Service Mesh Release Notes

The Red Hat OpenShift Service Mesh Operator is now available on ARM-based clusters as a Technology Preview feature.

The envoyExtAuthzGrpc field has been added, which is used to configure an external authorization provider using the gRPC API.

Common Vulnerabilities and Exposures (CVEs) have been addressed.

This release is supported on OpenShift Container Platform 4.10 and newer versions.

Added an option to the must-gather data collection tool that gathers information from a specified Service Mesh control plane namespace. For more information, see OSSM-351.

Improved performance for Service Mesh control planes with hundreds of namespaces

Simplifies installation, upgrades, and management of the Service Mesh control plane.

Reduces the Service Mesh control plane’s resource usage and startup time.

Improves performance by reducing inter-control plane communication over networking.

Removes the need to use Kubernetes Secrets, which have well known security risks.

Improves performance during certificate rotation, as proxies no longer require a restart to recognize new certificates.

ECDHE-ECDSA-AES128-SHA

ECDHE-RSA-AES128-SHA

AES128-GCM-SHA256

AES128-SHA

ECDHE-ECDSA-AES256-SHA

ECDHE-RSA-AES256-SHA

AES256-GCM-SHA384

AES256-SHA

ECDHE-ECDSA-AES128-SHA

ECDHE-RSA-AES128-SHA

AES128-GCM-SHA256

AES128-SHA

ECDHE-ECDSA-AES256-SHA

ECDHE-RSA-AES256-SHA

AES256-GCM-SHA384

AES256-SHA

Policy (authentication.istio.io/v1alpha1) is no longer supported. Depending on the specific configuration in your Policy resource, you may have to configure multiple resources to achieve the same effect.

ServiceMeshPolicy (maistra.io/v1) is no longer supported.

RbacConfig (rbac.istio.io/v1alpha1) is no longer supported.

ServiceMeshRbacConfig (maistra.io/v1) is no longer supported.

ServiceRole (rbac.istio.io/v1alpha1) is no longer supported.

ServiceRoleBinding (rbac.istio.io/v1alpha1) is no longer supported.

In Kiali, the login and LDAP strategies are deprecated. A future version will introduce authentication using OpenID providers.

OSSM-3890 Attempting to use the Gateway API in a multitenant mesh deployment generates an error message similar to the following:

2023-05-02T15:20:42.541034Z	error	watch error in cluster Kubernetes: failed to list *v1alpha2.TLSRoute: the server could not find the requested resource (get tlsroutes.gateway.networking.k8s.io)
2023-05-02T15:20:42.616450Z	info	kube	controller "gateway.networking.k8s.io/v1alpha2/TCPRoute" is syncing...

$ kubectl get crd gateways.gateway.networking.k8s.io ||   { kubectl kustomize "github.com/kubernetes-sigs/gateway-api/config/crd/experimental?ref=v0.5.1" | kubectl apply -f -; }

OSSM-2042 deployment of SMCP named default fails. If you are creating an SMCP object, and set its version field to v2.3, the name of the object cannot be default. If the name is default, then the control plane fails to deploy, and OpenShift generates a Warning event with the following message:

OSSM-1655 Kiali dashboard shows error after enabling mTLS in SMCP.

OSSM-1505 This issue only occurs when using the ServiceMeshExtension resource on OpenShift Container Platform 4.11. When you use ServiceMeshExtension on OpenShift Container Platform 4.11 the resource never becomes ready. If you inspect the issue using oc describe ServiceMeshExtension you will see the following error: stderr: Error creating mount namespace before pivot: function not implemented.

OSSM-1396 If a gateway resource contains the spec.externalIPs setting, instead of being recreated when the ServiceMeshControlPlane is updated, the gateway is removed and never recreated.

OSSM-1168 When service mesh resources are created as a single YAML file, the Envoy proxy sidecar is not reliably injected into pods. When the SMCP, SMMR, and deployment resources are created individually, the deployment works as expected.

OSSM-1115 The concurrency field of the spec.proxy API did not propagate to the istio-proxy. The concurrency field works when set with ProxyConfig. The concurrency field specifies the number of worker threads to run. If the field is set to 0, then the number of worker threads available is equal to the number of CPU cores. If the field is not set, then the number of worker threads available defaults to 2.

apiVersion: networking.istio.io/v1beta1
kind: ProxyConfig
metadata:
  name: mesh-wide-concurrency
  namespace: <istiod-namespace>
spec:
  concurrency: 0

OSSM-1052 When configuring a Service ExternalIP for the ingressgateway in the Service Mesh control plane, the service is not created. The schema for the SMCP is missing the parameter for the service.

OSSM-882 This applies for Service Mesh 2.1 and earlier. Namespace is in the accessible_namespace list but does not appear in Kiali UI. By default, Kiali will not show any namespaces that start with "kube" because these namespaces are typically internal-use only and not part of a mesh.

api:
  namespaces:
    exclude:
    - "^istio-operator"
    - "^kube-.*"
    - "^openshift.*"
    - "^ibm.*"
    - "^kiali-operator"

MAISTRA-2692 With Mixer removed, custom metrics that have been defined in Service Mesh 2.0.x cannot be used in 2.1. Custom metrics can be configured using EnvoyFilter. Red Hat is unable to support EnvoyFilter configuration except where explicitly documented. This is due to tight coupling with the underlying Envoy APIs, meaning that backward compatibility cannot be maintained.

MAISTRA-2648 Service mesh extensions are currently not compatible with meshes deployed on IBM Z.

MAISTRA-1959 Migration to 2.0 Prometheus scraping (spec.addons.prometheus.scrape set to true) does not work when mTLS is enabled. Additionally, Kiali displays extraneous graph data when mTLS is disabled.

spec:
  proxy:
    networking:
      trafficControl:
        inbound:
          excludedPorts:
          - 15020

MAISTRA-453 If you create a new project and deploy pods immediately, sidecar injection does not occur. The operator fails to add the maistra.io/member-of before the pods are created, therefore the pods must be deleted and recreated for sidecar injection to occur.

MAISTRA-158 Applying multiple gateways referencing the same hostname will cause all gateways to stop functioning.

KIALI-2206 When you are accessing the Kiali console for the first time, and there is no cached browser data for Kiali, the “View in Grafana” link on the Metrics tab of the Kiali Service Details page redirects to the wrong location. The only way you would encounter this issue is if you are accessing Kiali for the first time.

KIALI-507 Kiali does not support Internet Explorer 11. This is because the underlying frameworks do not support Internet Explorer. To access the Kiali console, use one of the two most recent versions of the Chrome, Edge, Firefox or Safari browser.

OSSM-4851 Previously, an error occurred in the operator deploying new pods in a namespace scoped inside the mesh when runAsGroup, runAsUser, or fsGroup parameters were nil. Now, a yaml validation has been added to avoid the nil value.

OSSM-3771 Previously, OpenShift routes could not be disabled for additional ingress gateways defined in a Service Mesh Control Plane (SMCP). Now, a routeConfig block can be added to each additionalIngress gateway so the creation of OpenShift routes can be enabled or disabled for each gateway.

OSSM-4197 Previously, if you deployed a v2.2 or v2.1 of the 'ServiceMeshControlPlane' resource, the /etc/cni/multus/net.d/ directory was not created. As a result, the istio-cni pod failed to become ready, and the istio-cni pods log contained the following message:

$ error   Installer exits with open /host/etc/cni/multus/net.d/v2-2-istio-cni.kubeconfig.tmp.841118073: no such file or directory

OSSM-3993 Previously, Kiali only supported OpenShift OAuth via a proxy on the standard HTTPS port of 443. Now, Kiali supports OpenShift OAuth over a non-standard HTTPS port. To enable the port, you must set the spec.server.web_port field to the proxy’s non-standard HTTPS port in the Kiali CR.

OSSM-3936 Previously, the values for the injection_label_rev and injection_label_name attributes were hardcoded. This prevented custom configurations from taking effect in the Kiali Custom Resource Definition (CRD). Now, the attribute values are not hardcoded. You can customize the values for the injection_label_rev and injection_label_name attributes in the spec.istio_labels specification.

OSSM-3644 Previously, the federation egress-gateway received the wrong update of network gateway endpoints, causing extra endpoint entries. Now, the federation-egress gateway has been updated on the server side so it receives the correct network gateway endpoints.

OSSM-3595 Previously, the istio-cni plugin sometimes failed on RHEL because SELinux did not allow the utility iptables-restore to open files in the /tmp directory. Now, SELinux passes iptables-restore via stdin input stream instead of via a file.

OSSM-3586 Previously, Istio proxies were slow to start when Google Cloud Platform (GCP) metadata servers were not available. When you upgrade to Istio 1.14.6, Istio proxies start as expected on GCP, even if metadata servers are not available.

OSSM-3025 Istiod sometimes fails to become ready. Sometimes, when a mesh contained many member namespaces, the Istiod pod did not become ready due to a deadlock within Istiod. The deadlock is now resolved and the pod now starts as expected.

OSSM-2493 Default nodeSelector and tolerations in SMCP not passed to Kiali. The nodeSelector and tolerations you add to SMCP.spec.runtime.defaults are now passed to the Kiali resource.

OSSM-2492 Default tolerations in SMCP not passed to Jaeger. The nodeSelector and tolerations you add to SMCP.spec.runtime.defaults are now passed to the Jaeger resource.

OSSM-2374 If you deleted one of the ServiceMeshMember resources, then the Service Mesh operator deleted the ServiceMeshMemberRoll. While this is expected behavior when you delete the last ServiceMeshMember, the operator should not delete the ServiceMeshMemberRoll if it contains any members in addition to the one that was deleted. This issue is now fixed and the operator only deletes the ServiceMeshMemberRoll when the last ServiceMeshMember resource is deleted.

OSSM-2373 Error trying to get OAuth metadata when logging in. To fetch the cluster version, the system:anonymous account is used. With the cluster’s default bundled ClusterRoles and ClusterRoleBinding, the anonymous account can fetch the version correctly. If the system:anonymous account loses its privileges to fetch the cluster version, OpenShift authentication becomes unusable.

OSSM-2371 Despite Kiali being configured as "view-only," a user can change the proxy logging level via the Workload details' Logs tab’s kebab menu. This issue has been fixed so the options under "Set Proxy Log Level" are disabled when Kiali is configured as "view-only."

OSSM-2344 Restarting Istiod causes Kiali to flood CRI-O with port-forward requests. This issue occurred when Kiali could not connect to Istiod and Kiali simultaneously issued a large number of requests to istiod. Kiali now limits the number of requests it sends to istiod.

OSSM-2335 Dragging the mouse pointer over the Traces scatterchart plot sometimes caused the Kiali console to stop responding due to concurrent backend requests.

OSSM-2221 Previously, gateway injection in the ServiceMeshControlPlane namespace was not possible because the ignore-namespace label was applied to the namespace by default.

$ oc label namespace <istio_system> maistra.io/ignore-namespace-

OSSM-2053 Using Red Hat OpenShift Service Mesh Operator 2.2 or 2.3, during SMCP reconciliation, the SMMR controller removed the member namespaces from SMMR.status.configuredMembers. This caused the services in the member namespaces to become unavailable for a few moments.

OSSM-1962 Use EndpointSlices in federation controller. The federation controller now uses EndpointSlices, which improves scalability and performance in large deployments. The PILOT_USE_ENDPOINT_SLICE flag is enabled by default. Disabling the flag prevents use of federation deployments.

OSSM-1668 A new field spec.security.jwksResolverCA was added to the Version 2.1 SMCP but was missing in the 2.2.0 and 2.2.1 releases. When upgrading from an Operator version where this field was present to an Operator version that was missing this field, the .spec.security.jwksResolverCA field was not available in the SMCP.

OSSM-1325 istiod pod crashes and displays the following error message: fatal error: concurrent map iteration and map write.

OSSM-1211 Configuring Federated service meshes for failover does not work as expected.

OSSM-1099 The Kiali console displayed the message Sorry, there was a problem. Try a refresh or navigate to a different page.

OSSM-1074 Pod annotations defined in SMCP are not injected in the pods.

OSSM-999 Kiali retention did not work as expected. Calendar times were greyed out in the dashboard graph.

OSSM-797 Kiali Operator pod generates CreateContainerConfigError while installing or updating the operator.

OSSM-722 Namespace starting with kube is hidden from Kiali.

OSSM-569 There is no CPU memory limit for the Prometheus istio-proxy container. The Prometheus istio-proxy sidecar now uses the resource limits defined in spec.proxy.runtime.container.

OSSM-535 Support validationMessages in SMCP. The ValidationMessages field in the Service Mesh Control Plane can now be set to True. This writes a log for the status of the resources, which can be helpful when troubleshooting problems.

OSSM-449 VirtualService and Service causes an error "Only unique values for domains are permitted. Duplicate entry of domain."

OSSM-419 Namespaces with similar names will all show in Kiali namespace list, even though namespaces may not be defined in Service Mesh Member Role.

OSSM-296 When adding health configuration to the Kiali custom resource (CR) is it not being replicated to the Kiali configmap.

OSSM-291 In the Kiali console, on the Applications, Services, and Workloads pages, the "Remove Label from Filters" function is not working.

OSSM-289 In the Kiali console, on the Service Details pages for the 'istio-ingressgateway' and 'jaeger-query' services there are no Traces being displayed. The traces exist in Jaeger.

OSSM-287 In the Kiali console there are no traces being displayed on the Graph Service.

OSSM-285 When trying to access the Kiali console, receive the following error message "Error trying to get OAuth Metadata".

MAISTRA-2735 The resources that the Service Mesh Operator deletes when reconciling the SMCP changed in Red Hat OpenShift Service Mesh version 2.1. Previously, the Operator deleted a resource with the following labels:

MAISTRA-2687 Red Hat OpenShift Service Mesh 2.1 federation gateway does not send the full certificate chain when using external certificates. The Service Mesh federation egress gateway only sends the client certificate. Because the federation ingress gateway only knows about the root certificate, it cannot verify the client certificate unless you add the root certificate to the federation import ConfigMap.

MAISTRA-2635 Replace deprecated Kubernetes API. To remain compatible with OpenShift Container Platform 4.8, the apiextensions.k8s.io/v1beta1 API was deprecated as of Red Hat OpenShift Service Mesh 2.0.8.

MAISTRA-2631 The WASM feature is not working because podman is failing due to nsenter binary not being present. Red Hat OpenShift Service Mesh generates the following error message: Error: error configuring CNI network plugin exec: "nsenter": executable file not found in $PATH. The container image now contains nsenter and WASM works as expected.

MAISTRA-2534 When istiod attempted to fetch the JWKS for an issuer specified in a JWT rule, the issuer service responded with a 502. This prevented the proxy container from becoming ready and caused deployments to hang. The fix for the community bug has been included in the Service Mesh 2.0.7 release.

MAISTRA-2411 When the Operator creates a new ingress gateway using spec.gateways.additionaIngress in the ServiceMeshControlPlane, Operator is not creating a NetworkPolicy for the additional ingress gateway like it does for the default istio-ingressgateway. This is causing a 503 response from the route of the new gateway.

MAISTRA-2401 CVE-2021-3586 servicemesh-operator: NetworkPolicy resources incorrectly specified ports for ingress resources. The NetworkPolicy resources installed for Red Hat OpenShift Service Mesh did not properly specify which ports could be accessed. This allowed access to all ports on these resources from any pod. Network policies applied to the following resources are affected:

MAISTRA-2378 When the cluster is configured to use OpenShift SDN with ovs-multitenant and the mesh contains a large number of namespaces (200+), the OpenShift Container Platform networking plugin is unable to configure the namespaces quickly. Service Mesh times out causing namespaces to be continuously dropped from the service mesh and then reenlisted.

MAISTRA-2370 Handle tombstones in listerInformer. The updated cache codebase was not handling tombstones when translating the events from the namespace caches to the aggregated cache, leading to a panic in the go routine.

MAISTRA-2117 Add optional ConfigMap mount to operator. The CSV now contains an optional ConfigMap volume mount, which mounts the smcp-templates ConfigMap if it exists. If the smcp-templates ConfigMap does not exist, the mounted directory is empty. When you create the ConfigMap, the directory is populated with the entries from the ConfigMap and can be referenced in SMCP.spec.profiles. No restart of the Service Mesh operator is required.

MAISTRA-2010 AuthorizationPolicy does not support request.regex.headers field. The validatingwebhook rejects any AuthorizationPolicy with the field, and even if you disable that, Pilot tries to validate it using the same code, and it does not work.

MAISTRA-1979 Migration to 2.0 The conversion webhook drops the following important fields when converting SMCP.status from v2 to v1:

MAISTRA-1947 Technology Preview Updates to ServiceMeshExtensions are not applied.

MAISTRA-1983 Migration to 2.0 Upgrading to 2.0.0 with an existing invalid ServiceMeshControlPlane cannot easily be repaired. The invalid items in the ServiceMeshControlPlane resource caused an unrecoverable error. The fix makes the errors recoverable. You can delete the invalid resource and replace it with a new one or edit the resource to fix the errors. For more information about editing your resource, see [Configuring the Red Hat OpenShift Service Mesh installation].

MAISTRA-1502 As a result of CVEs fixes in version 1.0.10, the Istio dashboards are not available from the Home Dashboard menu in Grafana. To access the Istio dashboards, click the Dashboard menu in the navigation panel and select the Manage tab.

MAISTRA-1399 Red Hat OpenShift Service Mesh no longer prevents you from installing unsupported CNI protocols. The supported network configurations has not changed.

MAISTRA-1089 Migration to 2.0 Gateways created in a non-control plane namespace are automatically deleted. After removing the gateway definition from the SMCP spec, you need to manually delete these resources.

MAISTRA-858 The following Envoy log messages describing deprecated options and configurations associated with Istio 1.1.x are expected:

MAISTRA-806 Evicted Istio Operator Pod causes mesh and CNI not to deploy.

MAISTRA-681 When the Service Mesh control plane has many namespaces, it can lead to performance issues.

MAISTRA-193 Unexpected console info messages are visible when health checking is enabled for citadel.

Bugzilla 1821432 The toggle controls in OpenShift Container Platform Custom Resource details page does not update the CR correctly. UI Toggle controls in the Service Mesh Control Plane (SMCP) Overview page in the OpenShift Container Platform web console sometimes updates the wrong field in the resource. To update a SMCP, edit the YAML content directly or update the resource from the command line instead of clicking the toggle controls.