Upgrade Notes

When you upgrade from Istio 1.7.x to Istio 1.8.x, you need to consider the changes on this page. These notes detail the changes which purposefully break backwards compatibility with Istio 1.7.x. The notes also mention changes which preserve backwards compatibility while introducing new behavior. Changes are only included if the new behavior would be unexpected to a user of Istio 1.7.x.

Mixer is no longer supported in Istio

If you are using the istio-policy or istio-telemetry services, or any related Mixer configuration, you will not be able to upgrade without taking action to either (a) convert your existing configuration and code to the new extension model for Istio or (b) use the gRPC shim developed to bridge transition to the new model. For more details, please refer to the developer wiki.

The semantics of revision for gateways in IstioOperator has changed from 1.7 to 1.8

In 1.7, revision means you are creating a new gateway with a different revision so it would not conflict with the default gateway. In 1.8, it means the revision of istiod the gateway is configuring with. If you are using revision for gateways in IstioOperator in 1.7, before moving to 1.8, you must upgrade it to the revision of the Istiod (or delete the revision if you don’t use revision). See Issue #28849.

Istio CoreDNS Plugin Deprecation

The Istio sidecar now provides native support for DNS resolution with ServiceEntries using meshConfig.defaultConfig.proxyMetadata.ISTIO_META_DNS_CAPTURE="true". Previously, this support was provided by the third party Istio CoreDNS plugin. As a result, the istio-coredns-plugin is now deprecated and will be removed in a future release.

Use the new filter names for EnvoyFilter

If you are using EnvoyFilter API, it is recommended to change to the new filter names as described in Envoy’s deprecation notice The deprecated filter names will be supported in this release for backward compatibility but will be removed in future releases.

Inbound Cluster Name Format

The format of inbound Envoy cluster names has changed. Previously, they included the Service hostname and port name, such as inbound|80|http|httpbin.default.svc.cluster.local. This lead to issues when multiple Services select the same pod. As a result, we have removed the port name and hostname - the new format will instead resemble inbound|80||.

For most users, this is an implementation detail, and will only impact debugging or tooling that directly interacts with Envoy configuration.

Avoid use of mesh expansion installation flags

To ease setup for multicluster and virtual machines while giving more control to users, the meshExpansion and meshExpansionPorts installation flags have been deprecated, and port 15012 has been added to the default list of ports for the istio-ingressgateway Service.

For users with values.global.meshExpansion.enabled=true, perform the following steps before upgrading Istio:

  1. Apply the code sample for exposing Istiod through ingress.
Zip
$ kubectl apply -f @samples/multicluster/expose-istiod.yaml@

This removes operator.istio.io/managed labels from the associated Istio networking resources so that the Istio installer won’t delete them. After this step, you can modify these resources freely.

  1. If components.ingressGateways[name=istio-ingressgateway].k8s.service.ports is overridden, add port 15012 to the list of ports:
- port: 15012
    targetPort: 15012
    name: tcp-istiod
  1. If values.gateways.istio-ingressgateway.meshExpansionPorts is set, move all ports to components.ingressGateways[name=istio-ingressgateway].k8s.service.ports if they’re not already present. Then, unset this value.

  2. Unset values.global.meshExpansion.enabled.

Protocol Detection Timeout Changes

In order to support permissive mTLS traffic as well as automatic protocol detection, the proxy will sniff the first few bytes of traffic to determine the protocol used. For certain “server first” protocols, such as the protocol used by MySQL, there will be no initial bytes to sniff. To mitigate this issue in the past, Istio introduced a detection timeout. However, we found this caused frequent telemetry and traffic failures during slow connections, while increasing latency for misconfigured server first protocols rather than failing fast.

This timeout has been disabled by default. This has the following impacts:

  • Non “server first” protocols will no longer have a risk of telemetry or traffic failures during slow connections
  • Properly configured “server first” protocols will no longer have an extra 5 seconds latency on each connection
  • Improperly configured “server first” protocols will experience connection timeouts. Please ensure you follow the steps listed in Server First Protocols to ensure you do not run into traffic issues.

Update AuthorizationPolicy resources to use remoteIpBlocks/notRemoteIpBlocks instead of ipBlocks/notIpBlocks if using the Proxy Protocol

If using the Proxy Protocol on a load balancer in front an ingress gateway in conjunction with ipBlocks/notIpBlocks on an AuthorizationPolicy to perform IP-based access control, then please update the AuthorizationPolicy to use remoteIpBlocks/notRemoteIpBlocks instead after upgrading. The ipBlocks/notIpBlocks fields now strictly refer to the source IP address of the packet that arrives at the sidecar.

AUTO_PASSTHROUGH Gateway mode

Previously, gateways were configured with multiple Envoy cluster configurations for each Service in the cluster, even those not referenced by any Gateway or VirtualService. This was added to support the AUTO_PASSTHROUGH mode on Gateway, generally used for exposing Services across networks.

However, this came at an increased CPU and memory cost in the gateway and Istiod. As a result, we have disabled these by default on the istio-ingressgateway and istio-egressgateway.

If you are relying on this feature for multi-network support, please ensure you apply one of the following changes:

  1. Follow our new Multicluster Installation documentation.

This documentation will guide you through running a dedicate gateway deployment for this type of traffic (generally referred to as the eastwest-gateway). This eastwest-gateway will automatically be configured to support AUTO_PASSTHROUGH.

  1. Modify your installation of the gateway deployment to include this configuration. This is controlled by the ISTIO_META_ROUTER_MODE environment variable. Setting this to sni-dnat enables these clusters, while standard (the new default) disables them.
ingressGateways:
- name: istio-ingressgateway
    enabled: true
    k8s:
    env:
        - name: ISTIO_META_ROUTER_MODE
          value: "sni-dnat"
Was this information useful?
Do you have any suggestions for improvement?

Thanks for your feedback!