Change Notes

Known Issues

  • Memory leak in TCP Wasm extensions affecting TCP telemetry (see Issue #24720). The leak occurs when upstream connections are interrupted mid-stream.

  • Wasm extension configuration updates are disruptive (see Issue #13690). The configuration is immediately applied for existing requests and connections, and is not reverted if the outer xDS is rejected.

  • Race condition with Envoy aggregate cluster when creating an EnvoyFilter and ServiceEntry for the same service. Istio-injected pods are unable to start up due to istio-proxy crashing with a segfault. See Issue #28620 for more information.

Traffic Management

  • Deprecated the use of the .global stub domain for multi-primary (formerly “replicated control planes”) multicluster configurations. The new onboarding flow uses a simpler configuration which allows services across the mesh to be accessed via *.cluster.local. There were several limitations with .global, such as poor load balancing when using gateways, which are no longer an issue with the new configuration.

  • Added DNS capture in istio-agent by default for VMs installed using istioctl x workload entry configure.

  • Added holdApplicationUntilProxyStarts field to ProxyConfig, allowing it to be configured at the pod level. Should not be used in conjunction with the deprecated values.global.proxy.holdApplicationUntilProxyStarts value. (Issue #27696)

  • Added support for INSERT_FIRST, INSERT_BEFORE, INSERT_AFTER insert operations for HTTP_ROUTE in EnvoyFilter (Issue #26692)

  • Added REPLACE operation for EnvoyFilter. REPLACE operation can replace the contents of a named filter with new contents. It is only valid for HTTP_FILTER and NETWORK_FILTER. (Issue #27425)

  • Added Istio resource status now includes observed generation (Issue #28003)

  • Fixed remove endpoints when the new labels in WorkloadEntry do not match the workloadSelector in ServiceEntry. (Issue #25678)

  • Fixed when a node has multiple IP addresses (e.g., a VM in the mesh expansion scenario), Istio Proxy will now bind inbound listeners to the first applicable address in the list (new behavior) rather than to the last one (former behavior). (Issue #28269)

Security

  • Improved Gateway certificates to be read and distributed from Istiod, rather than in the gateway pods. This reduces the permissions required in the gateways, improves performance, and makes certificate reading more extensible. This change is fully backwards compatible with the old mechanism, and requires no changes to your cluster. If required, it can be disabled by setting the ISTIOD_ENABLE_SDS_SERVER=false environment variable in Istiod. (Pull Request #27744)

  • Improved TLS configuration on sidecar server side inbound paths to enforce TLSv2 version along with recommended cipher suites. If this is not needed or creates problems with non Envoy clients, it can disabled by setting Istiod env variable PILOT_SIDECAR_ENABLE_INBOUND_TLS_V2 to false. (Pull Request #27500)

  • Updated The ipBlocks/notIpBlocks fields of an AuthorizationPolicy now strictly refer to the source IP address of the IP packet as it arrives at the sidecar. Prior to this release, if using the Proxy Protocol, then the ipBlocks/notIpBlocks would refer to the IP address determined by the Proxy Protocol. Now the remoteIpBlocks/notRemoteIpBlocks fields must be used to refer to the client IP address from the Proxy Protocol. (reference)(usage)(usage) (Issue #22341)

  • Added AuthorizationPolicy now supports nested JWT claims. (Issue #21340)

  • Added support for client side Envoy secure naming config when trust domain alias is used. This fixes the multi-cluster service discovery client SAN generation to use all endpoints’ service accounts rather than the first found service registry. (Pull Request #26185)

  • Added Experimental feature support allowing Istiod to integrate with external certificate authorities using Kubernetes CSR API (>=1.18 only). (Issue #27606)(usage)

  • Added Enable user to set the custom VM identity provider for credential authentication (Issue #27947)

  • Added action ‘AUDIT’ to Authorization Policy that can be used to determine which requests should be audited. (Issue #25591)

  • Added support for migration and concurrent use of regular K8S tokens as well as new K8S tokens with audience. This feature is enabled by default, can be disabled by REQUIRE_3P_TOKEN environment variable in Istiod, which will require new tokens with audience. The TOKEN_AUDIENCES environment variable allows customizing the checked audience, default remains istio-ca. (Pull Request #26482)

  • Added AuthorizationPolicy now supports a Source of type remoteIpBlocks/notRemoteIpBlocks that map to a new Condition attribute called remote.ip that can also be used in the “when” clause. If using an http/https load balancer in front of the ingress gateway, the remote.ip attribute is set to the original client IP address determined by the X-Forwarded-For http header from the trusted proxy configured through the numTrustedProxies field of the gatewayTopology under the meshConfig when you install Istio or set it via an annotation on the ingress gateway. See the documentation here: Configuring Gateway Network Topology. If using a TCP load balancer with the Proxy Protocol in front of the ingress gateway, the remote.ip is set to the original client IP address as given by the Proxy Protocol. (reference)(usage)(usage) (Issue #22341)

Telemetry

  • Updated the “Control Plane Dashboard” and the “Performance Dashboard” to use the container_memory_working_set_bytes metric to display memory. This metric only counts memory that cannot be reclaimed by the kernel even under memory pressure, and therefore more relevant for tracking. It is also consistent with kubectl top. The reported values are lower than the previous values.

  • Updated the Istio Workload and Istio Service dashboards resulting in faster load time. (Issue #22408)

  • Added datasource parameter to Grafana dashboards (Issue #22408)

  • Added Listener Access Logs when ResponseFlag from Envoy is set. (Issue #26851)

  • Added support for OpenCensusAgent formatted trace export with configurable trace context headers.

  • Added Proxy config to control Envoy native stats generation. (Issue #26546)

  • Added Istio Wasm Extension Grafana Dashboard. (Issue #25843)

  • Added gRPC streaming message count proxy Prometheus metrics istio_request_messages_total and istio_response_messages_total (Pull Request #3048)

  • Added support for properly labeling traffic in client metrics for cases when the destination is not reached or is not behind a proxy. (Issue #20538)

  • Fixed interpretation of $(HOST_IP) in Zipkin and Datadog tracer address. (Issue #27911)

  • Removed all Mixer-related features and functionality. This is a scheduled removal of a deprecated Istio services and deployments, as well as Mixer-focused CRDs and component and related functionality. (Issue #25333),(Issue #24300)

Installation

  • Promoted external control plane to alpha. (Issue #11)

  • Updated Kiali addon to version 1.26.

  • Added support for installing and upgrading Istio using Helm 3

  • Improved multi-network configuration so that labeling a service with topology.istio.io/network=network-name can configure cross-network gateways without using mesh networks.

  • Improved sidecar injection to not modify the pod securityPolicy.fsGroup which could conflict with existing settings and secret mounts. This option is enabled automatically on Kubernetes 1.19+ and is not supported on older versions. (Issue #26882)

  • Improved Generated operator manifests for use with kustomize are available in the manifests directory. (Issue #27139)

  • Updated install script to bypass GitHub API Rate Limiting. (Pull Request #23469)

  • Added port 15012 to the default list of ports for the istio-ingressgateway Service. (Issue #25933)

  • Added support for Kubernetes versions 1.16 to 1.19 to Istio 1.8. (Issue #28814)

  • Added the ability to specify the network for a Pod using the label topology.istio.io/network. This overrides the setting for the cluster’s installation values (values.globalnetwork). If the label isn’t set, it is injected based on the global value for the cluster. (Issue #25500)

  • Deprecated installation flags values.global.meshExpansion.enabled in favor of user-managed config and values.gateways.istio-ingressgateway.meshExpansionPorts in favor of components.ingressGateways[name=istio-ingressgateway].k8s.service.ports (Issue #25933)

  • Fixed Istio operator manager to allow configuring RENEW_DEADLINE. (Issue #27509)

  • Fixed an issue preventing NodePort services from being used as the registryServiceName in meshNetworks.

  • Removed support for installing third-party telemetry applications with istioctl. These applications (Prometheus, Grafana, Zipkin, Jaeger, and Kiali), often referred to as the Istio addons, must now be installed separately. This does not impact Istio’s ability to produce telemetry for those use in the addons. See Reworking our Addon Integrations for more info. (Issue #23868),(Issue #23583)

  • Removed istio-telemetry and istio-policy services and deployments from installation by istioctl. (Issue #23868),(Issue #23583)

  • Fixed Istio Grafana Dashboards queries which have reporter field. (Issue #27595)

istioctl

  • Improved istioctl analyze to find the exact line number with configuration errors when analyzing yaml files. Before, it would return the first line of the resource with the error. (Issue #22872)

  • Updated istioctl experimental version and proxy-status to use token security. A new option, --plaintext, has been created for testing without tokens. (Issue #24905)

  • Added istioctl commands may now refer to pods indirectly, for example istioctl dashboard envoy deployment/httpbin (Issue #26080)

  • Added io as short name for Istio Operator resources in addition to iop. (Issue #27159)

  • Added --type for istioctl experimental create-remote-secret to allow user specify type for the created secret.

  • Added an experimental OpenShift Kubernetes platform profile to istioctl. To install with the OpenShift profile, use istioctl install --set profile=openshift. (OpenShift Platform Setup)(Install OpenShift using istioctl)

  • Added istioctl bug-report command to generate an archive of Istio and cluster information to assist with debugging. (Issue #26045)

  • Added new command istioctl experimental istiod log to enable managing logging levels of istiod components. (Issue #25276),(Issue #27797)

  • Deprecated centralIstiod flag in favor of externalIstiod to better support external control plane model. (Issue #24471)

  • Fixed an issue which allowed an empty revision flag on install. (Issue #26940)

Documentation

  • Improved Multicluster install docs to include current best practices, incorporating recent updates to onboarding tooling. In particular, the multi-primary configuration (formerly known as “replicated control planes”) no longer relies on manually configuring the .global stub domain, preferring instead to use *.svc.cluster.local for accessing services throughout the mesh.
Was this information useful?
Do you have any suggestions for improvement?

Thanks for your feedback!