux capabilities, network access, file permissions, etc.

See: About AppArmor.

Support for Vault

This release re-introduces support for Vault as it was intermittently unavailable in StarlingX. The supported version vault: 1.14.8 or later / vault-k8s: 1.2.1 / helm-chart: 0.25.0 after the helm-v3 up-version to 3.6+

StarlingX integrates open source Vault containerized security application (Optional) into the StarlingX solution, that requires PVCs as a storage backend to be enabled.

See: Vault Overview.

Support for Portieris

StarlingX now supports version 0.13.10. Portieris is an open source Kubernetes admission controller which ensures only policy-compliant images, such as signed images from trusted registries, can run. The Portieris application uses images from the icr.io registry. You must configure service parameters for the icr.io registry prior to applying the Portieris application, see: About Changing External Registries for StarlingX Installation. For Distributed Cloud deployments, the images must be present on the System Controller registry.

See: Portieris Overview.

Configurable Power Manager

Configurable Power Manager focuses on containerized applications that use power profiles individually by the core and/or the application.

StarlingX has the capability to regulate the frequency of the entire processor. However, this control is primarily directed towards the classification of the core, distinguishing between application and platform cores. Consequently, if a user requires to control over an individual core, such as Core 10 in a 24-core CPU, adjustments must be applied to all cores collectively. In the context of containerized operations, it becomes imperative to establish personalized configurations. This entails assigning each container the requisite power configuration. In essence, this involves providing specific and individualized power configurations to each core or group of cores.

See: Configurable Power Manager.

Technology Preview - Install Power Metrics Application

The Power Metrics app deploys two containers, cAdvisor and Telegraf that collect metrics about hardware usage.

See: Install Power Metrics Application.

Install Node Feature Discovery (NFD) StarlingX Application

Node Feature Discovery (NFD) version 0.15.0 detects hardware features available on each node in a kubernetes cluster and advertises those features using Kubernetes node labels. This procedure walks you through the process of installing the NFD StarlingX Application.

See: Install Node Feature Discovery Application.

Partial Disk (Transparent) Encryption Support via Software Encryption (LUKS)

A new encrypted filesystem using Linux Unified Key Setup (LUKS) is created automatically on all hosts to store security-sensitive files. This is mounted at ‘/var/luks/stx/luks_fs’ and the files kept in ‘/var/luks/stx/luks_fs/controller’ directory are replicated between controllers.

K8s API/CLI OIDC (Dex) Authentication with Local LDAP Backend

StarlingX offers LDAP commands to create and manage LDAP Linux groups as part of a StarlingX local LDAP server (serving the local StarlingX cluster and, in the case of Distributed Cloud, the entire Distribute Cloud System).

StarlingX provides procedures to configure the oidc-auth-apps OIDC Identity Provider (Dex) system application to use the StarlingX local LDAP server (in addition to, or in place of the already supported remote Windows Active Directory) to authenticate users of the Kubernetes API.

See:

Create LDAP Linux Groups

StarlingX offers LDAP commands to create and manage LDAP Linux groups as part of the ldapscripts library.

StarlingX OpenStack now supports Antelope

Currently stx-openstack has been updated and now deploys OpenStack services based on the Antelope release.

Pod Security Policy

PSP ONLY applies if running on Kubernetes v1.24 or earlier. PSP is deprecated as of Kubernetes v1.21 and is removed in Kubernetes v1.25. Instead of using PSP, you can enforce similar restrictions on Pods using Pod Security Admission Controller.

Since it has been introduced PSP has had usability problems. The way PSPs are applied to pods has proven confusing especially when trying to use them. It is easy to accidentally grant broader permissions than intended, and difficult to inspect which PSPs apply in a certain situation. Kubernetes offers a built-in PSA controller that will replace PSPs in the future.

WAD users sudo and local linux group assignment

StarlingX 9.0 supports and provides procedures for centrally configured Window Active Directory (WAD) Users with sudo access and local linux group assignments; i.e. with only WAD configuration changes.

See:

Subcloud Error Root Cause Correction Action

This feature provides a root cause analysis of the subcloud deployment / upgrade failure. This includes:

  • existing ‘deploy_status’ that provides progress through phases of subcloud deployment and, on error, the phase that failed

  • introduces deploy_error_desc attribute that provides a summary of the key deployment/upgrade errors

  • Additional text that is added at the end of the ‘deploy_error_desc’ error message, with information on:

    • trouble shooting commands

    • root cause of the errors and

    • suggested recovery action

See: Manage Subclouds Using the CLI

Patch Orchestration Phase Operations

The distributed cloud patch orchestration has the option to separate the upload from the apply, remove, install and reboot operations. This facilitates performing the upload operations outside of the system maintenance window to reduce the total execution time during the patch activation that occurs during the maintenance window. With the separation of operations, systems can be prestaged with the updates prior to applying the changes to the system.

See: Distributed Cloud Guide

Long Latency Between System Controller and Subclouds

Rehoming procedure of a subcloud that has been powered off for a long period of time will differ from the regular rehoming procedure. Based on how long the subcloud has been offline, the platform certificates will expire and will need to be regenerated.

See: Rehoming Subcloud with Expired Certificates

GEO Redundancy

StarlingX may be deployed across a geographically distributed set of regions. A region consists of a local Kubernetes cluster with local redundancy and access to high-bandwidth, low-latency networking between hosts within that region.

StarlingX Distributed Cloud GEO redundancy configuration supports the ability to recover from a catastrophic event that requires subclouds to be rehomed away from the failed system controller site to the available site(s) which have enough spare capacity. This way, even if the failed site cannot be restored in short time, the subclouds can still be rehomed to available peer system controller(s) for centralized management.

In this release, the following items are addressed:

  • 1+1 GEO redundancy

    • Active-Active redundancy model

    • Total number of subclouds should not exceed 1K

  • Automated operations

    • Synchronization and liveness check between peer systems

    • Alarm generation if peer system controller is down

  • Manual operations

    • Batch rehoming from alive peer system controller

See: GEO Redundancy

Redfish Virtual Media Robustness

Redfish virtual media operations has been observed to frequently fail with transient errors. While the conditions for those failures are not always known (network, BMC timeouts, etc), it has been observed that if the Subcloud install operation is retried, the operation is successful.

To alleviate the transient conditions, the robustness of the Redfish virtual media controller (RVMC) is improved by introducing additional error handling and retry attempts.

See: Install a Subcloud Using Redfish Platform Management Service

Hardware Updates

See:

Bug status

Fixed bugs

This release provides fixes for a number of defects. Refer to the StarlingX bug database to review the R9.0 Fixed Bugs.

Known Limitations and Workarounds

The following are known limitations you may encounter with your StarlingX Release 9.0 and earlier releases. Workarounds are suggested where applicable.

Note

These limitations are considered temporary and will likely be resolved in a future release.

Suspend/Resume on VMs with SR-IOV (direct) Ports

When using VMs with SR-IOV ports created with the -vnic-type=direct option after a Suspend action, if one wants to Resume the instance it might come up with all virtual NICs created but missing the IP Address of the vNIC connected to the SR-IOV port.

Workaround: Manually Power-Off and Power-On (or Hard-Reboot) the instance and the IP should be assigned correctly again (no information is lost).

Error on Restoring OpenStack after Backup

The ansible command for restoring the app will fail with StarlingX Release 9.0 with an error message mentioning the absence of an Armada directory.

Workaround: Manually change the backup tarball adding the Armada directory using the following the steps:

tar -xzf wr_openstack_backup_file.tgz  # this will create a opt directory
cp -r  opt/platform/fluxcd/ opt/platform/armada # copy fluxd to armada
tar -czf new_wr-openstack_backu.tgz opt/  # tar the opt directory into a new backup tarball

Subcloud Upgrade with Kubernetes Versions

Subcloud Kubernetes versions are upgraded along with the System Controller. You can add a new subcloud while the System Controller is on intermediate versions of Kubernetes as long as the needed k8s images are available at the configured sources.

Workaround: In a Distributed Cloud configuration, when upgrading from StarlingX Release 7.0 the Kubernetes version is v1.23.1. The default version of the new install for Kubernetes is v1.24.4. Kubernetes must be upgraded one version at a time on the System Controller.

Note

New subclouds should not be added until the System Controller has been upgraded to Kubernetes v1.24.4.

AIO-SX Restore Fails during puppet-manifest-apply.sh

Restore fails using a backup file created after a fresh install.

Workaround: During the restore process, after reinstalling the controllers, the OAM interface must be configured with the same IP address protocol version used during installation.

Subcloud Controller-0 is in a degraded state after upgrade and host unlock

During an upgrade orchestration of the subcloud from StarlingX Release 7.0 to StarlingX Release 8.0, and after host unlock, the subcloud is in a degraded state, and alarm 200.004 is raised, displaying “controller-0 experienced a service-affecting failure. Auto-recovery in progress”.

Workaround: You can recover the subcloud to the available state by locking and unlocking controller-0 .

Limitations when using Multiple Driver Versions for the same NIC Family

The capability to support multiple NIC driver versions has the following limitations:

  • Intel NIC family supports only: ice, i40e and iavf drivers

  • Driver versions must respect the compatibility matrix between drivers

  • Multiple driver versions cannot be loaded simultaneously and applies to the entire system.

  • Latest driver version will be loaded by default, unless specifically configured to use a legacy driver version.

  • Drivers used by the installer will always use the latest version, therefore firmware compatibility must support basic NIC operations for each version to facilitate installation

  • Host reboot is required to activate the configured driver versions

  • For Backup and Restore, the host must be rebooted a second time for in order to activate the drivers versions.

Workaround: NA

Quartzville Tools

The following celo64e and nvmupdate64e commands are not supported in StarlingX, Release 8.0 due to a known issue in Quartzville tools that crashes the host.

Workaround: Reboot the host using the boot screen menu.

Controller SWACT unavailable after System Restore

After performing a restore of the system, the user is unable to swact the controller.

Workaround: NA

Intermittent Kubernetes Upgrade failure due to missing Images

During a Kubernetes upgrade, the upgrade may intermittently fail when you run system kube-host-upgrade <host> control-plane due to the containerd cache being cleared.

Workaround: If the above failure is encountered, run the following commands on the host encountering the failure:

Procedure

  1. Ensure the failure is due to missing images by running crictl images and confirming the following are not present:

    registry.local:9001/k8s.gcr.io/kube-apiserver:v1.24.4
    registry.local:9001/k8s.gcr.io/kube-controller-manager:v1.24.4
    registry.local:9001/k8s.gcr.io/kube-scheduler:v1.24.4
    registry.local:9001/k8s.gcr.io/kube-proxy:v1.24.4
    
  2. Manually pull the image into containerd cache by running the following commands, replacing <admin_password> with your password for the admin user.

    ~(keystone_admin)]$ crictl pull --creds admin:<admin_password> registry.local:9001/k8s.gcr.io/kube-apiserver:v1.24.4
    ~(keystone_admin)]$ crictl pull --creds admin:<admin_password> registry.local:9001/k8s.gcr.io/kube-controller-manager:v1.24.4
    ~(keystone_admin)]$ crictl pull --creds admin:<admin_password> registry.local:9001/k8s.gcr.io/kube-scheduler:v1.24.4
    ~(keystone_admin)]$ crictl pull --creds admin:<admin_password> registry.local:9001/k8s.gcr.io/kube-proxy:v1.24.4
    
  3. Ensure the images are present when running crictl images. Rerun the system kube-host-upgrade <host> control-plane` command.

Docker Network Bridge Not Supported

The Docker Network Bridge, previously created by default, is removed and no longer supported in StarlingX Release 8.0 as the default bridge IP address collides with addresses already in use.

As a result, docker can no longer be used for running containers. This impacts building docker images directly on the host.

Workaround: Create a Kubernetes pod that has network access, log in to the container, and build the docker images.

Impact of Kubenetes Upgrade to v1.24

In Kubernetes v1.24 support for the RemoveSelfLink feature gate was removed. In previous releases of StarlingX this has been set to “false” for backward compatibility, but this is no longer an option and it is now hardcoded to “true”.

Workaround: Any application that relies on this feature gate being disabled (i.e. assumes the existence of the “self link”) must be updated before upgrading to Kubernetes v1.24.

Password Expiry Warning Message is not shown for LDAP user on login

In StarlingX Release 8.0, the password expiry warning message is not shown for LDAP users on login when the password is nearing expiry. This is due to the pam-sssd integration.

Workaround: It is highly recommend that LDAP users maintain independent notifications and update their passwords every 3 months.

The expired password can be reset by a user with root privileges using the following command:

Console Session Issues during Installation

After bootstrap and before unlocking the controller, if the console session times out (or the user logs out), systemd does not work properly. fm, sysinv and mtcAgent do not initialize.

Workaround: If the console times out or the user logs out between bootstrap and unlock of controller-0, then, to recover from this issue, you must re-install the ISO.

PTP O-RAN Spec Compliant Timing API Notification

  • The ptp-notification <minor_version>.tgz application tarball and the corresponding notificationservice-base:stx8.0-v2.0.2 image are not backwards compatible with applications using the v1 ptp-notification API and the corresponding notificationclient-base:stx.8.0-v2.0.2 image.

    Backward compatibility will be provided in StarlingX Release 9.0.

    Note

    For O-RAN Notification support (v2 API), deploy and use the ptp-notification-<minor_version>.tgz application tarball. Instructions for this can be found in the StarlingX Release 8.0 documentation.

    See:

  • The v1 API only supports monitoring a single ptp4l + phc2sys instance.

    Workaround: Ensure the system is not configured with multiple instances when using the v1 API.

  • The O-RAN Cloud Notification defines a /././sync API v2 endpoint intended to allow a client to subscribe to all notifications from a node. This endpoint is not supported StarlingX Release 8.0.

    Workaround: A specific subscription for each resource type must be created instead.

  • v1 / v2

    • v1: Support for monitoring a single ptp4l instance per host - no other services can be queried/subscribed to.

    • v2: The API conforms to O-RAN.WG6.O-Cloud Notification API-v02.01 with the following exceptions, that are not supported in StarlingX Release 8.0.

      • O-RAN SyncE Lock-Status-Extended notifications

      • O-RAN SyncE Clock Quality Change notifications

      • O-RAN Custom cluster names

      • /././sync endpoint

    Workaround: See the respective PTP-notification v1 and v2 document subsections for further details.

    v1: https://docs.starlingx.io/api-ref/ptp-notification-armada-app/api_ptp_notifications_definition_v1.html

    v2: https://docs.starlingx.io/api-ref/ptp-notification-armada-app/api_ptp_notifications_definition_v2.html

Upper case characters in host names cause issues with kubernetes labelling

Upper case characters in host names cause issues with kubernetes labelling.

Workaround: Host names should be lower case.

Debian Bootstrap

On CentOS bootstrap worked even if dns_servers were not present in the localhost.yml. This does not work for Debian bootstrap.

Workaround: You need to configure the dns_servers parameter in the localhost.yml, as long as no FQDNs were used in the bootstrap overrides in the localhost.yml file for Debian bootstrap.

Installing a Debian ISO

The disks and disk partitions need to be wiped before the install. Installing a Debian ISO may fail with a message that the system is in emergency mode if the disks and disk partitions are not completely wiped before the install, especially if the server was previously running a CentOS ISO.

Workaround: When installing a lab for any Debian install, the disks must first be completely wiped using the following procedure before starting an install.

Use the following wipedisk commands to run before any Debian install for each disk (eg: sda, sdb, etc):

sudo wipedisk
# Show
sudo sgdisk -p /dev/sda
# Clear part table
sudo sgdisk -o /dev/sda

Note

The above commands must be run before any Debian install. The above commands must also be run if the same lab is used for CentOS installs after the lab was previously running a Debian ISO.

Security Audit Logging for K8s API

A custom policy file can only be created at bootstrap in apiserver_extra_volumes. If a custom policy file was configured at bootstrap, then after bootstrap the user has the option to configure the parameter audit-policy-file to either this custom policy file (/etc/kubernetes/my-audit-policy-file.yml) or the default policy file /etc/kubernetes/default-audit-policy.yaml. If no custom policy file was configured at bootstrap, then the user can only configure the parameter audit-policy-file to the default policy file.

Only the parameter audit-policy-file is configurable after bootstrap, so the other parameters (audit-log-path, audit-log-maxsize, audit-log-maxage and audit-log-maxbackup) cannot be changed at runtime.

Workaround: NA

See: Kubernetes Operator Command Logging.

Installing subcloud with patches in Partial-Apply is not supported

When a patch has been uploaded and applied, but not installed, it is in a Partial-Apply state. If a remote subcloud is installed via Redfish (miniboot) at this point, it will run the patched software. Any patches in this state will be applied on the subcloud as it is installed. However, this is not reflected in the output from the sw-patch query command on the subcloud.

Workaround: For remote subcloud install operations using the Redfish protocol, you should avoid installing any subclouds if there are System Controller patches in the Partial-Apply state.

PTP is not supported on Broadcom 57504 NIC

PTP is not supported on the Broadcom 57504 NIC.

Workaround: None. Do not configure PTP instances on the Broadcom 57504 NIC.

Metrics Server Update across Upgrades

After a platform upgrade, the Metrics Server will NOT be automatically updated.

Workaround: To update the Metrics Server, See: Install Metrics Server

Horizon Drop-Down lists in Chrome and Firefox causes issues due to the new branding

Drop-down menus in Horizon do not work due to the ‘select’ HTML element on Chrome and Firefox.

It is considered a ‘replaced element’ as it is generated by the browser and/or operating system. This element has a limited range of customizable CSS properties.

Workaround: The system should be 100% usable even with this limitation. Changing browser’s and/or operating system’s theme could solve display issues in case they limit the legibility of the elements (i.e. white text and white background).

Deploying an App using nginx controller fails with internal error after controller.name override

An Helm override of controller.name to the nginx-ingress-controller app may result in errors when creating ingress resources later on.

Example of Helm override:

Workaround: NA

Kata Container is not supported on StarlingX 8.0

Kata Containers that were supported on CentOS in earlier releases of StarlingX will not be supported on StarlingX Release 8.0.

Vault is not supported on StarlingX Release 8.0

The Vault application is not supported on StarlingX Release 8.0.

Workaround: NA

Portieris is not supported on StarlingX Release 8.0

The Portieris application is not supported on StarlingX Release 8.0.

Workaround: NA

DCManager Patch Orchestration

Warning

Patches must be applied or removed on the System Controller prior to using the dcmanager patch-strategy command to propagate changes to the subclouds.

Optimization with a Large number of OSDs

As Storage nodes are not optimized, you may need to optimize your Ceph configuration for balanced operation across deployments with a high number of OSDs. This results in an alarm being generated even if the installation succeeds.

800.001 - Storage Alarm Condition: HEALTH_WARN. Please check ‘ceph -s’

Workaround: To optimize your storage nodes with a large number of OSDs, it is recommended to use the following commands:

$ ceph osd pool set kube-rbd pg_num 256
$ ceph osd pool set kube-rbd pgp_num 256

PTP tx_timestamp_timeout causes ptp4l port to transition to FAULTY

NICs using the Intel Ice NIC driver may report the following in the ptp4l` logs, which might coincide with a PTP port switching to FAULTY before re-initializing.

ptp4l[80330.489]: timed out while polling for tx timestamp
ptp4l[80330.CGTS-30543489]: increasing tx_timestamp_timeout may correct this issue, but it is likely caused by a driver bug

This is due to a limitation of the Intel ICE driver.

Workaround: The recommended workaround is to set the tx_timestamp_timeout parameter to 700 (ms) in the ptp4l config using the following command.

~(keystone_admin)]$ system ptp-instance-parameter-add ptp-inst1 tx_timestamp_timeout=700

BPF is disabled

BPF cannot be used in the PREEMPT_RT/low latency kernel, due to the inherent incompatibility between PREEMPT_RT and BPF, see, https://lwn.net/Articles/802884/.

Some packages might be affected when PREEMPT_RT and BPF are used together. This includes the following, but not limited to these packages.

  • libpcap

  • libnet

  • dnsmasq

  • qemu

  • nmap-ncat

  • libv4l

  • elfutils

  • iptables

  • tcpdump

  • iproute

  • gdb

  • valgrind

  • kubernetes

  • cni

  • strace

  • mariadb

  • libvirt

  • dpdk

  • libteam

  • libseccomp

  • binutils

  • libbpf

  • dhcp

  • lldpd

  • containernetworking-plugins

  • golang

  • i40e

  • ice

Workaround: It is recommended not to use BPF with real time kernel. If required it can still be used, for example, debugging only.

crashkernel Value

crashkernel=auto is no longer supported by newer kernels, and hence the v5.10 kernel will not support the “auto” value.

Workaround: StarlingX uses crashkernel=2048m instead of crashkernel=auto.

Note

StarlingX Release 8.0 has increased the amount of reserved memory for the crash/kdump kernel from 512 MiB to 2048 MiB.

Control Group parameter

The control group (cgroup) parameter kmem.limit_in_bytes has been deprecated, and results in the following message in the kernel’s log buffer (dmesg) during boot-up and/or during the Ansible bootstrap procedure: “kmem.limit_in_bytes is deprecated and will be removed. Please report your use case to linux-mm@kvack.org if you depend on this functionality.” This parameter is used by a number of software packages in StarlingX, including, but not limited to, systemd, docker, containerd, libvirt etc.

Workaround: NA. This is only a warning message about the future deprecation of an interface.

Kubernetes Taint on Controllers for Standard Systems

In Standard systems, a Kubernetes taint is applied to controller nodes in order to prevent application pods from being scheduled on those nodes; since controllers in Standard systems are intended ONLY for platform services. If application pods MUST run on controllers, a Kubernetes toleration of the taint can be specified in the application’s pod specifications.

Workaround: Customer applications that need to run on controllers on Standard systems will need to be enabled/configured for Kubernetes toleration in order to ensure the applications continue working after an upgrade from StarlingX Release 6.0 to StarlingX future Releases. It is suggested to add the Kubernetes toleration to your application prior to upgrading to StarlingX Release 8.0.

You can specify toleration for a pod through the pod specification (PodSpec). For example:

spec:
....
template:
....
    spec
      tolerations:
        - key: "node-role.kubernetes.io/control-plane"
        operator: "Exists"
        effect: "NoSchedule"
        - key: "node-role.kubernetes.io/control-plane"
        operator: "Exists"
        effect: "NoSchedule"

See: Taints and Tolerations.

New Kubernetes Taint on Controllers for Standard Systems

A new Kubernetes taint will be applied to controllers for Standard systems in order to prevent application pods from being scheduled on controllers; since controllers in Standard systems are intended ONLY for platform services. If application pods MUST run on controllers, a Kubernetes toleration of the taint can be specified in the application’s pod specifications. You will also need to change the nodeSelector / nodeAffinity to use the new label.

Workaround: Customer applications that need to run on controllers on Standard systems will need to be enabled/configured for Kubernetes toleration in order to ensure the applications continue working after an upgrade to StarlingX Release 8.0 and StarlingX future Releases.

You can specify toleration for a pod through the pod specification (PodSpec). For example:

spec:
....
template:
....
    spec
      tolerations:
        - key: "node-role.kubernetes.io/control-plane"
        operator: "Exists"
        effect: "NoSchedule"

See: Taints and Tolerations.

Ceph alarm 800.001 interrupts the AIO-DX upgrade orchestration

Upgrade orchestration fails on AIO-DX systems that have Ceph enabled.

Workaround: Clear the Ceph alarm 800.001 by manually upgrading both controllers and using the following command:

~(keystone_admin)]$ ceph mon enable-msgr2

Ceph alarm 800.001 is cleared.

Storage Nodes are not considered part of the Kubernetes cluster

When running the system kube-host-upgrade-list command the output must only display controller and worker hosts that have control-plane and kubelet components. Storage nodes do not have any of those components and so are not considered a part of the Kubernetes cluster.

Workaround: Do not include Storage nodes.

Backup and Restore of ACC100 (Mount Bryce) configuration requires double unlock attempt

After restoring from a previous backup with an Intel ACC100 processing accelerator device, the first unlock attempt will be refused since this specific kind of device will be updated in the same context.

Workaround: A second attempt after few minutes will accept and unlock the host.

Application Pods with SRIOV Interfaces

Application Pods with SR-IOV Interfaces require a restart-on-reboot: “true” label in their pod spec template.

Pods with SR-IOV interfaces may fail to start after a platform restore or Simplex upgrade and persist in the Container Creating state due to missing PCI address information in the CNI configuration.

Workaround: Application pods that require|SRIOV| should add the label restart-on-reboot: “true” to their pod spec template metadata. All pods with this label will be deleted and recreated after system initialization, therefore all pods must be restartable and managed by a Kubernetes controller (i.e. DaemonSet, Deployment or StatefulSet) for auto recovery.

Pod Spec template example:

template:
    metadata:
      labels:
        tier: node
        app: sriovdp
        restart-on-reboot: "true"

Management VLAN Failure

If the Management VLAN fails on the active System Controller, communication failure 400.005 is detected, and alarm 280.001 is raised indicating subclouds are offline.

Workaround: System Controller will recover and subclouds are manageable when the Management VLAN is restored.

Host Unlock During Orchestration

If a host unlock during orchestration takes longer than 30 minutes to complete, a second reboot may occur. This is due to the delays, VIM tries to abort. The abort operation triggers the second reboot.

Workaround: NA

Storage Nodes Recovery on Power Outage

Storage nodes take 10-15 minutes longer to recover in the event of a full power outage.

Workaround: NA

Ceph OSD Recovery on an AIO-DX System

In certain instances a Ceph OSD may not recover on an AIO-DX system (for example, if an OSD comes up after a controller reboot and a swact occurs), and remains in the down state when viewed using the ceph -s command.

Workaround: Manual recovery of the OSD may be required.

Using Helm with Container-Backed Remote CLIs and Clients

If Helm is used within Container-backed Remote CLIs and Clients:

  • You will NOT see any helm installs from StarlingX Platform’s system FluxCD applications.

    Workaround: Do not directly use Helm to manage StarlingX Platform’s system FluxCD applications. Manage these applications using system application commands.

  • You will NOT see any helm installs from end user applications, installed using Helm on the controller’s local CLI.

    Workaround: It is recommended that you manage your Helm applications only remotely; the controller’s local CLI should only be used for management of the StarlingX Platform infrastructure.

Remote CLI Containers Limitation for StarlingX Platform HTTPS Systems

The python2 SSL lib has limitations with reference to how certificates are validated. If you are using Remote CLI containers, due to a limitation in the python2 SSL certificate validation, the certificate used for the ‘ssl’ certificate should either have:

  1. CN=IPADDRESS and SAN=empty or,

  2. CN=FQDN and SAN=FQDN

Workaround: Use CN=FQDN and SAN=FQDN as CN is a deprecated field in the certificate.

Cert-manager does not work with uppercase letters in IPv6 addresses

Cert-manager does not work with uppercase letters in IPv6 addresses.

Workaround: Replace the uppercase letters in IPv6 addresses with lowercase letters.

apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
    name: oidc-auth-apps-certificate
    namespace: test
spec:
    secretName: oidc-auth-apps-certificate
    dnsNames:
    - ahost.com
    ipAddresses:
    - fe80::903a:1c1a:e802::11e4
    issuerRef:
        name: cloudplatform-interca-issuer
        kind: Issuer

Kubernetes Root CA Certificates

Kubernetes does not properly support k8s_root_ca_cert and k8s_root_ca_key being an Intermediate CA.

Workaround: Accept internally generated k8s_root_ca_cert/key or customize only with a Root CA certificate and key.

Windows Active Directory

  • Limitation: The Kubernetes API does not support uppercase IPv6 addresses.

    Workaround: The issuer_url IPv6 address must be specified as lowercase.

  • Limitation: The refresh token does not work.

    Workaround: If the token expires, manually replace the ID token. For more information, see, Configure Kubernetes Client Access.

  • Limitation: TLS error logs are reported in the oidc-dex container on subclouds. These logs should not have any system impact.

    Workaround: NA

  • Limitation: stx-oidc-client liveness probe sometimes reports failures. These errors may not have system impact.

    Workaround: NA

BMC Password

The BMC password cannot be updated.

Workaround: In order to update the BMC password, de-provision the BMC, and then re-provision it again with the new password.

Application Fails After Host Lock/Unlock

In some situations, application may fail to apply after host lock/unlock due to previously evicted pods.

Workaround: Use the kubectl delete command to delete the evicted pods and reapply the application.

Application Apply Failure if Host Reset

If an application apply is in progress and a host is reset it will likely fail. A re-apply attempt may be required once the host recovers and the system is stable.

Workaround: Once the host recovers and the system is stable, a re-apply may be required.

Pod Recovery after a Host Reboot

On occasions some pods may remain in an unknown state after a host is rebooted.

Workaround: To recover these pods kill the pod. Also based on https://github.com/kubernetes/kubernetes/issues/68211 it is recommended that applications avoid using a subPath volume configuration.

Rare Node Not Ready Scenario

In rare cases, an instantaneous loss of communication with the active kube-apiserver may result in kubernetes reporting node(s) as stuck in the “Not Ready” state after communication has recovered and the node is otherwise healthy.

Workaround: A restart of the kublet process on the affected node(s) will resolve the issue.

Platform CPU Usage Alarms

Alarms may occur indicating platform cpu usage is >90% if a large number of pods are configured using liveness probes that run every second.

Workaround: To mitigate either reduce the frequency for the liveness probes or increase the number of platform cores.

Pods Using isolcpus

The isolcpus feature currently does not support allocation of thread siblings for cpu requests (i.e. physical thread +HT sibling).

Workaround: NA

system host-disk-wipe command

The system host-disk-wipe command is not supported in this release.

Workaround: NA

Restrictions on the Size of Persistent Volume Claims (PVCs)

There is a limitation on the size of Persistent Volume Claims (PVCs) that can be used for all StarlingX Platform Releases.

Workaround: It is recommended that all PVCs should be a minimum size of 1GB. For more information, see, https://bugs.launchpad.net/starlingx/+bug/1814595.

Sub-Numa Cluster Configuration not Supported on Skylake Servers

Sub-Numa cluster configuration is not supported on Skylake servers.

Workaround: For servers with Skylake Gold or Platinum CPUs, Sub-NUMA clustering must be disabled in the BIOS.

The ptp-notification-demo App is Not a System-Managed Application

The ptp-notification-demo app is provided for demonstration purposes only. Therefore, it is not supported on typical platform operations such as Backup and Restore.

Workaround: NA

Deleting image tags in registry.local may delete tags under the same name

When deleting image tags in the registry.local docker registry, you should be aware that the deletion of an <image-name:tag-name> will delete all tags under the specified <image-name> that have the same ‘digest’ as the specified <image-name:tag-name>. For more information, see, Delete Image Tags in the Docker Registry.

Workaround: NA

Deprecated Notices

Airship Armada is deprecated

Note

Airship Armada is removed in stx.9.0 and replaced with FluxCD. All Armada based applications have to be removed before you perform an upgrade from StarlingX Release 9.0 to StarlingX Release 10.0.

Note

Some application repositories may still have “armada” in the file path but are now supported by FluxCD. See https://opendev.org/starlingx/?sort=recentupdate&language=&q=armada.

StarlingX Release 7.0 introduces FluxCD based applications that utilize FluxCD Helm/source controller pods deployed in the flux-helm Kubernetes namespace. Airship Armada support is now considered to be deprecated. The Armada pod will continue to be deployed for use with any existing Armada based applications but will be removed in StarlingX Release 8.0, once the stx-openstack Armada application is fully migrated to FluxCD.

Cert-manager API Version deprecation

The upgrade of cert-manager from 0.15.0 to 1.7.1, deprecated support for cert manager API versions cert-manager.io/v1alpha2 and cert-manager.io/v1alpha3. When creating cert-manager CRDs (certificates, issuers, etc) with StarlingX Release 8.0, use cert-manager.io/v1.

Kubernetes APIs

Kubernetes APIs that will be removed in K8s 1.25 are listed below:

See: https://kubernetes.io/docs/reference/using-api/deprecation-guide/#v1-25

Release Information for other versions

You can find details about a release on the specific release page.

Version

Release Date

Notes

Status

StarlingX R8.0

2023-02

https://docs.starlingx.io/r/stx.8.0/releasenotes/index.html

Maintained

StarlingX R7.0

2022-07

https://docs.starlingx.io/r/stx.7.0/releasenotes/index.html

Maintained

StarlingX R6.0

2021-12

https://docs.starlingx.io/r/stx.6.0/releasenotes/index.html

Maintained

StarlingX R5.0.1

2021-09

https://docs.starlingx.io/r/stx.5.0/releasenotes/index.html

EOL

StarlingX R5.0

2021-05

https://docs.starlingx.io/r/stx.5.0/releasenotes/index.html

EOL

StarlingX R4.0

2020-08

EOL

StarlingX R3.0

2019-12

EOL

StarlingX R2.0.1

2019-10

EOL

StarlingX R2.0

2019-09

EOL

StarlingX R12.0

2018-10

EOL

StarlingX follows the release maintenance timelines in the StarlingX Release Plan.

The Status column uses OpenStack maintenance phase definitions.