Difference between revisions of "Kubernetes/Istio"

From Ever changing code
Jump to navigation Jump to search
 
(90 intermediate revisions by the same user not shown)
Line 1: Line 1:
= Architecture Istio v1.7 =
= Architecture Istio ~v1.7+ =
:[[File:ClipCapIt-200930-172701.PNG]]
:[[File:ClipCapIt-200930-172701.PNG]]


Line 23: Line 23:
Security Architecture
Security Architecture
:[[File:ClipCapIt-210120-161513.PNG|600px]]
:[[File:ClipCapIt-210120-161513.PNG|600px]]
[https://istio.io/latest/blog/2021/upcoming-networking-changes/ Pod networking changed v1.10] matching default Kubernetes behaviour
:[[File:ClipCapIt-210531-145953.PNG|600px]]
[https://istio.io/latest/docs/ops/deployment/requirements/ Ports]
<source lang=bash>
service/istiod:
15010 - grpc-xds
15012 - https-dns
15014 - http-monitoring
443  - https-webhook, container port 15017
</source>


= Istio components =
= Istio components =
Line 88: Line 102:
<source lang=bash>
<source lang=bash>
# Minimum requirements are 8G and 4 CPUs
# Minimum requirements are 8G and 4 CPUs
PROFILE=minikube-v1.17.6-istio
K8SVERSION=v1.18.9
minikube start --memory=8192 --cpus=4 --kubernetes-version=v1.17.6 --profile $PROFILE
PROFILE=minikube-$K8SVERSION-istio
minikube start --memory=8192 --cpus=4 --kubernetes-version=v1.17.6 --driver kvm --profile $PROFILE-kvm2
minikube profile $PROFILE # set default profile
 
minikube start --kubernetes-version=$K8SVERSION
minikube start --memory=8192 --cpus=4 --kubernetes-version=$K8SVERSION --driver minikube
minikube start --memory=8192 --cpus=4 --kubernetes-version=$K8SVERSION --driver kvm


minikube tunnel --profile $PROFILE
minikube tunnel
minikube addons enable istio --profile $PROFILE # [1] error
minikube addons enable istio # [1] error
</source>
</source>


Line 107: Line 125:
</source>
</source>


= Install istioctl cli =
 
''The <code>curl ISTIO_VERSION=1.6.8 sh -</code> just downloads istio files into its own directory. It does not install anything''
Install [https://istio.io/latest/docs/examples/bookinfo/#deploying-the-application bookinfo] example app.
 
= Install =
== istioctl cli ==
''The <code>curl ISTIO_VERSION=1.10.0 sh -</code> just downloads istio files into its own directory. It does not install anything''




<source lang=bash>
<source lang=bash>
# Istio 1.7.x - option-1 - more official approche
# Option_1 - the official approach
curl -L https://istio.io/downloadIstio | ISTIO_VERSION=1.7.6 sh -
curl -L https://istio.io/downloadIstio | ISTIO_VERSION=1.10.0 sh -
cd istio-1.7.6/  # istio package directory
cd istio-1.10.0/  # istio package directory
export PATH=$PWD/bin:$PATH
export PATH=$PWD/bin:$PATH
export PATH=$PATH:/git3rd/istio-1.6.8/bin


# Istio 1.7.x - option-2
# Option_2
export ISTIO_VERSION=1.7.6
export ISTIO_VERSION=1.10.0
curl -L https://istio.io/downloadIstio | sh -
curl -L https://istio.io/downloadIstio | sh -
export PATH=$PWD/istio-$ISTIO_VERSION/bin:$PATH
export PATH=$PWD/istio-$ISTIO_VERSION/bin:$PATH
Line 126: Line 147:
# Check version
# Check version
istioctl version --remote
istioctl version --remote
client version: 1.7.3
client version: 1.9.1
control plane version: 1.7.3
control plane version: 1.9.1
data plane version: 1.7.3 (7 proxies)
data plane version: 1.9.1 (7 proxies)
 
# Pre-flight check
istioctl x precheck


# make sure you can connect to k8s cluster, then verify the install
# Verify the install
istioctl verify-install
istioctl verify-install
...
...
Line 139: Line 163:
Istio is installed successfully
Istio is installed successfully


# Verify mesh coverage and config status  
# Verify mesh coverage and the config status  
istioctl proxy-status
istioctl proxy-status
NAME                                                  CDS        LDS        EDS        RDS        ISTIOD                      VERSION
NAME                                                  CDS        LDS        EDS        RDS        ISTIOD                      VERSION
details-v1-5974b67c8-xbp8l.default                    SYNCED    SYNCED    SYNCED    SYNCED    istiod-5c6b7b5b8f-9npdz    1.7.3
details-v1-5974b67c8-xbp8l.default                    SYNCED    SYNCED    SYNCED    SYNCED    istiod-5c6b7b5b8f-9npdz    1.9.1
istio-ingressgateway-5689f7c67-gvrh8.istio-system    SYNCED    SYNCED    SYNCED    SYNCED    istiod-5c6b7b5b8f-9npdz    1.7.3
istio-ingressgateway-5689f7c67-gvrh8.istio-system    SYNCED    SYNCED    SYNCED    SYNCED    istiod-5c6b7b5b8f-9npdz    1.9.1
...
...
# Analyse the mesh configuration
istioctl analyze --all-namespaces
</source>
</source>


= Install the control plane =
== control plane ==
*[https://istio.io/latest/docs/setup/install/istioctl/ Install with Istioctl] - recommended
*[https://istio.io/latest/docs/setup/install/operator/ Istio Operator Install]
*[https://istio.io/latest/docs/setup/install/helm/ Install with Helm]
 
 
Istio maintainers with increasing complexity of the project that goes against the user friendliness still support ''helm manifest'' based configuration although there is fair movement towards the operator pattern. See below for differences, v1.6 and v1.7 still support both methods.
Istio maintainers with increasing complexity of the project that goes against the user friendliness still support ''helm manifest'' based configuration although there is fair movement towards the operator pattern. See below for differences, v1.6 and v1.7 still support both methods.


Line 158: Line 190:
* [deprecated] <code>istioctl manifest install</code> with <code>--set</code> or <code>-f <manifests.yaml></code>
* [deprecated] <code>istioctl manifest install</code> with <code>--set</code> or <code>-f <manifests.yaml></code>


 
== Install ==
<source lang=bash>
<source lang=bash>
# Tested with 1.7.4
# Tested with 1.7.3 [depreciating method via helm charts], --set values are prefixed with 'values.'
istioctl          install --skip-confirmation --set profile=default --dry-run # via operator
istioctl          upgrade --skip-confirmation --set profile=default --dry-run # via operator
 
# Tested with 1.7.3
istioctl manifest install --skip-confirmation --set profile=default --dry-run # via helm chart, depreciating
 
# Using istio.operator via 'istioctl' cli, the install can be replaced with 'upgrade'
istioctl install --skip-confirmation --set profile=default \
  --set        kiali.enabled=true      \
  --set        prometheus.enabled=true \
  --dry-run
 
# Using helm templates, notice the prefix 'values.'
istioctl manifest install --skip-confirmation --set profile=default \
istioctl manifest install --skip-confirmation --set profile=default \
   --set values.kiali.enabled=true      \
   --set values.kiali.enabled=true      \
   --set values.prometheus.enabled=true \
   --set values.prometheus.enabled=true \
   --dry-run
   --dry-run
</source>


Bring all to defaults set by a profile, you can see a number of resources getting removed/pruned.
# Tested with 1.8.2 [istio.operator via 'istioctl' cli]
<source lang=bash>
istioctl x precheck
istioctl manifest install --skip-confirmation --set profile=default
istioctl install --skip-confirmation --set profile=default --dry-run # via operator
✔ Istio core installed                                                             
istioctl upgrade --skip-confirmation --set profile=default --dry-run # via operator
✔ Istiod installed                                                                 
✔ Ingress gateways installed                                                       
- Pruning removed resources                                                            Removed Deployment:istio-system:prometheus.
  Removed Service:istio-system:prometheus.
  Removed ConfigMap:istio-system:prometheus.
  Removed ServiceAccount:istio-system:prometheus.
  Removed ClusterRole::kiali.
  Removed ClusterRole::kiali-viewer.
  Removed ClusterRole::prometheus-istio-system.
  Removed ClusterRoleBinding::kiali.
  Removed ClusterRoleBinding::prometheus-istio-system.
✔ Installation complete
</source>
</source>


Line 275: Line 280:
|}
|}


= Uninstall Istio =
;Uninstall
[https://istio.io/v1.6/docs/setup/getting-started/#uninstall Uninstall v1.6.8], it's safe to ignore RBAC not existing resources.
[https://istio.io/v1.6/docs/setup/getting-started/#uninstall Uninstall v1.6.8], it's safe to ignore RBAC not existing resources.
<source lang=bash>
<source lang=bash>
Line 288: Line 293:
✔ Uninstall complete
✔ Uninstall complete
</source>
</source>
== Install - day 2 operation ==
{{Note|Update 2021-06-01 the guide focus on Istio v1.10}}
This should be a preferred method for day 2 operations - [https://istio.io/latest/docs/setup/upgrade/canary/ canary upgrade] where control-plane gets installed with revisions in <code>istio-system</code> namespace and ingress-gateway(s) components are installed separately in another namespace eg. <code>istio-ingress</code>.
;Pre checks
<source lang=bash>
istioctl experimental precheck
istioctl analyze -A
# Optional labels for data plane operations
## Automatic proxy side injection
kubectl label namespace default istio-injection=enabled
kubectl label namespace default istio-injection=disabled
kubectl label namespace default istio.io/rev=1-10-0 # revision label
## Discovery namespace selectors v1.10+ !!!Warning labels should exist before enabling this feature otherwise you may break cp what can wach from api-server
kubectl label namespace istio-ingress istio-discovery=enabled # ingressgateway namespaces also need to be labelled (if in different ns than cp), otherwise the traffic won't be able to flow from ingress to our services
kubectl label namespace default      istio-discovery=enabled
</source>
=== Install control-plane ===
* [https://istio.io/latest/blog/2021/revision-tags/ Safely upgrade the Istio control plane with revisions and tags]
* [https://istio.io/latest/docs/setup/upgrade/canary/#stable-revision-labels-experimental Stable revision labels (experimental)]
{{Note|For installing v1.9.x you should first install without revisions}} eg.<code>REVISION=""</code> this is because revisions upgrade expects to have <code>default</code> revision set.
{{Note|Due to [https://github.com/istio/istio/issues/28880 a bug] in the creation of the ValidatingWebhookConfiguration during install, '''initial installations of Istio must not specify a revision'''. A workaround is to run command below, where <REVISION> is the name of the revision that should handle validation. This command creates an istiod service pointed to the target revision.
<source lang=bash>
kubectl get service -n istio-system -o json istiod-<REVISION> | jq '.metadata.name = "istiod" | del(.spec.clusterIP) | del(.spec.clusterIPs)' | kubectl apply -f -
</source>}}
{{Note|The difference in between operations <code>install</code> and <code>upgrade</code> is fuzzy. Upgrade safely upgrades components, and show diff in between the profile and what is being configured. On the other hand for eg. discoverySelectors feature only the install applies the config where using upgrade seems not having any effect.}}
<source lang=bash>
REVISION="" # don't use revisions
REVISION='--revision 1-10-0'
istioctl install -y -n istio-system $REVISION -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
metadata:
  name: control-plane
  #namespace: istio-system # it's default
spec:
  profile: minimal # installs only control-plane
  hub: gcr.io/istio-release
  components:
    pilot:
      k8s:
        hpaSpec:
          minReplicas: 2
        env:
        - name: PILOT_FILTER_GATEWAY_CLUSTER_CONFIG # Reduce Gateway Config to services that have routing rules defined
          value: "true"
  meshConfig:
    defaultConfig:
      proxyMetadata:
        # https://preliminary.istio.io/latest/blog/2020/dns-proxy/
        # curl localhost:8080/debug/nsdz?proxyID=<POD-NAME>
        # dig @localhost -p 15053 postgres-stolon-proxy.sample.svc
        ISTIO_META_DNS_CAPTURE: "true"
    enablePrometheusMerge: true
    #discoverySelectors: # new in v1.10, choose which namespaces Istiod should watch (from api-server)
    #- matchLabels:      # | for building the internal mesh registry based on configurations of services,endpoints,deployments
                        # | dynamically restrict the set of namespaces that are part of the mesh
    #    istio-discovery: enabled # <- example label
    #    #env: prod
    #    #region: us-east-1
    #- matchExpressions:
    #  - key: app
    #    operator: In
    #    values:
    #    - cassandrs
    #    - spark
EOF
) --dry-run
</source>
;Verify if there any clients are connected to Istio v1.10.0
<source lang=bash>
kubectl create deployment sleep --image=curlimages/curl -- /bin/sleep 3650d
kubectl exec -it deploy/sleep -- curl istiod-1-10-0.istio-system:15014/debug/connections | jq .
{
  "totalClients": 0,
  "clients": null
}
# example of output when client(s) are conencted
{
  "totalClients": 1,
  "clients": [
    {
      "connectionId": "sidecar~172.17.0.8~sleep-76f59b4b9c-4sw6t.default~default.svc.cluster.local-3",
      "connectedAt": "2021-06-01T04:57:10.663650097Z",
      "address": "172.17.0.1:8415",
      "watches": null
    }
  ]
}
# Show all discovered endpoints
kubectl exec -it deploy/sleep -- curl istiod-1-10-0.istio-system:15014/debug/endpointz | jq . # ~1800+ lines for empty Minikube
</source>
;Sidecar proxy automatic injection by telling a namespace which revision (aka control plane) to select/use.
<source lang=bash>
kubectl label namespace default istio.io/rev=1-10-0 istio-injection-
</source>
{{Note|The istio-injection label must be removed because it takes precedence over the istio.io/rev label for backward compatibility.}}
;Upgrade with Canary tag revision for isolated testing
<source lang=bash>
# Tag the new control-plane version with 'canary' tag
istioctl experimental revision tag set canary --revision 1-10-0
Revision tag "canary" created, referencing control plane revision "1-10-0". To enable injection using this
revision tag, use 'kubectl label namespace <NAMESPACE> istio.io/rev=canary'
# Show tagged revision
istioctl experimental revision list
REVISION TAG    ISTIO-OPERATOR-CR                                PROFILE REQD-COMPONENTS
1-10-0  canary istio-system/installed-state-control-plane-1-10-0 minimal base
                                                                          istiod
# Update a namespace label to use the newly tagged control-plane for the injected sidecars to use it/connect to it
kubectl label namespace default istio.io/rev=canary --overwrite
# Restart deployments for new proxy to be injected, test the application and if all is good delete the "app canary deployment" and continue with full rollout of all deployments.
</source>
=== Install ingress-gateway ===
We install the ingress-gateway components separated in a separated namespace <code>istio-ingress</code>. This is to allow to upgrade control-plane and ingressgateway(s) components to be independent. We achieve this using 'minimal' profile that includes only control CP components. Then for IGW install we use 'empty' profile that adds only the ingress-gateway. See a list of what is included in each [https://istio.io/latest/docs/setup/additional-setup/config-profiles/ profile].
:[[File:ClipCapIt-210317-090509.PNG]]
{{Note| '''tag''' and '''revision''' - in v1.10 ingressgateway is injected into the pod}} thus this becomes and behaves almost like any other workload. Thus the revision label should be in use eg. <code>istio.io/rev=1-10-0</code> to control the proxy version and to which control-plane the ingressgateway connects to
<source lang=bash>
# Create ingress own namespace
kubectl create ns istio-ingress
# Install revision
REVISION='--revision 1-10-0'
istioctl install -y -n istio-ingress $REVISION -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
metadata:
  name: istio-ingressgateway
spec:
  profile: empty # uses the 'empty' profile and enables the istio-ingressgateway component
  hub: gcr.io/istio-release
  components:
    ingressGateways:
    - name: istio-ingressgateway
      namespace: istio-ingress # if not set defaults to 'istio-system'
      enabled: true
      k8s:
        hpaSpec:
          minReplicas: 2
        service:
          type: LoadBalancer
        overlays:
        - apiVersion: apps/v1
          kind: Deployment
          name: istio-ingressgateway
          patches:
          - path: spec.template.spec.containers[name:istio-proxy].lifecycle
            value:
              preStop:
                exec:
                  command: ["sh", "-c", "sleep 5"]
EOF
) --dry-run
# Verify
# | Note that 'revision' relates to control-plane thus ingressgateway 'installed-state-istio-ingressgateway-1-10-0' installation
# | does hot have revision set
## Show IstioOperator installations
kubectl -n istio-system get istiooperators.install.istio.io
NAME                                          REVISION  STATUS  AGE
installed-state-control-plane-1-10-0          1-10-0              165m
installed-state-istio-ingressgateway-1-10-0  1-10-0              8m47s
## Show revisions
istioctl experimental revision list
REVISION TAG    ISTIO-OPERATOR-CR                                        PROFILE REQD-COMPONENTS
1-10-0  canary istio-system/installed-state-control-plane-1-10-0        minimal base
                                                                                istiod
                istio-system/installed-state-istio-ingressgateway-1-10-0 empty  ingress:istio-ingressgateway
## Test
GATEWAY_IP=$(kubectl get svc -n istio-system istio-ingressgateway -o jsonpath="{.status.loadBalancer.ingress[0].ip}")
# With Gateway and VirtualService we can route traffic to a destination service via IngressGateway
curl -H "Host: example.com" http://$GATEWAY_IP
</source>
{{Note|As the next step IngressGateway can be [https://istio.io/latest/docs/ops/integrations/certmanager/#kubernetes-ingress integrated] with Cert Manager}}
;Reduce Gateway Config
<source lang=bash>
# Show what services IngressGateway has discovered, the list has even those that no routing is defined
istioctl pc clusters deploy/istio-ingressgateway -n istio-system
# Set PILOT_FILTER_GATEWAY_CLUSTER_CONFIG environment variable in the istiod deployment
#  kind: IstioOperator
#  spec:
#    components:
#      pilot:
#        k8s:
#          env:                                                                           
#          - name: PILOT_FILTER_GATEWAY_CLUSTER_CONFIG                                                 
#            value: "true"
</source>
;(Optional) Access logging for gateway only using EnvoyFilter
* [https://istio.io/latest/docs/tasks/observability/logs/access-log/ Istio docs] explain to enable Envoy's Access Logs for the entire service mesh
<source lang=bash>
kubectl apply -f <(cat <<EOF
apiVersion: networking.istio.io/v1alpha3
kind: EnvoyFilter
metadata:
  name: ingressgateway-access-logging
  namespace: istio-system
spec:
  workloadSelector:
    labels:
      istio: ingressgateway
  configPatches:
  - applyTo: NETWORK_FILTER
    match:
      context: GATEWAY
      listener:
        filterChain:
          filter:
            name: "envoy.filters.network.http_connection_manager"
    patch:
      operation: MERGE
      value:
        typed_config:
          "@type": "type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager"
          access_log:
          - name: envoy.access_loggers.file
            typed_config:
              "@type": "type.googleapis.com/envoy.extensions.access_loggers.file.v3.FileAccessLog"
              path: /dev/stdout
              format: "[%START_TIME%] \"%REQ(:METHOD)% %REQ(X-ENVOY-ORIGINAL-PATH?:PATH)% %PROTOCOL%\" %RESPONSE_CODE% %RESPONSE_FLAGS% \"%UPSTREAM_TRANSPORT_FAILURE_REASON%\" %BYTES_RECEIVED% %BYTES_SENT% %DURATION% %RESP(X-ENVOY-UPSTREAM-SERVICE-TIME)% \"%REQ(X-FORWARDED-FOR)%\" \"%REQ(USER-AGENT)%\" \"%REQ(X-REQUEST-ID)%\" \"%REQ(:AUTHORITY)%\" \"%UPSTREAM_HOST%\" %UPSTREAM_CLUSTER% %UPSTREAM_LOCAL_ADDRESS% %DOWNSTREAM_LOCAL_ADDRESS% %DOWNSTREAM_REMOTE_ADDRESS% %REQUESTED_SERVER_NAME% %ROUTE_NAME%\n"
EOF
) --dry-run=server
</source>
;Check status, list and describe revisions
<source lang=bash>
kubectl get -n istio-system istiooperators.install.istio.io
NAME                                          REVISION  STATUS  AGE
installed-state-control-plane                                    22m
installed-state-control-plane-1-10-0          1-10-0              17h
installed-state-istio-ingressgateway-1-10-0  1-10-0              15h
# List revisions
istioctl x revision list
REVISION TAG      ISTIO-OPERATOR-CR                                        PROFILE REQD-COMPONENTS
1-10-0  canary  istio-system/installed-state-control-plane-1-10-0        minimal base
                                                                                  istiod
                  istio-system/installed-state-istio-ingressgateway-1-10-0 empty  ingress:istio-ingressgateway
default  <no-tag> istio-system/installed-state-control-plane              minimal base
                                                                                  istiod
# Describe revisions
istioctl x revision describe 1-10-0
ISTIO-OPERATOR CUSTOM RESOURCE: (2)
1. istio-system/installed-state-control-plane-1-10-0
  COMPONENTS:
  - base
  - istiod
  CUSTOMIZATIONS:
  - values.global.jwtPolicy=first-party-jwt
  - hub=gcr.io/istio-release
2. istio-system/installed-state-istio-ingressgateway-1-10-0
  COMPONENTS:
  - ingress:istio-ingressgateway
  CUSTOMIZATIONS:
  - components.ingressGateways[0].enabled=true
  - values.global.jwtPolicy=first-party-jwt
  - hub=gcr.io/istio-release
MUTATING WEBHOOK CONFIGURATIONS: (2)
WEBHOOK                      TAG
istio-revision-tag-canary    canary
istio-sidecar-injector-1-10-0 <no-tag>
CONTROL PLANE PODS (ISTIOD): (2)
NAMESPACE    NAME                          ADDRESS    STATUS  AGE
istio-system istiod-1-10-0-7c5b7865df-dcd92 172.17.0.5 Running 18h
istio-system istiod-1-10-0-7c5b7865df-xh5r9 172.17.0.3 Running 18h
INGRESS GATEWAYS: (0)
Ingress gateway is enabled for this revision. However there are no such pods. It could be that it is replaced by ingress-gateway from another revision (as it is still upgraded in-place) or it could be some issue with installation
EGRESS GATEWAYS: (0)
Egress gateway is disabled for this revision
</source>
{{Note|''Ingress gateway is enabled for this revision. However there are no such pods.'' - this is a bit blury message}}
;Uninstall
<source lang=bash>
# This command uninstalls all revisions in default namespace of 'istio-system'
istioctl x uninstall --purge # uninstall all
istioctl x uninstall -f <(cat <<EOF
...
EOF
)
</source>
Install book info application
<source lang=bash>
# Namespaces
kubectl create ns bookstore-1
kubectl create ns bookstore-2
kubectl label namespace bookstore-1 istio-injection=enabled
kubectl label namespace bookstore-2 istio.io/rev=1-9-1 istio-injection-
kubectl get ns --show-labels
NAME              STATUS  AGE    LABELS
bookstore-1      Active  155m  istio-injection=enabled
bookstore-2      Active  155m  istio.io/rev=1-10-0
# Deploy application
kubectl apply -f samples/bookinfo/platform/kube/bookinfo.yaml -n bookstore-1
kubectl apply -f samples/bookinfo/platform/kube/bookinfo.yaml -n bookstore-2
istioctl proxy-status --revision 1-10-0 # show only proxies of revision '1-10-0'
istioctl proxy-status                  # show all proxies
</source>
You can see proxies connected only to revision '1-9-1' and at the bottom all proxies
:[[File:ClipCapIt-210317-134114.PNG]]
{{Note|Screen grab from previous Istio v1.9 version}}
= Ingress flow from ALB to the service =
This example graph shows the traffic flow from ALB -> a container
:[[File:ClipCapIt-210828-101105.PNG]]


= Get [https://istio.io/latest/docs/setup/install/istioctl/ info] =
= Get [https://istio.io/latest/docs/setup/install/istioctl/ info] =
Line 463: Line 828:
* [https://medium.com/@anshul.prakash17/creating-an-internal-load-balancer-in-aws-eks-using-istio-3defcb51c6a2 Creating an Internal Load Balancer in AWS EKS using Istio]
* [https://medium.com/@anshul.prakash17/creating-an-internal-load-balancer-in-aws-eks-using-istio-3defcb51c6a2 Creating an Internal Load Balancer in AWS EKS using Istio]
* [https://stackoverflow.com/questions/62407364/how-to-set-aws-alb-instead-of-elb-in-istio how-to-set-aws-alb-instead-of-elb-in-istio]
* [https://stackoverflow.com/questions/62407364/how-to-set-aws-alb-instead-of-elb-in-istio how-to-set-aws-alb-instead-of-elb-in-istio]
* [https://rtfm.co.ua/en/istio-external-aws-application-loadbalancer-and-istio-ingress-gateway/ Istio: external AWS Application LoadBalancer and Istio Ingress Gateway] published on 2021-04-22


Gateways are a special type of component, since multiple ingress and egress gateways can be defined. In the IstioOperator API, gateways are defined as a list type. The default profile installs one ingress gateway, called istio-ingressgateway.  
Gateways are a special type of component, since multiple ingress and egress gateways can be defined. In the IstioOperator API, gateways are defined as a list type. The default profile installs one ingress gateway, called istio-ingressgateway.  
Line 484: Line 850:
istioctl install \
istioctl install \
   --set profile=default \
   --set profile=default \
  --set addonComponents.prometheus.enabled=false \
  --set addonComponents.grafana.enabled=false \
  --set addonComponents.kiali.enabled=false \
  --set addonComponents.tracing.enabled=false \
   --set components.ingressGateways[0].enabled="true" \
   --set components.ingressGateways[0].enabled="true" \
   --set components.ingressGateways[0].k8s.serviceAnnotations."service\.beta\.kubernetes\.io/aws-load-balancer-internal"=\"true\"
   --set components.ingressGateways[0].k8s.serviceAnnotations."service\.beta\.kubernetes\.io/aws-load-balancer-internal"=\"true\"
Line 536: Line 898:
</source>
</source>


== Istio-ingressgateway - object and traffic flow
== Ingressgateway Kubernetes service ==
The istio <code>istio-ingressgateway</code> works as any other ingress controller, it has service type: LoadBalancer, that creates an endpoint with 4 ports, these in turn make external-loadbalancer to listen on these ports:
The istio <code>istio-ingressgateway</code> works as any other ingress controller, it has service type: LoadBalancer, that creates an endpoint with 4 ports, these in turn make external-loadbalancer to listen on these ports:
<source>
<source>
Line 592: Line 954:
     status-port  15021  TCP
     status-port  15021  TCP
     http2        8080  TCP
     http2        8080  TCP
</source>
== General traffic flow ==
{{Note|This overview applies to each workload including Ingress-gateways}}
The arrows in the diagram show the flow of a request through the configuration, and the five key elements are the ‘listener,’ ‘filter chains,’ ‘routes,’ ‘clusters,’ and ‘endpoints’. The route is part of the filter chain, which is part of the listener.
:[[File:ClipCapIt-210503-164443.PNG]]
* listener - 'binds' to a port and listens for inbound requests. Any request that comes in via another port would not be seen or handled by Envoy. In Envoy terms, it manages downstream requests
* filter chain - after being accepted by 'the listener', describes how the request should be handled once it’s entered Envoy
* route - extension of the filter chain, takes the HTTP request information and directs it to the correct service
* cluster - address IP:port that are the backend for a service. A cluster is a group of logically similar upstream hosts that Envoy connects to. Envoy discovers the members of a cluster via CDS(Cluster Discovery Service).
* endpoints - network nodes that implement a logical service. They are grouped into clusters. Endpoints in a cluster are upstream of an Envoy proxy.
== Ingress-gateway - routes verification ==
<source lang=bash>
INGRESS_GATEWAY_LABEL=ingressgateway # ingressgateway-internal|external
INGRESS_GATEWAY_POD=$(kubectl -n istio-system get pod  -l istio=$INGRESS_GATEWAY_LABEL -o jsonpath='{.items[0].metadata.name}')
# Get listeners, other options: bootstrap,cluster,endpoint,listener,log,route,secret
istioctl proxy-config listeners -n istio-system $INGRESS_GATEWAY_POD
ADDRESS PORT  MATCH DESTINATION
0.0.0.0 8080  ALL  Route: http.80
0.0.0.0 15021 ALL  Inline Route: /healthz/ready*
0.0.0.0 15090 ALL  Inline Route: /stats/prometheus*
# Show routes, and virtual-services
istioctl proxy-config routes -n istio-system $INGRESS_GATEWAY_POD
NOTE: This output only contains routes loaded via RDS.
NAME        DOMAINS                                                        MATCH                  VIRTUAL SERVICE
http.80    alertmanager.staging.net                    /*                    alertmanager.monitoring
http.80    grafana.staging.net                          /*                    grafana.monitoring
http.80    prometheus.staging.net                      /*                    prometheus.monitoring
            *                                                              /healthz/ready*       
            *                                                              /stats/prometheus*
</source>
</source>


Line 792: Line 1,191:


;<code>requestauthentications</code>: it's for humans outside of the cluster to validate their identity and decide to allow enter the cluster, used for end-user authentication to verify credentials attached to the request.
;<code>requestauthentications</code>: it's for humans outside of the cluster to validate their identity and decide to allow enter the cluster, used for end-user authentication to verify credentials attached to the request.
<syntaxhighlight lang="js">code</syntaxhighlight>
<syntaxhighlight lang="js">
apiVersion: "security.istio.io/v1beta1"
apiVersion: "security.istio.io/v1beta1"
kind: "RequestAuthentication"
kind: "RequestAuthentication"
Line 819: Line 1,218:
If any <tt>AuthorizationPolicy</tt> applies to a workload, access to that workload is denied by default, unless explicitly allowed by a rule declared in the policy.
If any <tt>AuthorizationPolicy</tt> applies to a workload, access to that workload is denied by default, unless explicitly allowed by a rule declared in the policy.


<syntaxhighlight lang="js">code</syntaxhighlight>
<syntaxhighlight lang="js">
# allow-all policy full access to all workloads in the default namespace
# allow-all policy full access to all workloads in the default namespace
apiVersion: security.istio.io/v1beta1
apiVersion: security.istio.io/v1beta1
Line 1,016: Line 1,415:
<job shell script aka command:>
<job shell script aka command:>
curl -X POST http://127.0.0.0/quitquitquit
curl -X POST http://127.0.0.0/quitquitquit
</source>
= [https://istio.io/latest/docs/ops/diagnostic-tools/ Debugging] =
* [https://istio.io/latest/docs/ops/diagnostic-tools/proxy-cmd/ Debugging Envoy and Istiod]
Analize
<source lang=bash>
istioctl analyze --all-namespaces
istioctl analyze samples/bookinfo/networking/*.yaml
istioctl analyze --use-kube=false samples/bookinfo/networking/*.yaml
# Enable live analyzer, it will add to the status field eg.<code>status.validationMessages.type.code: IST0101</code>
istioctl install --set values.global.istiod.enableAnalysis=true
</source>
Debug a singular pod or a deployment pods
<syntaxhighlight lang="yaml">
apiVersion: v1
kind: Namespace
metadata:
  name: redis
  labels:
    istio.io/rev: asm-173-6
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: debug-pod
  namespace: redis
  labels:
    app: debug-pod
spec:
  replicas: 1
  selector:
    matchLabels:
      app: debug-pod
  template:
    metadata:
      labels:
        app: debug-pod
      annotations:
        sidecar.istio.io/logLevel: debug # <- enable debug mode
    spec:
      containers:
        - image: samos123/docker-toolbox
          name: debug-pod
</syntaxhighlight>
Istiocli debug commands
<source lang=bash>
istioctl experimental describe pod <pod-name>[.<namespace>]
istioctl proxy-status
# | SYNCED means that Envoy has acknowledged the last configuration Istiod has sent to it.
# | NOT SENT means that Istiod hasn’t sent anything to Envoy. This usually is because Istiod has nothing to send.
# | STALE means that Istiod has sent an update to Envoy but has not received an acknowledgement. This usually indicates a networking issue between Envoy and Istiod or a bug with Istio itself.
# Retrieve information about cluster configuration for the Envoy instance in a specific pod:
istioctl proxy-config cluster <pod-name> [flags]
# To retrieve information about bootstrap configuration for the Envoy instance in a specific pod:
istioctl proxy-config bootstrap <pod-name> [flags]
# To retrieve information about listener configuration for the Envoy instance in a specific pod:
istioctl proxy-config listener <pod-name> [flags]
# To retrieve information about route configuration for the Envoy instance in a specific pod:
istioctl proxy-config route <pod-name> [flags]
# To retrieve information about endpoint configuration for the Envoy instance in a specific pod:
istioctl proxy-config endpoints <pod-name> [flags]
istioctl proxy-config endpoints productpage-v1-6c886ff494-7vxhs --cluster "outbound|9080||reviews.default.svc.cluster.local"
</source>
= [https://github.com/salesforce/helm-starter-istio Istio helm starter] =
Project by Salesforce to scaffold a helm chart with Istio that uses helm-starter plugin.
<source lang=bash>
# Install `helm-starter` helm plugin
helm plugin install https://github.com/salesforce/helm-starter.git
# Fetch starter templates: mesh-service, ingress-service, mesh-egress, auth-policy
helm starter fetch https://github.com/salesforce/helm-starter-istio.git
# Create a helm chart from the starter
helm create NAME --starter helm-starter-istio/[starter-name]
for starter in mesh-service ingress-service mesh-egress auth-policy; do helm create authservice-$starter --starter helm-starter-istio/$starter; done
</source>
== Important labels ==
Deployment
<syntaxhighlightjs lang="yaml">
apiVersion: apps/v1
kind: Deployment
spec:
  selector:
    matchLabels:
      app: {{ .Values.service | lower | quote }} # <- Istio label
      version: {{ .Values.version | quote }}    # <- Istio label
  template:
    metadata:
      labels:
        # Kubernetes recommended labels
        app.kubernetes.io/name: {{ .Values.service | lower | quote }}
        app.kubernetes.io/part-of: {{ .Values.system | quote }}
        app.kubernetes.io/version: {{ .Values.version | quote }}
        # Isio required labels
        app: {{ .Values.service | lower | quote }} # <- Istio label
        version: {{ .Values.version | quote }}    # <- Istio label
{{- if .Values.configMap }}
      annotations:
        checksum/config: {{ include (print $.Template.BasePath "/configmap.yaml") . | sha256sum }}
{{- end }}
    spec:
      containers:
{{- range .Values.ports }}
            - name: {{ .name | quote }}
              containerPort: {{ .targetPort }}
              protocol: TCP # <-- TCP|HTTP or any other that is accepted
{{- end }}
</syntaxhighlightjs>
Service
<syntaxhighlightjs lang="yaml">
apiVersion: v1
kind: Service
spec:
  ports:
{{- range .Values.ports }}
    - name: {{ .name | quote }} # <- tcp-<name>, http-<name>, grcp-<name>
      port: {{ .port }}
      targetPort: {{ .targetPort }}
      protocol: TCP # <- TCP|HTTP or any other that is accepted
{{- end }}
</syntaxhighlightjs>
= Envoy =
* Upstream connections are the service Envoy is initiating the connection to.
* Downstream connections are the client that is initiating a request through Envoy.
:[[File:ClipCapIt-210525-080849.PNG|none|left|400px]]
Terminology in my own words:
;Listeners: Declaration of ports to open on the proxy and listen for incoming connections
;Clusters: Backend service to which we can route traffic. Also a group of logically similar upstream hosts that Envoy connects to
;Envoy Discovery (XDS)
Envoy is driven by a set of APIs that configure certain aspects of the proxy. We saw earlier how we specified clusters, listeners, and routes. We can also configure those over Envoy's xDS APIs:
* Listener discovery service (LDS)
* Route discovery service (RDS)
* Cluster discovery service (CDS)
* Service discovery service (SDS/EDS)
* Aggregated discovery service (ADS)
Now we don't have to specify listeners ahead of time. We can open them up or close them at runtime. Below example of Envoy config for dynamic Listeners configuration, set/received from <code>api_config_source</code>.
<syntaxhighlightjs lang="yaml">
dynamic_resources:
    lds_config: 
    api_config_source:
        api_type: GRPC
        grpc_services:
        - envoy_grpc:
            cluster_name: xds_cluster
</syntaxhighlightjs>
Configuring Envoy through this API is exactly what Istio's control plane does. With Istio we specify configurations in a more user-friendly format and Istio translates that configuration into something Envoy can understand and delivers this configuration through Envoy's xDS API.
== Service-to-service communication ==
:[[File:ClipCapIt-210324-180344.PNG]]
This is zoom in on inbound connection from the <code>sleep</code> to <code>http</code> pod.
[[File:istio-envoy-2.jpg|left|Istio]]
.
== Troubleshooting ==
<source lang=bash>
# Interrogate envoy stats endpoint
kubectl exec -it deploy/sleep -- curl http://envoy:15000/stats| grep retry
</source>
= Troubleshooting istio =
Grab whole mesh config (new in v1.10) and paste into https://envoyui.solo.io/ analyzer
<source lang=bash>
istioctl proxy-config all deploy/sleep -n default -o json > sleep.deploy.all-config_dump.json
</source>
Check istiod endpoints, get registry
<source lang=bash>
# Get all services in the Istio registry, workloads are also included even if they are not officially part of the mesh
kubectl exec -n istio-system -it deploy/istiod-1-8-4 -- pilot-discovery request GET /debug/registryz
kubectl exec -n istio-system -it deploy/istiod-1-8-4 -- pilot-discovery version # [discovery,help,request,version]
</source>
Debug info from istiod control-plane
<source lang=bash>
# istiod has istiod:8080/debug endpoint 'internal-debug' enables this using secure channel
## curl /debug endpoint
kubectl -n istio-system port-forward deploy/istiod 8080
firefox http://127.0.0.1:8080/debug      # all debug endpoints
firefox http://127.0.0.1:8080/debug/mesh # show mesh config
## internal-debug (not working)
istioctl x internal-debug syncz
</source>
Verify an app/pod iptables modified by istio-proxy. This rules are in the Network kernel namespace hold by the pod.
<source lang=bash>
# List iptables rules - this should fail
kubectl -n default exec -it deploy/httpbin -c istio-proxy -- sudo iptables -L -t nat
sudo: effective uid is not 0, is /usr/bin/sudo on a file system with the 'nosuid' option set or an NFS file system without root privileges?
# Grant the proxy sidecar privileges escalation
kubectl edit deploy/httpbin -n default
      containers:
      - name: istio-proxy
        image: docker.io/istio/proxyv2:1.8.4
        securityContext:
          allowPrivilegeEscalation: true # by default it's false
          privileged: true              # by default it's false
# Show iptables - 15006(incoming) and 15001(outgoing) traffic interception
kubectl -n default exec -it deploy/httpbin -c istio-proxy -- sudo iptables -L -t nat | grep -e ISTIO_IN_REDIRECT -e 15006 -e ISTIO_REDIRECT -e 15001
ISTIO_IN_REDIRECT  tcp  --  anywhere            anywhere           
Chain ISTIO_IN_REDIRECT (3 references)
REDIRECT  tcp  --  anywhere            anywhere            redir ports 15006
ISTIO_IN_REDIRECT  all  --  anywhere            !localhost            owner UID match istio-proxy
ISTIO_IN_REDIRECT  all  --  anywhere            !localhost            owner GID match istio-proxy
ISTIO_REDIRECT  all  --  anywhere            anywhere           
Chain ISTIO_REDIRECT (1 references)
REDIRECT  tcp  --  anywhere            anywhere            redir ports 15001
</source>
We can see here iptables is used to redirect incoming and outgoing traffic to Istio's data plane proxy. Incoming traffic goes to port <code>15006</code> of the Istio proxy while outgoing traffic will go through <code>15001</code>. If we check the Envoy listeners for those ports, we can see exactly how the traffic gets handled.
;IngressGateway
Query the gateway configuration to get all routes
<source lang=bash>
istioctl proxy-config routes deploy/istio-ingressgateway.istio-system
NOTE: This output only contains routes loaded via RDS.
NAME        DOMAINS              MATCH                  VIRTUAL SERVICE
http.80    example.com          /*                    web-api-gw-vs.default
            *                    /stats/prometheus*   
            *                    /healthz/ready*
# Show details of a single route
istioctl proxy-config routes deploy/istio-ingressgateway.istio-system --name http.80 -o json
</source>
</source>


Line 1,031: Line 1,683:
* [https://www.youtube.com/watch?v=zVlepkF-hJc Istio Architecture Part 3 - The Control Plane]
* [https://www.youtube.com/watch?v=zVlepkF-hJc Istio Architecture Part 3 - The Control Plane]
* [https://www.youtube.com/watch?v=BlQryF29L3Q Istio Architecture Part 4 - Going Deeper into Envoy]
* [https://www.youtube.com/watch?v=BlQryF29L3Q Istio Architecture Part 4 - Going Deeper into Envoy]
* [https://getistio.io/istio-cheatsheet/ istio-cheatsheet] by Tetratate A+++
Istio v1.7
Istio v1.7
* [https://www.youtube.com/watch?v=HZonKiu7onk VM <-> K8s mesh] demo at 30 min
* [https://www.youtube.com/watch?v=HZonKiu7onk VM <-> K8s mesh] demo at 30 min
Semi-related
* [https://github.com/csantanapr/knative-kind Knative on Kind (KonK)]
Related technology
* [https://developers.google.com/web/fundamentals/performance/http2 Introduction to HTTP/2] http/2 is used by Envoy internally
Tutorials and demos:
* Upgrading Istio without Downtime 1-9-7 to 1-10-3
** [https://www.solo.io/blog/upgrading-istio-without-downtime/ upgrading-istio-without-downtime] Blog
** [https://github.com/nmnellis/istio-upgrade-demo istio-upgrade-demo] GitHub
* [https://dzone.com/articles/safely-upgrade-the-istio-control-plane-with-revisi?oid&utm_medium=email&_hsmi=148688354&_hsenc=p2ANqtz-8dBVjZA0BJoWbS4e-Guer014AS5F2-5lgkBH63X87QKdnr5N4XB7d4AmRHjjWXDO7NfmsE4hVgOG-y-RLAtNawPzLvQbEXuVEfsRJVcAadDTxrxJ8&utm_content=148687162&utm_source=hs_email# Safely Upgrade the Istio Control Plane With Revisions and Tags] 05-Aug-2021
Investigate TLS:
* [https://redhat-scholars.github.io/istio-tutorial/istio-tutorial/1.6.x/8mTLS.html Mutual TLS and Istio] inspect with tcpdump and ksniff
* [https://www.programmersought.com/article/35504882940/ Use ksniff and Wireshark to verify service mesh TLS in Kubernetes]
Solutions:
* [https://dev.to/peterj/what-are-sticky-sessions-and-how-to-configure-them-with-istio-1e1a What are sticky sessions and how to configure them with Istio?]
* [https://medium.com/@cy.chiang/how-to-integrate-aws-alb-with-istio-v1-0-b17e07cae156 integrate AWS ALB with istio] 2018

Latest revision as of 10:20, 28 August 2021

Architecture Istio ~v1.7+

ClipCapIt-200930-172701.PNG
Namespace: <app namespace>
  | app1  |                | app2  |  # main container
  | proxy |  <---------->  | proxy |  # Data Plane (all Envoy sidecar proxies)
  |  pod  |                |  pod  |

Namespace: istio-system
| |citadel|        |mixer|       |pilot| |
| |  pod  |        | pod |       | pod | | 
|      C o n t r o l  P l a n e  A P I   |
 ----------------------------------------

Note: All proxies are collectively named Data Plane and everything else that Istio deployed is called Control Plane


Note: Proxy term meaning is when someone has authority to represent someone. In software proxy components are invisible to clients. proxies


Security Architecture

ClipCapIt-210120-161513.PNG


Pod networking changed v1.10 matching default Kubernetes behaviour

ClipCapIt-210531-145953.PNG


Ports

service/istiod:
15010 - grpc-xds
15012 - https-dns
15014 - http-monitoring
443   - https-webhook, container port 15017

Istio components

  • Istio-telemetry
  • Istio-pilot
  • Istio-tracing


Istio components
Envoy L7 proxy Pilot Citadel Mixer[deprecate] Galley
  • Dynamic service discovery
  • Load balancing
  • TLS termination
  • Health checks
  • Staged rollouts
  • Fault injection

Converts Istio configuration into a format that Envoy can understand.

Aware about pods health, what pods are available and sends to the proxy pods that are alive with any other configuration updates.

  • Propagates the configuration to the Proxies
  • Service discovery
  • Intelligent routing
  • Resiliency

Manages certificates, allows to enable TLS/SSL across entire cluster.

  • User authentication
  • Credential management
  • Certificate management
  • Traffic encryption

Pods

  • istio-citadel-*

It's certificate store.

  • handles Access control
  • Usage policies, rate limiting
  • Telemetry data (data scraping)

It has a lot of modules/plugins. Pods: istio-policy-* istio-telemetry-*

Interface for underlying Istio API gateway(aka server). It reads in k8s yaml and transforms it into internal structure Istio understand.


Istio UI components:

  • grafana:3000 - dashboards
  • kiali:31000 - visualisation, tells what services are part of istio, how are they connected and performing
  • jaeger:31001 - tracing


Noticeable changes
  • In Istio 1.6, completed transition and fully moved functionality into Istiod. This has allow to remove the separate deployments for Citadel, the sidecar injector, and Galley.

Istio on minikube

# Minimum requirements are 8G and 4 CPUs
K8SVERSION=v1.18.9
PROFILE=minikube-$K8SVERSION-istio
minikube profile $PROFILE # set default profile

minikube start --kubernetes-version=$K8SVERSION
minikube start --memory=8192 --cpus=4 --kubernetes-version=$K8SVERSION --driver minikube
minikube start --memory=8192 --cpus=4 --kubernetes-version=$K8SVERSION --driver kvm

minikube tunnel
minikube addons enable istio # [1] error

Troubleshooting

[1] - no matches for kind "IstioOperator"
💣  enable failed: run callbacks: running callbacks: [sudo KUBECONFIG=/var/lib/minikube/kubeconfig /var/lib/minikube/binaries/v1.17.6/kubectl apply -f /etc/kubernetes/addons/istio-default-profile.yaml: Process exited with status 1
stdout:
namespace/istio-system unchanged

stderr:
error: unable to recognize "/etc/kubernetes/addons/istio-default-profile.yaml": no matches for kind "IstioOperator" in version "install.istio.io/v1alpha1"


Install bookinfo example app.

Install

istioctl cli

The curl ISTIO_VERSION=1.10.0 sh - just downloads istio files into its own directory. It does not install anything


# Option_1 - the official approach
curl -L https://istio.io/downloadIstio | ISTIO_VERSION=1.10.0 sh -
cd istio-1.10.0/  # istio package directory
export PATH=$PWD/bin:$PATH

# Option_2
export ISTIO_VERSION=1.10.0
curl -L https://istio.io/downloadIstio | sh -
export PATH=$PWD/istio-$ISTIO_VERSION/bin:$PATH
export ISTIO_INSTALL_DIR=$PWD/istio-$ISTIO_VERSION

# Check version
istioctl version --remote
client version: 1.9.1
control plane version: 1.9.1
data plane version: 1.9.1 (7 proxies)

# Pre-flight check
istioctl x precheck

# Verify the install
istioctl verify-install
...
CustomResourceDefinition: templates.config.istio.io.default checked successfully
CustomResourceDefinition: istiooperators.install.istio.io.default checked successfully
Checked 25 custom resource definitions
Checked 1 Istio Deployments
Istio is installed successfully

# Verify mesh coverage and the config status 
istioctl proxy-status
NAME                                                  CDS        LDS        EDS        RDS        ISTIOD                      VERSION
details-v1-5974b67c8-xbp8l.default                    SYNCED     SYNCED     SYNCED     SYNCED     istiod-5c6b7b5b8f-9npdz     1.9.1
istio-ingressgateway-5689f7c67-gvrh8.istio-system     SYNCED     SYNCED     SYNCED     SYNCED     istiod-5c6b7b5b8f-9npdz     1.9.1
...

# Analyse the mesh configuration
istioctl analyze --all-namespaces

control plane


Istio maintainers with increasing complexity of the project that goes against the user friendliness still support helm manifest based configuration although there is fair movement towards the operator pattern. See below for differences, v1.6 and v1.7 still support both methods.


There are a few operational modes to configure control plane:

  • istio operator
    • install - istioctl install with --set or -f istio-overlay.yaml will be overlayed on top of a chosen profile
    • reconfigure - istioctl manifest apply -f istio-install.yaml --dry-run
    • istio operator init installs the operator pod istio-operator. Then using kubectl apply -f istio-overlay.yaml will trigger the operator to sync the changes. Also manually changing istio-system/istiooperators.install.istio.io/installed-state object will trigger an the operator event to sync the config and reconfigure teh control-plane.
  • [deprecated] istioctl manifest install with --set or -f <manifests.yaml>

Install

# Tested with 1.7.3 [depreciating method via helm charts], --set values are prefixed with 'values.'
istioctl manifest install --skip-confirmation --set profile=default \
  --set values.kiali.enabled=true      \
  --set values.prometheus.enabled=true \
  --dry-run

# Tested with 1.8.2 [istio.operator via 'istioctl' cli]
istioctl x precheck
istioctl install --skip-confirmation --set profile=default --dry-run # via operator
istioctl upgrade --skip-confirmation --set profile=default --dry-run # via operator


Using istio.operator via manifest yaml
Basic +Kiali,Prometheus +ServiceEntry, mesh traffic only +egress
<syntaxhighlightjs lang=yaml>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default

EOF ) </syntaxhighlightjs>

<syntaxhighlightjs lang=yaml>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default
 values:
   kiali:
     enabled: true
   prometheus:
     enabled: true

EOF ) </syntaxhighlightjs>

<syntaxhighlightjs lang=yaml>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default
 values:
   kiali:
     enabled: true
   # Kiali uses Prometheus to populate its dashboards
   prometheus:
     enabled: true
   meshConfig:
     # debugging
     accessLogFile: /dev/stdout
     outboundTrafficPolicy:
       mode: REGISTRY_ONLY

EOF ) </syntaxhighlightjs>

<syntaxhighlightjs lang=yaml>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default
 components:
   egressGateways:
   - enabled: true
     name: istio-egressgateway
 values:
   kiali:
     enabled: true
   prometheus:
     enabled: true
   meshConfig:
     accessLogFile: /dev/stdout
     outboundTrafficPolicy:
       mode: REGISTRY_ONLY

EOF ) </syntaxhighlightjs>

Uninstall

Uninstall v1.6.8, it's safe to ignore RBAC not existing resources.

istioctl manifest generate --set profile=default | kubectl delete --ignore-not-found=true -f -
kubectl delete namespace istio-system

Uninstall v1.7.x - view logs

# Removes istio-system resources and istio-operator
istioctl x uninstall --purge # cmd:experimental, aliases: experimental, x, exp
✔ Uninstall complete

Install - day 2 operation

Note: Update 2021-06-01 the guide focus on Istio v1.10


This should be a preferred method for day 2 operations - canary upgrade where control-plane gets installed with revisions in istio-system namespace and ingress-gateway(s) components are installed separately in another namespace eg. istio-ingress.


Pre checks
istioctl experimental precheck
istioctl analyze -A

# Optional labels for data plane operations
## Automatic proxy side injection
kubectl label namespace default istio-injection=enabled
kubectl label namespace default istio-injection=disabled
kubectl label namespace default istio.io/rev=1-10-0 # revision label

## Discovery namespace selectors v1.10+ !!!Warning labels should exist before enabling this feature otherwise you may break cp what can wach from api-server
kubectl label namespace istio-ingress istio-discovery=enabled # ingressgateway namespaces also need to be labelled (if in different ns than cp), otherwise the traffic won't be able to flow from ingress to our services
kubectl label namespace default       istio-discovery=enabled

Install control-plane

Note: For installing v1.9.x you should first install without revisions eg.REVISION="" this is because revisions upgrade expects to have default revision set.


Note: Due to a bug in the creation of the ValidatingWebhookConfiguration during install, initial installations of Istio must not specify a revision. A workaround is to run command below, where <REVISION> is the name of the revision that should handle validation. This command creates an istiod service pointed to the target revision.

kubectl get service -n istio-system -o json istiod-<REVISION> | jq '.metadata.name = "istiod" | del(.spec.clusterIP) | del(.spec.clusterIPs)' | kubectl apply -f -


Note: The difference in between operations install and upgrade is fuzzy. Upgrade safely upgrades components, and show diff in between the profile and what is being configured. On the other hand for eg. discoverySelectors feature only the install applies the config where using upgrade seems not having any effect.


REVISION="" # don't use revisions
REVISION='--revision 1-10-0'

istioctl install -y -n istio-system $REVISION -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
metadata:
  name: control-plane
  #namespace: istio-system # it's default
spec:
  profile: minimal # installs only control-plane
  hub: gcr.io/istio-release
  components:
    pilot:
      k8s:
        hpaSpec:
          minReplicas: 2
        env:
        - name: PILOT_FILTER_GATEWAY_CLUSTER_CONFIG # Reduce Gateway Config to services that have routing rules defined
          value: "true"
  meshConfig:
    defaultConfig:
      proxyMetadata:
        # https://preliminary.istio.io/latest/blog/2020/dns-proxy/
        # curl localhost:8080/debug/nsdz?proxyID=<POD-NAME>
        # dig @localhost -p 15053 postgres-stolon-proxy.sample.svc
        ISTIO_META_DNS_CAPTURE: "true"
    enablePrometheusMerge: true
    #discoverySelectors: # new in v1.10, choose which namespaces Istiod should watch (from api-server)
    #- matchLabels:      # | for building the internal mesh registry based on configurations of services,endpoints,deployments
                         # | dynamically restrict the set of namespaces that are part of the mesh
    #    istio-discovery: enabled # <- example label
    #    #env: prod
    #    #region: us-east-1
    #- matchExpressions:
    #  - key: app
    #    operator: In
    #    values:
    #    - cassandrs
    #    - spark
EOF
) --dry-run


Verify if there any clients are connected to Istio v1.10.0
kubectl create deployment sleep --image=curlimages/curl -- /bin/sleep 3650d
kubectl exec -it deploy/sleep -- curl istiod-1-10-0.istio-system:15014/debug/connections | jq .
{
  "totalClients": 0,
  "clients": null
}

# example of output when client(s) are conencted
{
  "totalClients": 1,
  "clients": [
    {
      "connectionId": "sidecar~172.17.0.8~sleep-76f59b4b9c-4sw6t.default~default.svc.cluster.local-3",
      "connectedAt": "2021-06-01T04:57:10.663650097Z",
      "address": "172.17.0.1:8415",
      "watches": null
    }
  ]
}

# Show all discovered endpoints
kubectl exec -it deploy/sleep -- curl istiod-1-10-0.istio-system:15014/debug/endpointz | jq . # ~1800+ lines for empty Minikube


Sidecar proxy automatic injection by telling a namespace which revision (aka control plane) to select/use.
kubectl label namespace default istio.io/rev=1-10-0 istio-injection-

Note: The istio-injection label must be removed because it takes precedence over the istio.io/rev label for backward compatibility.


Upgrade with Canary tag revision for isolated testing
# Tag the new control-plane version with 'canary' tag
istioctl experimental revision tag set canary --revision 1-10-0
Revision tag "canary" created, referencing control plane revision "1-10-0". To enable injection using this
revision tag, use 'kubectl label namespace <NAMESPACE> istio.io/rev=canary'

# Show tagged revision
istioctl experimental revision list 
REVISION TAG    ISTIO-OPERATOR-CR                                 PROFILE REQD-COMPONENTS
1-10-0   canary istio-system/installed-state-control-plane-1-10-0 minimal base
                                                                          istiod

# Update a namespace label to use the newly tagged control-plane for the injected sidecars to use it/connect to it
kubectl label namespace default istio.io/rev=canary --overwrite

# Restart deployments for new proxy to be injected, test the application and if all is good delete the "app canary deployment" and continue with full rollout of all deployments.

Install ingress-gateway

We install the ingress-gateway components separated in a separated namespace istio-ingress. This is to allow to upgrade control-plane and ingressgateway(s) components to be independent. We achieve this using 'minimal' profile that includes only control CP components. Then for IGW install we use 'empty' profile that adds only the ingress-gateway. See a list of what is included in each profile.

ClipCapIt-210317-090509.PNG


Note: tag and revision - in v1.10 ingressgateway is injected into the pod thus this becomes and behaves almost like any other workload. Thus the revision label should be in use eg. istio.io/rev=1-10-0 to control the proxy version and to which control-plane the ingressgateway connects to


# Create ingress own namespace
kubectl create ns istio-ingress

# Install revision
REVISION='--revision 1-10-0'
istioctl install -y -n istio-ingress $REVISION -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
metadata:
  name: istio-ingressgateway
spec:
  profile: empty # uses the 'empty' profile and enables the istio-ingressgateway component
  hub: gcr.io/istio-release
  components:
    ingressGateways:
    - name: istio-ingressgateway
      namespace: istio-ingress # if not set defaults to 'istio-system'
      enabled: true
      k8s:
        hpaSpec:
          minReplicas: 2
        service:
          type: LoadBalancer
        overlays:
        - apiVersion: apps/v1
          kind: Deployment
          name: istio-ingressgateway
          patches:
          - path: spec.template.spec.containers[name:istio-proxy].lifecycle
            value:
              preStop:
                exec:
                  command: ["sh", "-c", "sleep 5"]
EOF
) --dry-run

# Verify
# | Note that 'revision' relates to control-plane thus ingressgateway 'installed-state-istio-ingressgateway-1-10-0' installation
# | does hot have revision set

## Show IstioOperator installations
kubectl -n istio-system get istiooperators.install.istio.io 
NAME                                          REVISION   STATUS   AGE
installed-state-control-plane-1-10-0          1-10-0              165m
installed-state-istio-ingressgateway-1-10-0   1-10-0              8m47s

## Show revisions
istioctl experimental revision list 
REVISION TAG    ISTIO-OPERATOR-CR                                        PROFILE REQD-COMPONENTS
1-10-0   canary istio-system/installed-state-control-plane-1-10-0        minimal base
                                                                                 istiod
                istio-system/installed-state-istio-ingressgateway-1-10-0 empty   ingress:istio-ingressgateway

## Test
GATEWAY_IP=$(kubectl get svc -n istio-system istio-ingressgateway -o jsonpath="{.status.loadBalancer.ingress[0].ip}")
# With Gateway and VirtualService we can route traffic to a destination service via IngressGateway
curl -H "Host: example.com" http://$GATEWAY_IP


Note: As the next step IngressGateway can be integrated with Cert Manager


Reduce Gateway Config
# Show what services IngressGateway has discovered, the list has even those that no routing is defined
istioctl pc clusters deploy/istio-ingressgateway -n istio-system

# Set PILOT_FILTER_GATEWAY_CLUSTER_CONFIG environment variable in the istiod deployment
#  kind: IstioOperator
#  spec:
#    components:
#      pilot:
#        k8s:
#          env:                                                                            
#          - name: PILOT_FILTER_GATEWAY_CLUSTER_CONFIG                                                  
#            value: "true"


(Optional) Access logging for gateway only using EnvoyFilter
  • Istio docs explain to enable Envoy's Access Logs for the entire service mesh
kubectl apply -f <(cat <<EOF
apiVersion: networking.istio.io/v1alpha3
kind: EnvoyFilter
metadata:
  name: ingressgateway-access-logging
  namespace: istio-system
spec:
  workloadSelector:
    labels:
      istio: ingressgateway
  configPatches:
  - applyTo: NETWORK_FILTER
    match:
      context: GATEWAY
      listener:
        filterChain:
          filter:
            name: "envoy.filters.network.http_connection_manager"
    patch:
      operation: MERGE
      value:
        typed_config:
          "@type": "type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager"
          access_log:
          - name: envoy.access_loggers.file
            typed_config:
              "@type": "type.googleapis.com/envoy.extensions.access_loggers.file.v3.FileAccessLog"
              path: /dev/stdout
              format: "[%START_TIME%] \"%REQ(:METHOD)% %REQ(X-ENVOY-ORIGINAL-PATH?:PATH)% %PROTOCOL%\" %RESPONSE_CODE% %RESPONSE_FLAGS% \"%UPSTREAM_TRANSPORT_FAILURE_REASON%\" %BYTES_RECEIVED% %BYTES_SENT% %DURATION% %RESP(X-ENVOY-UPSTREAM-SERVICE-TIME)% \"%REQ(X-FORWARDED-FOR)%\" \"%REQ(USER-AGENT)%\" \"%REQ(X-REQUEST-ID)%\" \"%REQ(:AUTHORITY)%\" \"%UPSTREAM_HOST%\" %UPSTREAM_CLUSTER% %UPSTREAM_LOCAL_ADDRESS% %DOWNSTREAM_LOCAL_ADDRESS% %DOWNSTREAM_REMOTE_ADDRESS% %REQUESTED_SERVER_NAME% %ROUTE_NAME%\n"
EOF
) --dry-run=server


Check status, list and describe revisions
kubectl get -n istio-system istiooperators.install.istio.io 
NAME                                          REVISION   STATUS   AGE
installed-state-control-plane                                     22m
installed-state-control-plane-1-10-0          1-10-0              17h
installed-state-istio-ingressgateway-1-10-0   1-10-0              15h

# List revisions
istioctl x revision list 
REVISION TAG      ISTIO-OPERATOR-CR                                        PROFILE REQD-COMPONENTS
1-10-0   canary   istio-system/installed-state-control-plane-1-10-0        minimal base
                                                                                   istiod
                  istio-system/installed-state-istio-ingressgateway-1-10-0 empty   ingress:istio-ingressgateway
default  <no-tag> istio-system/installed-state-control-plane               minimal base
                                                                                   istiod

# Describe revisions
istioctl x revision describe 1-10-0

ISTIO-OPERATOR CUSTOM RESOURCE: (2)
1. istio-system/installed-state-control-plane-1-10-0
  COMPONENTS:
  - base
  - istiod
  CUSTOMIZATIONS:
  - values.global.jwtPolicy=first-party-jwt
  - hub=gcr.io/istio-release

2. istio-system/installed-state-istio-ingressgateway-1-10-0
  COMPONENTS:
  - ingress:istio-ingressgateway
  CUSTOMIZATIONS:
  - components.ingressGateways[0].enabled=true
  - values.global.jwtPolicy=first-party-jwt
  - hub=gcr.io/istio-release

MUTATING WEBHOOK CONFIGURATIONS: (2)
WEBHOOK                       TAG
istio-revision-tag-canary     canary
istio-sidecar-injector-1-10-0 <no-tag>

CONTROL PLANE PODS (ISTIOD): (2)
NAMESPACE    NAME                           ADDRESS    STATUS  AGE
istio-system istiod-1-10-0-7c5b7865df-dcd92 172.17.0.5 Running 18h
istio-system istiod-1-10-0-7c5b7865df-xh5r9 172.17.0.3 Running 18h

INGRESS GATEWAYS: (0)
Ingress gateway is enabled for this revision. However there are no such pods. It could be that it is replaced by ingress-gateway from another revision (as it is still upgraded in-place) or it could be some issue with installation

EGRESS GATEWAYS: (0)
Egress gateway is disabled for this revision

Note: Ingress gateway is enabled for this revision. However there are no such pods. - this is a bit blury message


Uninstall
# This command uninstalls all revisions in default namespace of 'istio-system'
istioctl x uninstall --purge # uninstall all
istioctl x uninstall -f <(cat <<EOF
...
EOF
)


Install book info application

# Namespaces
kubectl create ns bookstore-1
kubectl create ns bookstore-2
kubectl label namespace bookstore-1 istio-injection=enabled
kubectl label namespace bookstore-2 istio.io/rev=1-9-1 istio-injection-

kubectl get ns --show-labels 
NAME              STATUS   AGE    LABELS
bookstore-1       Active   155m   istio-injection=enabled
bookstore-2       Active   155m   istio.io/rev=1-10-0

# Deploy application
kubectl apply -f samples/bookinfo/platform/kube/bookinfo.yaml -n bookstore-1
kubectl apply -f samples/bookinfo/platform/kube/bookinfo.yaml -n bookstore-2

istioctl proxy-status --revision 1-10-0 # show only proxies of revision '1-10-0'
istioctl proxy-status                  # show all proxies


You can see proxies connected only to revision '1-9-1' and at the bottom all proxies

ClipCapIt-210317-134114.PNG

Note: Screen grab from previous Istio v1.9 version

Ingress flow from ALB to the service

This example graph shows the traffic flow from ALB -> a container

ClipCapIt-210828-101105.PNG

Get info

Profiles are istiooperators.install.istio.io CRD manifests located in istio-1.7.3/manifests/profiles

# List profiles
istioctl profile list
ls istio-1.7.3/manifests/profiles
default.yaml  demo.yaml  empty.yaml  minimal.yaml  preview.yaml  remote.yaml

# profile configuration
istioctl profile dump demo
istioctl profile dump --config-path components.pilot demo

# Differences in the profiles
istioctl profile diff default demo


Available dashboards commands

istioctl dashboard <dashboard>
Available Commands:
  controlz    Open ControlZ web UI
  envoy       Open Envoy admin web UI
  grafana     Open Grafana web UI
  jaeger      Open Jaeger web UI
  kiali       Open Kiali web UI
  prometheus  Open Prometheus web UI
  zipkin      Open Zipkin web UI


List all Istio CRDs

k get crd -A | grep istio | cut -f1 -d" "
adapters.config.istio.io
attributemanifests.config.istio.io
authorizationpolicies.security.istio.io
destinationrules.networking.istio.io
envoyfilters.networking.istio.io
gateways.networking.istio.io
handlers.config.istio.io
httpapispecbindings.config.istio.io
httpapispecs.config.istio.io
instances.config.istio.io
istiooperators.install.istio.io
peerauthentications.security.istio.io
quotaspecbindings.config.istio.io
quotaspecs.config.istio.io
requestauthentications.security.istio.io
rules.config.istio.io
serviceentries.networking.istio.io
sidecars.networking.istio.io
templates.config.istio.io
virtualservices.networking.istio.io
workloadentries.networking.istio.io

IstioOperator

Output logs of istio-operator Kubernetes/Istio-logs-default-install when installing default istio+kiali=prometheus.

Initialize istio.operator

Note: This is being unfolded, check out here the second paragraph.


istioctl operator init # dump, remove, --force
Using operator Deployment image: docker.io/istio/operator:1.7.3
✔ Istio operator installed                                                                                                                                     
✔ Installation complete

# The operator deployment gets installed in a new namespace 'istio-operator'
kubectl get deployment -n istio-operator  -owide
NAME             READY   UP-TO-DATE   AVAILABLE   AGE     CONTAINERS       IMAGES                           SELECTOR
istio-operator   1/1     1            1           7m14s   istio-operator   docker.io/istio/operator:1.7.3   name=istio-operator

kubectl get all -n istio-operator 
NAME                                 READY   STATUS    RESTARTS   AGE
pod/istio-operator-9dc6b7fd7-xf2m5   1/1     Running   0          5m42s
NAME                     TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
service/istio-operator   ClusterIP   10.119.252.73   <none>        8383/TCP   5m43s
NAME                             READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/istio-operator   1/1     1            1           5m43s
NAME                                       DESIRED   CURRENT   READY   AGE
replicaset.apps/istio-operator-9dc6b7fd7   1         1         1       5m42s


Istio operator (v1.7.3), this is not super clear to me

find . -iname *operator*
./samples/operator
./samples/addons/extras/prometheus-operator.yaml
./manifests/charts/istio-operator # it's a chart to deploy operator
./manifests/charts/istio-operator/crds/crd-operator.yaml
./manifests/charts/base/crds/crd-operator.yaml
./manifests/charts/istio-telemetry/prometheusOperator
./manifests/deploy/crds/istio_v1alpha1_istiooperator_crd.yaml
./manifests/deploy/crds/istio_v1alpha1_istiooperator_cr.yaml
./manifests/deploy/operator.yaml # this [1]
./manifests/examples/customresource/istio_v1alpha1_istiooperator_cr.yaml


  • [1] manifests/deploy/operator.yaml

<syntaxhighlightjs lang=yaml> --- apiVersion: apps/v1 kind: Deployment metadata:

 namespace: istio-operator
 name: istio-operator

spec:

 replicas: 1
 selector:
   matchLabels:
     name: istio-operator

... </syntaxhighlightjs>


This API get always installed

k -n istio-system get istiooperators.install.istio.io # kind: IstioOperator

CRDs aka new kind: objects - demystified

  • Istio terminology
    • Use workload over service that can be misinterpreted by some
  • Envoy terminiology Host, Downstream, Upstream, Listener, Cluster, Mesh, Runtime configuration
    • Cluster: A cluster is a group of logically similar upstream hosts that Envoy connects to, eg. group of pods reviews service v1, v2 and v3.
    • Cluster Discovery Service (CDS)
    • Listener discovery service (LDS)
    • Endpoint discovery service (EDS); cluster members are called endpoint.
    • Route Discovery Service (RDS)


  • VirtualServices and DestinationRules are key resources for configuring Istio’s traffic routing functionality.
    • VirtualServices - is used to configure how requests are routed to a service within an Istio service mesh, define how traffic is routed to a given destination
    • DestinationRules define policies that apply to traffic intended for a service after routing has occurred. A DestinationRules is applied after VirtualServices routing rules are evaluated, so they apply to the traffic’s real destination.
      • subsets - defines a set of pods with common matching label eg. version: v1


  • gateways,gw - taps on to ingressgateway; describe which controller to use eg. plugs into real network received by service/istio-ingressgateway that has label istio=ingressgateway
  • virtualservices,vs - it bounds to a gateway(s), and describes how requests are routed to a service within the mesh
  • destinationrules,dr - applied after vs routing rules to real traffic destination, defines available versions called subsets, they name (using subsets) different revisions available and contain a logic (selector) of what corresponds to each revision.
  • serviceentries,se - holds a list of all endpoints services that they belong to, adds an entry into Istio’s internal service registry, then the Envoy proxies can send traffic to the specified host as if it was a service in the mesh. Allows traffic to be managed for services running outside of the mesh,can set redirect, forward, retry, timeout and fault injection policies fro external destinations. Two purposes:
    • used as external service
    • allows multi-cluster services


  • adapters
  • attributemanifests
  • authorizationpolicies - defines policies of what service can access what workloads
  • envoyfilters
  • handlers
  • httpapispecbindings
  • httpapispecs
  • instances
  • istiooperators
  • peerauthentications - defines what authenticated traffic within the cluster is allowed to access what workloads; used for service-to-service authentication to verify the client making the connection
  • quotaspecbindings
  • quotaspecs
  • requestauthentications - it's for humans outside of the cluster to validate their identity and decide to allow enter the cluster
  • rules
  • sidecars
  • templates
  • workloadentries

Customize istio installation

Configure ingress-gateways

Gateways are a special type of component, since multiple ingress and egress gateways can be defined. In the IstioOperator API, gateways are defined as a list type. The default profile installs one ingress gateway, called istio-ingressgateway.


Ingress Gateway
describes a load balancer operating at the edge of the mesh that receives incoming HTTP/TCP connections.
Egress Gateway
An egress gateway configures a dedicated exit node for the traffic leaving the mesh, limiting which services can or should access external networks, or to enable secure control of egress traffic to add security to the mesh.
ClipCapIt-201028-064439.PNG


Show default values of the ingressgateway

istioctl profile dump --config-path components.ingressGateways
istioctl profile dump --config-path values.gateways.istio-ingressgateway


Install istio with ingressgateway servioce as internal AWS loadbalancer:

istioctl install \
   --set profile=default \
   --set components.ingressGateways[0].enabled="true" \
   --set components.ingressGateways[0].k8s.serviceAnnotations."service\.beta\.kubernetes\.io/aws-load-balancer-internal"=\"true\"
✔ Istio core installed                                                                                                                                         
✔ Istiod installed                                                                                                                                             
✔ Ingress gateways installed                                                                                                                                   
✔ Installation complete

# --set gateways.istio-ingressgateway.serviceAnnotations."service\.beta\.kubernetes\.io/aws-load-balancer-internal"="0\.0\.0\.0/0"


Configure using IstioOperator

istioctl manifest install -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  profile: default
  components:
    ingressGateways:
    - enabled: true # default is true for default profile
      name: istio-ingressgateway # required key
    egressGateways:
    - enabled: true
      name: istio-egressgateway # required key
EOF
) --dry-run

Ingress Gateways - controller - get details

# manually inject the sidecar
kubectl -n bin apply -f <(istioctl kube-inject -f httpbin.yaml)

export INGRESS_HOST=$(       kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.status.loadBalancer.ingress[0].ip}')
# AWS, uses 'hostname'
export INGRESS_HOST=$(       kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.status.loadBalancer.ingress[0].hostname}')

export INGRESS_PORT=$(       kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.spec.ports[?(@.name=="http2")].port}')
export SECURE_INGRESS_PORT=$(kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.spec.ports[?(@.name=="https")].port}')

# Optional
export TCP_INGRESS_PORT=$(   kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.spec.ports[?(@.name=="tcp")].port}')

# Verify
env | grep INGRESS

Ingressgateway Kubernetes service

The istio istio-ingressgateway works as any other ingress controller, it has service type: LoadBalancer, that creates an endpoint with 4 ports, these in turn make external-loadbalancer to listen on these ports:

Port  Name        Protocol
8443  https       TCP
15443 tls         TCP
15021 status-port TCP
8080  http2       TCP

so, any incoming traffic via the ext-lb will end up being processed by istio-egressgateway-54884d6c57-mqgbg pod according to it's configuration that in reality is Envoy config built by pilot using using Istio CRDs.


Describe the service and the endpoint

vagrant@vagrant:~$ k -n istio-system describe svc istio-ingressgateway 
Name:                     istio-ingressgateway
Namespace:                istio-system
Labels:                   app=istio-ingressgateway
...
Selector:                 app=istio-ingressgateway,istio=ingressgateway
Type:                     LoadBalancer
IP:                       10.119.253.21
LoadBalancer Ingress:     34.76.42.198
Port:                     status-port  15021/TCP
TargetPort:               15021/TCP
NodePort:                 status-port  32231/TCP
Endpoints:                10.116.19.12:15021
Port:                     http2  80/TCP
TargetPort:               8080/TCP
NodePort:                 http2  31604/TCP
Endpoints:                10.116.19.12:8080
Port:                     https  443/TCP
TargetPort:               8443/TCP
NodePort:                 https  30747/TCP
Endpoints:                10.116.19.12:8443
Port:                     tls  15443/TCP
TargetPort:               15443/TCP
NodePort:                 tls  31901/TCP
Endpoints:                10.116.19.12:15443

vagrant@vagrant:~$ k -n istio-system describe ep istio-ingressgateway 
Name:         istio-ingressgateway
Namespace:    istio-system
Labels:       app=istio-ingressgateway
...
Annotations:  <none>
Subsets:
  Addresses:          10.116.19.12 # -> target: istio-ingressgateway-86f88b6f6-7mzf7
  NotReadyAddresses:  <none>
  Ports:
    Name         Port   Protocol
    ----         ----   --------
    https        8443   TCP
    tls          15443  TCP
    status-port  15021  TCP
    http2        8080   TCP


General traffic flow

Note: This overview applies to each workload including Ingress-gateways

The arrows in the diagram show the flow of a request through the configuration, and the five key elements are the ‘listener,’ ‘filter chains,’ ‘routes,’ ‘clusters,’ and ‘endpoints’. The route is part of the filter chain, which is part of the listener.

ClipCapIt-210503-164443.PNG


  • listener - 'binds' to a port and listens for inbound requests. Any request that comes in via another port would not be seen or handled by Envoy. In Envoy terms, it manages downstream requests
  • filter chain - after being accepted by 'the listener', describes how the request should be handled once it’s entered Envoy
  • route - extension of the filter chain, takes the HTTP request information and directs it to the correct service
  • cluster - address IP:port that are the backend for a service. A cluster is a group of logically similar upstream hosts that Envoy connects to. Envoy discovers the members of a cluster via CDS(Cluster Discovery Service).
  • endpoints - network nodes that implement a logical service. They are grouped into clusters. Endpoints in a cluster are upstream of an Envoy proxy.

Ingress-gateway - routes verification

INGRESS_GATEWAY_LABEL=ingressgateway # ingressgateway-internal|external
INGRESS_GATEWAY_POD=$(kubectl -n istio-system get pod  -l istio=$INGRESS_GATEWAY_LABEL -o jsonpath='{.items[0].metadata.name}')

# Get listeners, other options: bootstrap,cluster,endpoint,listener,log,route,secret
istioctl proxy-config listeners -n istio-system $INGRESS_GATEWAY_POD
ADDRESS PORT  MATCH DESTINATION
0.0.0.0 8080  ALL   Route: http.80
0.0.0.0 15021 ALL   Inline Route: /healthz/ready*
0.0.0.0 15090 ALL   Inline Route: /stats/prometheus*

# Show routes, and virtual-services
istioctl proxy-config routes -n istio-system $INGRESS_GATEWAY_POD
NOTE: This output only contains routes loaded via RDS.
NAME        DOMAINS                                                         MATCH                  VIRTUAL SERVICE
http.80     alertmanager.staging.net                     /*                     alertmanager.monitoring
http.80     grafana.staging.net                          /*                     grafana.monitoring
http.80     prometheus.staging.net                       /*                     prometheus.monitoring
            *                                                               /healthz/ready*        
            *                                                               /stats/prometheus*

Object examples

Gateway, virtualservice

# istio-1.7.3/samples/bookinfo/networking/bookinfo-gateway.yaml
---
# Entry point gateway
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
  name: bookinfo-gateway
spec:
  selector:
    istio: ingressgateway # use istio default controller
  servers:
  - port:
      number: 80
      name: http
      protocol: HTTP
    hosts:
    - "*"
---
# Routes traffic to destination, could use subset for Traffic Shifting
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: bookinfo
spec:
  hosts:
  - "*"
  gateways:
  - bookinfo-gateway
  http:
  - match:
    - uri:
        exact: /productpage
    - uri:
        prefix: /static
    - uri:
        exact: /login
    - uri:
        exact: /logout
    - uri:
        prefix: /api/v1/products
    route:
    - destination:
        host: productpage
        port:
          number: 9080
       #subset: v1 # [1] if defined in DestinationRule for the host ('productpage'), it will route to any pods
                   # matching a label(s) defined in the subset

# samples/bookinfo/networking/destination-rule-all.yaml
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: productpage
spec:
  host: productpage
  subsets:   # pod selectors matching labels
  - name: v1 # [1] subset name for the host 'productpage'
    labels:
      version: v1
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: reviews
spec:
  host: reviews
  subsets:
  - name: v1
    labels:
      version: v1
  - name: v2
    labels:
      version: v2
  - name: v3
    labels:
      version: v3
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: ratings
spec:
  host: ratings
  subsets:
  - name: v1
    labels:
      version: v1
  - name: v2
    labels:
      version: v2
  - name: v2-mysql
    labels:
      version: v2-mysql
  - name: v2-mysql-vm
    labels:
      version: v2-mysql-vm
---
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:
  name: details
spec:
  host: details
  subsets:
  - name: v1
    labels:
      version: v1
  - name: v2
    labels:
      version: v2
---
# Split traffic between reviews:v1 and reviews:v3
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: reviews
spec:
  hosts:
    - reviews
  http:
  - route:
    - destination:
        host: reviews
        subset: v1
      weight: 50
    - destination:
        host: reviews
        subset: v3
      weight: 50

ServiceEntry

It intercepts any traffic from pods in the mesh (pods having envoy proxy installed) that matches the spec.hosts then applies meshConfig.outboundTrafficPolicy policy ALLOW_ANY or REGISTRY_ONLY plus additional features like retry, timeout etc. if configured.


It depends on installation option meshConfig.outboundTrafficPolicy.mode that configures Envoy's handling of unknown services, that is, services that are not defined in Istio’s internal service registry:

  • ALLOW_ANY by default, so may not be defined in the configMap/istio, tells Envoy to let calls to unknown services pass through, this also means that Istio's capabilities cannot be applied to these endpoints
  • REGISTRY_ONLY tells Envoy to block any connections to endpoints without a registry entry
kubectl get configmap istio -n istio-system -o yaml | grep mode -m1 -B1


Set REGISTRY_ONLY policy

istioctl manifest install -f <(cat <<EOF
apiVersion: install.istio.io/v1alpha1
kind: IstioOperator
spec:
  profile: default
  values:
    meshConfig:
      accessLogFile: /dev/stdout
      outboundTrafficPolicy:
        mode: REGISTRY_ONLY
EOF
)


Example MESH_EXTERNAL service entry adds the ext-svc external dependency to Istio’s service registry. Required when REGISTRY_ONLY policy is set.

apiVersion: networking.istio.io/v1alpha3
kind: ServiceEntry
metadata:
  name: google
spec:
  hosts:
  - www.google.com # FQDN of external resource or wildcard prefix
  ports:
  - number: 443
    name: https
    protocol: HTTPS
  location: MESH_EXTERNAL # treat the service as external part of the mesh
  resolution: DNS         # | MESH_INTERNAL - Treat remote cluster services as part of the service mesh
                          # | as all clusters in the service mesh share the same root of trust.

peerauthentications and requestauthentications

peerauthentications
defines what authenticated traffic within the cluster is allowed to access what workloads; used for service-to-service authentication to verify the client making the connection. mutualTLS is used for transport authentication, Authentication policies apply to requests that a service receives.
apiVersion: "security.istio.io/v1beta1"
kind: "PeerAuthentication"
metadata:
  name: "example-peer-policy"
  namespace: "foo"
spec:
  selector:
    matchLabels:
      app: reviews
  mtls:
    mode: STRICT # PERMISSIVE, DISABLE


requestauthentications
it's for humans outside of the cluster to validate their identity and decide to allow enter the cluster, used for end-user authentication to verify credentials attached to the request.
apiVersion: "security.istio.io/v1beta1"
kind: "RequestAuthentication"
metadata:
  name: "jwt-example"
  namespace: istio-system
spec:
  selector:
    matchLabels:
      istio: ingressgateway # requires end-user JWT for ingress gateway
  jwtRules:
  - issuer: "testing@secure.istio.io"
  jwksUri: "https://raw.githubusercontent.com/istio/istio/release-1.5/..
  # ^ it's authentication token not authorization is being evaluated in this policy

authorizationpolicies

These can be compare to Kubernetes NetworkPolicies but much more powerful as being aware of L7 and not only packet level traffic.


authorizationpolicies
defines policies of what service can access what workloads; authorisation features provides mesh-level, namespace-level, and workload-level access control on workloads in an Istio Mesh.

There is no need to explicitly enable Istio’s authorisation feature, only the AuthorizationPolicy needs applying on workloads to enforce access control.

If no AuthorizationPolicy applies to a workload, no access control will be enforced, In other words, all requests will be allowed. If any AuthorizationPolicy applies to a workload, access to that workload is denied by default, unless explicitly allowed by a rule declared in the policy.

# allow-all policy full access to all workloads in the default namespace
apiVersion: security.istio.io/v1beta1
kind: AuthorizationPolicy
metadata:
  name: allow-all
  namespace: default
spec:
  rules:
  - {}
---
# deny-all policy to all workloads in admin namespace
apiVersion: security.istio.io/v1beta1
kind: AuthorizationPolicy
metadata:
  name: deny-all
  namespace: admin
spec:
  {}


Policies can distinguish in between authenticated and not authenticated users/services by specifying rules. There are reach configuration options so you may want to see docs.

code

spec:

 rules:
 - from:
   - source:
       principals: ["cluster.local/ns/default/sa/sleep"]
   - source:
       namespaces: ["dev"]

</syntaxhighlight>

Addons and integrations

Note: In v1.7.x the installation of telemetry addons by istioctl is deprecated, these are being managed now using addons integrations.

Kiali visualisations


Note: Example needs redoing. Istio v1.6.8 requires kiali secret, v1.7.x by default no authentication is enabled

! values.kiali.enabled is deprecated; use the samples/addons/ deployments instead
! values.prometheus.enabled is deprecated; use the samples/addons/ deployments instead
! addonComponents.kiali.enabled is deprecated; use the samples/addons/ deployments instead
! addonComponents.prometheus.enabled is deprecated; use the samples/addons/ deployments instead

You install kiali manually from scratch or pass arg to istioctl

# Create user name and password
KIALI_USERNAME=admin
KIALI_PASSPHRASE=admin
kubectl apply -f <(cat <<EOF
apiVersion: v1
kind: Secret
metadata:
  name: kiali
  namespace: istio-system
  labels:
    app: kiali
stringData:
  username: $KIALI_USERNAME
  passphrase: $KIALI_PASSPHRASE
EOF
)


Install - Option 1
cd istio-1.#.#/samples/addons
kubectl apply -f kiali.yaml


Deprecated installation options

Option 2 - istioctl - deprecated in v1.8

istioctl manifest install \
  --set values.kiali.enabled=true \
  --set values.prometheus.enabled=true


Option 3 - IstioOperator manifest, this is desired configuration defaults are set is not specified, then installed what is set, any resources not default and not specified will be pruned. Deprecated in v1.8.

IstioOperator manifest to install Kiali
Basic No auth
<syntaxhighlightjs lang=bash>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default
 values:
   kiali:
     enabled: true
   prometheus:
     enabled: true

EOF ) </syntaxhighlightjs>

<syntaxhighlightjs lang=bash>

istioctl manifest install -f <(cat <<EOF apiVersion: install.istio.io/v1alpha1 kind: IstioOperator spec:

 profile: default
 meshConfig:
   # For debugging
   accessLogFile: /dev/stdout
 addonComponents:
   # Kiali uses Prometheus to populate its dashboards
   prometheus:
     enabled: true
   kiali:
     enabled: true
     namespace: kiali
     k8s:
       podAnnotations:
         sidecar.istio.io/inject: "true"
 values:
   kiali:
     prometheusAddr: http://prometheus.istio-system:9090
     dashboard:
       auth:
         strategy: anonymous

EOF ) </syntaxhighlightjs>


Verify installation and access the console

kubectl wait --for=condition=Available deployment/kiali -n istio-system --timeout=300s

# Access the dashboard
istioctl dashboard kiali

Add custom headers

apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
  name: httpbin
spec:
  hosts:
  - "*"
  gateways:
  - httpbin-gateway
  http:
  - match:
    - uri:
        prefix: /headers
    route:
    - destination:
        port:
          number: 8000
        host: httpbin
      headers:
        response:            # add to response
          add:
           "key1": "abc"
        request:             # add to request
           add:
             "key2": "def"

Handling Istio Sidecars in Kubernetes Jobs

A Job/CronJob is not considered complete until all containers have stopped running, and Istio Sidecars run indefinitely.

Option 1: Disabling Istio Sidecar injection

spec:
  template:
    metadata:
      annotations:
        sidecar.istio.io/inject: 'false'


Option 2: Use `pkill` to stop the Istio Process

Add pkill -f /usr/local/bin/pilot-agent inside Dockerfile command and shareProcessNamespace: true to the Pod spec (not the Job spec).


Istio 1.3 and Core k8s Support (unverified)

<job shell script aka command:>
curl -X POST http://127.0.0.0/quitquitquit

Debugging


Analize

istioctl analyze --all-namespaces
istioctl analyze samples/bookinfo/networking/*.yaml
istioctl analyze --use-kube=false samples/bookinfo/networking/*.yaml

# Enable live analyzer, it will add to the status field eg.<code>status.validationMessages.type.code: IST0101</code>
istioctl install --set values.global.istiod.enableAnalysis=true


Debug a singular pod or a deployment pods

apiVersion: v1
kind: Namespace
metadata:
  name: redis
  labels:
    istio.io/rev: asm-173-6
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: debug-pod
  namespace: redis
  labels:
    app: debug-pod
spec:
  replicas: 1
  selector:
    matchLabels:
      app: debug-pod
  template:
    metadata:
      labels:
        app: debug-pod
      annotations:
        sidecar.istio.io/logLevel: debug # <- enable debug mode
    spec:
      containers:
        - image: samos123/docker-toolbox
          name: debug-pod


Istiocli debug commands

istioctl experimental describe pod <pod-name>[.<namespace>]

istioctl proxy-status
# | SYNCED means that Envoy has acknowledged the last configuration Istiod has sent to it.
# | NOT SENT means that Istiod hasn’t sent anything to Envoy. This usually is because Istiod has nothing to send.
# | STALE means that Istiod has sent an update to Envoy but has not received an acknowledgement. This usually indicates a networking issue between Envoy and Istiod or a bug with Istio itself.

# Retrieve information about cluster configuration for the Envoy instance in a specific pod:
istioctl proxy-config cluster <pod-name> [flags]

# To retrieve information about bootstrap configuration for the Envoy instance in a specific pod:
istioctl proxy-config bootstrap <pod-name> [flags]

# To retrieve information about listener configuration for the Envoy instance in a specific pod:
istioctl proxy-config listener <pod-name> [flags]

# To retrieve information about route configuration for the Envoy instance in a specific pod:
istioctl proxy-config route <pod-name> [flags]

# To retrieve information about endpoint configuration for the Envoy instance in a specific pod:
istioctl proxy-config endpoints <pod-name> [flags]
istioctl proxy-config endpoints productpage-v1-6c886ff494-7vxhs --cluster "outbound|9080||reviews.default.svc.cluster.local"

Istio helm starter

Project by Salesforce to scaffold a helm chart with Istio that uses helm-starter plugin.

# Install `helm-starter` helm plugin
helm plugin install https://github.com/salesforce/helm-starter.git

# Fetch starter templates: mesh-service, ingress-service, mesh-egress, auth-policy
helm starter fetch https://github.com/salesforce/helm-starter-istio.git

# Create a helm chart from the starter
helm create NAME --starter helm-starter-istio/[starter-name]
for starter in mesh-service ingress-service mesh-egress auth-policy; do helm create authservice-$starter --starter helm-starter-istio/$starter; done

Important labels

Deployment <syntaxhighlightjs lang="yaml"> apiVersion: apps/v1 kind: Deployment spec:

 selector:
   matchLabels:
     app: Template:.Values.service # <- Istio label
     version: Template:.Values.version     # <- Istio label
 template:
   metadata:
     labels:
       # Kubernetes recommended labels
       app.kubernetes.io/name: Template:.Values.service
       app.kubernetes.io/part-of: Template:.Values.system
       app.kubernetes.io/version: Template:.Values.version
       # Isio required labels
       app: Template:.Values.service # <- Istio label
       version: Template:.Values.version     # <- Istio label

Template:- if .Values.configMap

     annotations:
       checksum/config: Template:Include (print $.Template.BasePath "/configmap.yaml") .

Template:- end

   spec:
     containers:

Template:- range .Values.ports

           - name: Template:.name
             containerPort: Template:.targetPort
             protocol: TCP # <-- TCP|HTTP or any other that is accepted

Template:- end </syntaxhighlightjs>


Service <syntaxhighlightjs lang="yaml"> apiVersion: v1 kind: Service spec:

 ports:

Template:- range .Values.ports

   - name: Template:.name # <- tcp-<name>, http-<name>, grcp-<name>
     port: Template:.port
     targetPort: Template:.targetPort
     protocol: TCP # <- TCP|HTTP or any other that is accepted

Template:- end </syntaxhighlightjs>

Envoy

  • Upstream connections are the service Envoy is initiating the connection to.
  • Downstream connections are the client that is initiating a request through Envoy.
ClipCapIt-210525-080849.PNG

Terminology in my own words:

Listeners
Declaration of ports to open on the proxy and listen for incoming connections
Clusters
Backend service to which we can route traffic. Also a group of logically similar upstream hosts that Envoy connects to
Envoy Discovery (XDS)

Envoy is driven by a set of APIs that configure certain aspects of the proxy. We saw earlier how we specified clusters, listeners, and routes. We can also configure those over Envoy's xDS APIs:

  • Listener discovery service (LDS)
  • Route discovery service (RDS)
  • Cluster discovery service (CDS)
  • Service discovery service (SDS/EDS)
  • Aggregated discovery service (ADS)


Now we don't have to specify listeners ahead of time. We can open them up or close them at runtime. Below example of Envoy config for dynamic Listeners configuration, set/received from api_config_source. <syntaxhighlightjs lang="yaml"> dynamic_resources:

   lds_config:   
   api_config_source:
       api_type: GRPC
       grpc_services:
       - envoy_grpc:
           cluster_name: xds_cluster

</syntaxhighlightjs> Configuring Envoy through this API is exactly what Istio's control plane does. With Istio we specify configurations in a more user-friendly format and Istio translates that configuration into something Envoy can understand and delivers this configuration through Envoy's xDS API.

Service-to-service communication

ClipCapIt-210324-180344.PNG


This is zoom in on inbound connection from the sleep to http pod.

Istio


.

Troubleshooting

# Interrogate envoy stats endpoint 
kubectl exec -it deploy/sleep -- curl http://envoy:15000/stats| grep retry

Troubleshooting istio

Grab whole mesh config (new in v1.10) and paste into https://envoyui.solo.io/ analyzer

istioctl proxy-config all deploy/sleep -n default -o json > sleep.deploy.all-config_dump.json


Check istiod endpoints, get registry

# Get all services in the Istio registry, workloads are also included even if they are not officially part of the mesh
kubectl exec -n istio-system -it deploy/istiod-1-8-4 -- pilot-discovery request GET /debug/registryz
kubectl exec -n istio-system -it deploy/istiod-1-8-4 -- pilot-discovery version # [discovery,help,request,version]


Debug info from istiod control-plane

# istiod has istiod:8080/debug endpoint 'internal-debug' enables this using secure channel
## curl /debug endpoint
kubectl -n istio-system port-forward deploy/istiod 8080
firefox http://127.0.0.1:8080/debug      # all debug endpoints
firefox http://127.0.0.1:8080/debug/mesh # show mesh config

## internal-debug (not working)
istioctl x internal-debug syncz


Verify an app/pod iptables modified by istio-proxy. This rules are in the Network kernel namespace hold by the pod.

# List iptables rules - this should fail
kubectl -n default exec -it deploy/httpbin -c istio-proxy -- sudo iptables -L -t nat
sudo: effective uid is not 0, is /usr/bin/sudo on a file system with the 'nosuid' option set or an NFS file system without root privileges?

# Grant the proxy sidecar privileges escalation
kubectl edit deploy/httpbin -n default
      containers:
      - name: istio-proxy 
        image: docker.io/istio/proxyv2:1.8.4
        securityContext:
          allowPrivilegeEscalation: true # by default it's false
          privileged: true               # by default it's false

# Show iptables - 15006(incoming) and 15001(outgoing) traffic interception
kubectl -n default exec -it deploy/httpbin -c istio-proxy -- sudo iptables -L -t nat | grep -e ISTIO_IN_REDIRECT -e 15006 -e ISTIO_REDIRECT -e 15001
ISTIO_IN_REDIRECT  tcp  --  anywhere             anywhere            
Chain ISTIO_IN_REDIRECT (3 references)
REDIRECT   tcp  --  anywhere             anywhere             redir ports 15006
ISTIO_IN_REDIRECT  all  --  anywhere            !localhost            owner UID match istio-proxy
ISTIO_IN_REDIRECT  all  --  anywhere            !localhost            owner GID match istio-proxy
ISTIO_REDIRECT  all  --  anywhere             anywhere            
Chain ISTIO_REDIRECT (1 references)
REDIRECT   tcp  --  anywhere             anywhere             redir ports 15001

We can see here iptables is used to redirect incoming and outgoing traffic to Istio's data plane proxy. Incoming traffic goes to port 15006 of the Istio proxy while outgoing traffic will go through 15001. If we check the Envoy listeners for those ports, we can see exactly how the traffic gets handled.


IngressGateway

Query the gateway configuration to get all routes

istioctl proxy-config routes deploy/istio-ingressgateway.istio-system
NOTE: This output only contains routes loaded via RDS.
NAME        DOMAINS              MATCH                  VIRTUAL SERVICE
http.80     example.com          /*                     web-api-gw-vs.default
            *                    /stats/prometheus*     
            *                    /healthz/ready* 

# Show details of a single route
istioctl proxy-config routes deploy/istio-ingressgateway.istio-system --name http.80 -o json

Resources

  • gitops-istio Istio on a Kubernetes cluster and automating A/B testing and canary releases with GitOps pipelines example

Training Istio v1.5

Istio v1.7

Semi-related

Related technology

Tutorials and demos:

Investigate TLS:

Solutions: