Istio and gateway lifecycle manager
Troubleshoot the installation, upgrade, or uninstallation of Istio components.
If your installation settings are configured incorrectly, the Gloo agent might not start or you might experience issues with other Istio components such as gateways.
Before reviewing these specific Istio operator topics, try Debugging Istio.
IstioLifecycleManager does not fully uninstall or namespaces stuck in terminating
What’s happening
You installed the istiod control plane and gateways by using the Gloo Istio lifecycle manager. When you follow the steps to uninstall Istio with the Istio lifecycle manager, the operator namespace, such as gm-iop-1-20, in one or more workload clusters is stuck in Terminating.
kubectl get ns --context ${REMOTE_CONTEXT}
Example output:
NAME STATUS AGE
bookinfo Active 132d
default Active 132d
global Active 123d
gloo-mesh Active 59d
gm-iop-1-20 Terminating 59d
helloworld Active 132d
httpbin Active 132d
kube-node-lease Active 132d
kube-public Active 132d
kube-system Active 132d
Why it’s happening
This error can happen when the istiod pod is deleted before the Istio or gateway lifecycle manager is able to finish uninstalling all Istio resources. This error might also happen if you repeatedly overwrite the same IstioLifecycleManager or GatewayLifecycleManager resource with different revisions and configuration.
How to fix it
In one workload cluster, get the
IstioOperatorresources in the operator namespace, such as the following command.kubectl get istiooperator.install.istio.io -n gm-iop-1-20 --context ${REMOTE_CONTEXT}Example output, in which operators still exist for an east-west gateway and the
istiodcontrol plane:NAME REVISION STATUS AGE istio-eastwestgateway-1-20 1-20 HEALTHY 59d istiod-control-plane 1-20 HEALTHY 59dFor each
IstioOperatorresource, repeat these steps to remove the finalizer.- Edit the resource, such as the following command. You might have a different revision or name for the operator resource.
kubectl edit istiooperator.install.istio.io -n gm-iop-1-20 istiod-control-plane --context ${REMOTE_CONTEXT} - Delete all lines in the
finalizerssection, including thefinalizers:line.apiVersion: install.istio.io/v1alpha1 kind: IstioOperator metadata: annotations: cluster.solo.io/cluster: cluster1 creationTimestamp: "2023-07-27T15:41:55Z" finalizers: - istio-finalizer.install.istio.io generation: 3 ... - Save and close the file. The resource is now deleted.
- Repeat these steps for any other
IstioOperatorresources in the namespace.
- Edit the resource, such as the following command. You might have a different revision or name for the operator resource.
After all resources in the operator namespace are deleted, verify that the operator namespace no longer exists.
kubectl get ns --context ${REMOTE_CONTEXT}Example output:
NAME STATUS AGE bookinfo Active 132d default Active 132d global Active 123d gloo-mesh Active 59d helloworld Active 132d httpbin Active 132d kube-node-lease Active 132d kube-public Active 132d kube-system Active 132dRepeat steps 1 - 3 for each workload cluster where you used the Istio lifecycle manager to install Istio.
Conflicting IstioLifecycleManager errors
What’s happening
- When you check the management logs, you see an Istio lifecycle manager error similar to the following:
failed to upsert snapshot for istio lifecycle manager","parent":"gloo-platform~gloo-mesh~cluster-1~admin.gloo.solo.io/v2, Kind=IstioLifecycleManager","err":"conflicting IOPs have been created from a different parent Istio lifecycle manager ... - When you check the Istio lifecycle manager status, you see a conflicting message similar to the following:Example output:
kubectl get IstioLifecycleManager -A -o yamlstatus: clusters: cluster-1: installations: auto: message: 'Another conflicting IstioLifecycleManager has previously been used to install a IstioOperators in this cluster, please check on uninstall of : gm-iop-1-20.gloo-platform' observedRevision: 1-20 state: FAILED
Why it’s happening
You might have a conflicting IstioLifecycleManager or GatewayLifecycleManager resource. For example, you might have uninstalled a previous IstioLifecycleManager resource that did not completely delete. This error can happen when the namespace is deleted before the Istio lifecycle manager is able to finish uninstalling all Istio resources.
How to fix it
- List the
IstioLifecycleManagerorGatewayLifecycleManagerresources. If you have multiple in the same namespace for the same purpose, try uninstalling the one that you no longer need.kubectl get IstioLifecycleManager -A - If you already uninstalled the
istiodcontrol plane or gateways by using the Istio lifecycle manager, try to manually replace theIstioLifecycleManagerorGatewayLifecycleManagerCR by deleting the existing resource, such as the following example command.kubectl delete GatewayLifecycleManager istio-eastwestgateway -n gloo-mesh --context $MGMT_CONTEXT
Agent crashes with IstioOperator error
What’s happening
Your Gloo agent does not start and is in a CrashLoopBackOff state.
When you check the agent logs, you see an error similar to the following:
failed to list *v1alpha1.IstioOperator: unknown field \"target\" in v1alpha1.ResourceMetricSource
...
{"level":"error","ts":1678198470.502656,"logger":"controller.input-ConfigMap-cache","caller":"controller/controller.go:208","msg":"Could not wait for Cache to sync","error":"failed to wait for input-ConfigMap-cache caches to sync: timed out waiting for cache to be synced","stacktrace":"sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).Start.func2\n\t/go/pkg/mod/sigs.k8s.io/controller-runtime@v0.11.2/pkg/internal/controller/controller.go:208\nsigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).Start\n\t/go/pkg/mod/sigs.k8s.io/controller-runtime@v0.11.2/pkg/internal/controller/controller.go:234\nsigs.k8s.io/controller-runtime/pkg/manager.(*runnableGroup).reconcile.func1\n\t/go/pkg/mod/sigs.k8s.io/controller-runtime@v0.11.2/pkg/manager/runnable_group.go:218"}
Why it’s happening
You might have an error in your Istio operator configuration, such as using a field that is deprecated or no longer supported.
By default, Gloo expects the Istio ingress gateway to have the name istio-ingressgateway. If you use a custom name for the ingress gateway, you cannot set up horizontal pod autoscaling (HPA) for the Istio gateways.
How to fix it
- Review your Istio operator configuration file for any deprecated or unsupported fields. For example, review the upstream Istio
operator.protofile for unsupported fields in your version of Istio. - If you use a custom Istio ingress gateway name, remove the HPA section from your Istio operator gateway configuration file. The following example shows the
hpaSpecin theistio-eastwestgatewayandistio-ingressgatewaysections.... components: ingressGateways: - name: <custom-name> namespace: gloo-mesh-gateways enabled: true k8s: hpaSpec: maxReplicas: 5 metrics: - resource: name: cpu targetAverageUtilization: 60 type: Resource ... - name:namespace: gloo-mesh-gateways enabled: true k8s: hpaSpec: maxReplicas: 5 metrics: - resource: name: cpu targetAverageUtilization: 60 type: Resource - To add back similar HPA functionality, set autoscaling minimum and maximum values to the gateway configuration. The following example shows both the
istio-eastwestgatewayandistio-ingressgatewaysections. Note that thetargetAverageUtilizationfield is also removed, because that field is deprecated in Istio 1.14 and later.... spec: values: gateways: istio-ingressgateway: autoscaleMin: 2 autoscaleMax: 5 ... istio-eastwestgateway: autoscaleMin: 2 autoscaleMax: 5 - Update your Istio operator configuration.