85+ Istio Interview Questions and Answers (2027)
86 Istio interview questions with answers: istiod, Envoy, ambient mode, VirtualService, DestinationRule, mTLS, AuthorizationPolicy, istioctl, scenarios.
Istio interview questions test whether you can run a service mesh on Kubernetes: how istiod and the Envoy or ztunnel data plane fit together, how VirtualService and DestinationRule shape traffic, how mutual TLS and AuthorizationPolicy secure it, how sidecar and ambient modes differ, and how to debug the mesh with istioctl. This page has 86 questions with direct answers.
Answers were checked against the Istio documentation in November 2026, when the docs listed Istio 1.31 as current. Ambient mode has been generally available since Istio 1.24. If you are new to Kubernetes networking, read our Kubernetes architecture components guide first.
Key takeaways
- Freshers should explain what a service mesh solves, the control plane and data plane split, sidecar injection and the core CRDs.
- Mid-level candidates are asked to write VirtualService, DestinationRule, PeerAuthentication and AuthorizationPolicy YAML, and to read
istioctl proxy-configoutput. - Senior rounds cover ambient mode (ztunnel and waypoint proxies), multicluster topologies, revision-based upgrades, performance cost and Gateway API.
- Istio graduated in the CNCF in July 2023. Ambient mode reached general availability in 1.24 (November 2024).
- Most scenario questions come down to four tools:
istioctl analyze,istioctl proxy-status,istioctl proxy-configand Envoy access logs.
How to use this Istio question bank
This is the main Istio interview page on the site. For targeted practice, the Istio engineer questions focus on basics, the advanced Istio questions go deeper on architecture, and the real-time Istio questions are troubleshooting drills.
| Mode | Data plane | L4 (mTLS, L4 policy) | L7 (routing, retries, L7 policy) | Pod restart to join |
|---|---|---|---|---|
| Sidecar | Envoy in every pod | Sidecar | Sidecar | Yes |
| Ambient | ztunnel per node, optional waypoint proxies | ztunnel | Waypoint (Envoy) | No |
Istio interview questions for freshers
1. What is Istio?
Istio is an open-source service mesh. It adds traffic management, security (mutual TLS, identity and authorization) and observability to services without changing application code, by routing service-to-service traffic through proxies it controls.
2. What problem does a service mesh solve?
In a microservices system, every service needs retries, timeouts, encryption, access control and metrics. A mesh moves those concerns out of each codebase into a shared infrastructure layer that is configured centrally and applied consistently.
3. What are the control plane and data plane in Istio?
The control plane is istiod, which turns your configuration into proxy configuration, issues certificates and handles service discovery. The data plane is the set of proxies that carry traffic: Envoy sidecars in sidecar mode, or ztunnel and waypoint proxies in ambient mode.
4. What is istiod?
The single control plane binary. Since Istio 1.5 it combines what used to be separate components (Pilot for configuration, Citadel for certificates and Galley for validation) into one deployment.
5. What is Envoy?
A high-performance L4 and L7 proxy written in C++, originally built at Lyft and now a CNCF graduated project. Istio uses Envoy for sidecars, gateways and waypoint proxies and configures it dynamically over the xDS APIs.
6. What is a sidecar proxy?
An Envoy container added to each application pod. iptables rules (or the Istio CNI plugin) redirect the pod's inbound and outbound traffic through it, so the proxy can apply routing, mTLS and telemetry.
7. How do you enable automatic sidecar injection?
Label the namespace and restart the pods:
kubectl label namespace shop istio-injection=enabled
kubectl rollout restart deployment -n shop
A mutating admission webhook adds the sidecar when pods are created. With revision-based installs you use istio.io/rev=<revision> instead.
8. What is ambient mode?
A sidecarless data plane. A ztunnel DaemonSet on each node handles mTLS and L4 policy, and optional waypoint proxies handle L7 features. You enrol a namespace with the label istio.io/dataplane-mode=ambient, and pods join without a restart.
9. How do you install Istio?
With istioctl install and a profile, with Helm charts (base, istiod, gateway, and cni and ztunnel for ambient), or through a platform that bundles Istio. Run istioctl x precheck before installing.
10. What installation profiles exist?
default (production starting point), demo (extra features for learning, not for production), minimal (control plane only), remote (for multicluster), empty, preview and ambient.
11. What is a VirtualService?
A resource that defines how requests to a host are routed: which subsets or services receive them, based on path, headers or weights, plus timeouts, retries, fault injection and mirroring.
12. What is a DestinationRule?
A resource that applies policies after routing has chosen a destination: subsets (for example by version label), load balancing, connection pool limits, outlier detection and TLS settings for upstream connections.
13. What is the difference between a VirtualService and a DestinationRule?
The VirtualService decides where a request goes. The DestinationRule decides how traffic behaves once it is going there. A canary needs both: subsets in the DestinationRule and weights in the VirtualService.
14. What is an Istio Gateway?
A resource that configures a standalone Envoy at the mesh edge for inbound or outbound traffic: ports, protocols, hosts and TLS. Routing for that traffic is attached with a VirtualService. Istio also supports the Kubernetes Gateway API, which the docs recommend for new setups.
15. What is a ServiceEntry?
A resource that adds an entry to Istio's internal service registry, usually for an external API or database. Once registered, you can apply routing, TLS and telemetry to it like an in-mesh service.
16. What is mutual TLS?
TLS where both client and server present certificates. In Istio each workload gets a certificate tied to its identity, so services authenticate each other and traffic is encrypted, without code changes.
17. What is a workload identity in Istio?
A SPIFFE ID derived from the Kubernetes service account, in the form spiffe://<trust-domain>/ns/<namespace>/sa/<service-account>. It is embedded in the workload certificate that istiod signs.
18. What is PeerAuthentication?
A policy that sets the mTLS mode for receiving workloads: STRICT (only mTLS), PERMISSIVE (mTLS and plaintext, used during migration) or DISABLE. It can be applied mesh-wide, per namespace, per workload or per port.
19. What is AuthorizationPolicy?
A policy that allows or denies requests based on source identity, namespace, IP, request path, method, headers or JWT claims. Actions are ALLOW, DENY, AUDIT and CUSTOM (external authorization).
20. What is RequestAuthentication?
A policy that validates JSON Web Tokens on incoming requests against a JWKS. It rejects invalid tokens but allows requests without a token, so you pair it with an AuthorizationPolicy that requires a principal.
21. What observability does Istio give you without code changes?
Metrics for request count, duration and size per source and destination, access logs, and distributed tracing spans from the proxies. Applications still need to forward trace headers for traces to join up across services.
22. What is Kiali?
An open-source console for Istio that shows the service graph, traffic health and configuration validation. Run it locally with istioctl dashboard kiali once the add-on is installed.
23. What is istioctl?
The Istio command-line tool. It installs and upgrades Istio, analyses configuration, shows proxy sync status and configuration, manages waypoints and opens dashboards.
24. What does istioctl analyze do?
It scans live cluster resources or local files for problems such as a VirtualService pointing at a missing subset, conflicting policies, or namespaces without injection. Run it before and after every change.
25. What is traffic shifting?
Sending a percentage of traffic to different versions by setting weight on VirtualService routes. It is the basis of canary releases and blue-green cutovers.
26. What is fault injection?
Deliberately adding delays or aborts with HTTP error codes to a percentage of requests through a VirtualService. It is used to test how callers handle a slow or failing dependency.
27. What is traffic mirroring?
Copying live requests to a second destination while users only receive responses from the primary. Set mirror and mirrorPercentage in a VirtualService to test a new version with real traffic.
28. Which ports should you know for the sidecar?
15001 for outbound traffic, 15006 for inbound, 15000 for the Envoy admin interface (pod-local), 15020 for merged Prometheus metrics, 15021 for health checks and 15090 for Envoy's own metrics. HBONE tunnels use 15008.
29. Which ports does istiod use?
15012 for xDS and the certificate authority over TLS, 15010 for plaintext xDS (only for secure networks), 15014 for control plane monitoring, and 15017 for the admission webhook (exposed through Service port 443).
30. Is Istio a CNCF project?
Yes. Istio joined the CNCF in 2022 and graduated in July 2023.
Istio interview questions for 2 to 5 years of experience
31. Write a canary configuration that sends 10% of traffic to v2.
apiVersion: networking.istio.io/v1
kind: DestinationRule
metadata:
name: reviews
spec:
host: reviews
subsets:
- name: v1
labels: { version: v1 }
- name: v2
labels: { version: v2 }
---
apiVersion: networking.istio.io/v1
kind: VirtualService
metadata:
name: reviews
spec:
hosts: ["reviews"]
http:
- route:
- destination: { host: reviews, subset: v1 }
weight: 90
- destination: { host: reviews, subset: v2 }
weight: 10
32. How do you route by header, for example for internal testers?
Add an http rule with a match on the header before the default route. Rules are evaluated in order, and the first match wins, so put specific matches first.
33. How do you configure timeouts and retries?
Set timeout on the route and a retries block with attempts, perTryTimeout and retryOn. Make sure the total of retries times per-try timeout fits inside the overall timeout, and only retry idempotent requests.
34. What is outlier detection?
Passive health checking in a DestinationRule. Envoy ejects an endpoint from the load balancing pool after a number of consecutive errors (consecutive5xxErrors or consecutiveGatewayErrors) for baseEjectionTime, limited by maxEjectionPercent.
35. How do you implement circuit breaking?
Combine connectionPool limits (such as tcp.maxConnections, http.http1MaxPendingRequests and http.maxRequestsPerConnection) with outlier detection. Requests over the limit fail fast with a 503 instead of piling up.
36. Which load balancing algorithms are supported?
ROUND_ROBIN, LEAST_REQUEST, RANDOM and PASSTHROUGH under simple, and consistentHash for session affinity by header, cookie, source IP or query parameter. Locality-aware load balancing can be added on top.
37. How do you migrate a namespace to strict mTLS safely?
Start with PERMISSIVE, confirm in Kiali or metrics that all callers already use mTLS (watch the connection_security_policy label), fix any clients outside the mesh, then switch the namespace's PeerAuthentication to STRICT.
38. What does tls.mode: ISTIO_MUTUAL in a DestinationRule do?
It tells the client proxy to use Istio-issued certificates for the upstream connection. Auto mTLS normally handles this, so you set it explicitly mainly when you also override other TLS settings for that host.
39. How do you allow only one service to call another?
apiVersion: security.istio.io/v1
kind: AuthorizationPolicy
metadata:
name: payments-allow-orders
namespace: payments
spec:
selector:
matchLabels: { app: payments }
action: ALLOW
rules:
- from:
- source:
principals: ["cluster.local/ns/orders/sa/orders"]
Once any ALLOW policy selects a workload, requests that match none of its rules are denied.
40. In what order are AuthorizationPolicy actions evaluated?
CUSTOM first, then DENY, then ALLOW. If any DENY matches, the request is rejected. If ALLOW policies exist and none match, it is rejected. If no ALLOW policy applies, the request is allowed.
41. How do you set a default-deny policy for a namespace?
Create an AuthorizationPolicy in the namespace with an empty spec: {}. It selects all workloads and has no rules, so nothing is allowed until you add specific ALLOW policies.
42. How do you control egress traffic?
Set meshConfig.outboundTrafficPolicy.mode to REGISTRY_ONLY so only hosts in the registry are reachable, add ServiceEntries for approved external services, and optionally route them through an egress gateway for central control and logging.
43. What does the Sidecar resource do?
It limits which services a sidecar knows about and which ports it listens on. Restricting egress.hosts to the namespaces a workload actually calls reduces proxy memory use and the size of configuration pushes in big meshes.
44. What is the Telemetry API?
A resource (Telemetry) that configures metrics, access logs and tracing per mesh, namespace or workload: providers, sampling percentage, tag overrides and log filters. It replaces older MeshConfig-only settings.
45. How do you enable Envoy access logs?
Set meshConfig.accessLogFile: /dev/stdout at install time, or apply a Telemetry resource with an accessLogging provider. Then read the logs from the istio-proxy container.
46. What do Envoy response flags tell you?
Short codes in access logs that explain failures. Examples: UH no healthy upstream, UF upstream connection failure, UO upstream overflow (circuit breaker), NR no route configured, URX retry or connection limit reached, DC downstream connection terminated.
47. What does istioctl proxy-status show?
Each proxy and whether its cluster, listener, route and endpoint configuration is SYNCED, NOT SENT or STALE compared with istiod. STALE points to a push problem between istiod and the proxy.
48. What does istioctl proxy-config show?
The configuration a specific Envoy is actually using: cluster, listener, route, endpoint, secret and log subcommands. It proves whether your YAML reached the proxy the way you intended.
49. Why must applications propagate trace headers?
The proxy creates spans for each hop, but it cannot link an inbound request to the outbound calls the application makes. The app must copy headers such as traceparent or the B3 headers from incoming to outgoing requests.
50. How do you expose a service through the ingress gateway with TLS?
Store the certificate in a Kubernetes secret in the gateway's namespace, define a Gateway with a port 443 server using tls.mode: SIMPLE and credentialName, and bind a VirtualService to that Gateway for the host. With Gateway API, use a Gateway with an HTTPS listener and an HTTPRoute.
51. What is the difference between Istio's Gateway and the Kubernetes Gateway API?
Istio's networking.istio.io Gateway is Istio-specific and configures an existing gateway deployment. The Kubernetes Gateway API is a standard API with GatewayClass, Gateway and HTTPRoute, and Istio can deploy the gateway automatically. Istio uses Gateway API for waypoints too.
52. What does holdApplicationUntilProxyStarts do?
It makes the application container wait until the sidecar is ready, so the app does not fail calls made during start-up. Set it in proxy.istio.io/config annotations or mesh-wide in meshConfig.defaultConfig.
53. Why do Jobs sometimes never complete with a sidecar?
The main container exits but the sidecar keeps running, so the pod never finishes. Kubernetes native sidecar containers solve this because the sidecar stops after the main container. Istio supports running its proxy as a native sidecar; check your version's default, or use ambient mode for batch workloads.
54. What is an EnvoyFilter and why use it carefully?
A resource that patches the generated Envoy configuration directly. It can do things no other API does, but it depends on Envoy internals, can break silently on upgrade, and is not supported in all ambient paths. Prefer first-class APIs or a WasmPlugin.
55. What is a WasmPlugin?
A resource that loads a WebAssembly module into Envoy proxies to extend them, for example for custom authentication or header logic, without rebuilding Envoy.
56. How do you see Istio metrics in Prometheus?
Prometheus scrapes the merged metrics endpoint on port 15020 of each pod. Standard metrics include istio_requests_total, istio_request_duration_milliseconds and istio_tcp_sent_bytes_total, with labels such as source_workload, destination_service and response_code.
57. How does Istio rotate workload certificates?
The istio-agent in each proxy requests a certificate from istiod over SDS and renews it automatically before expiry. Workload certificates are short lived (24 hours by default), so revoking access mostly means removing authorization, not chasing old certificates.
58. How do you plug in your own certificate authority?
Create a cacerts secret in istio-system with your intermediate CA certificate and key, root certificate and chain before installing, or integrate an external CA through istiod's supported mechanisms such as cert-manager's istio-csr. All clusters that should trust each other need the same root.
Senior Istio interview questions
59. How does traffic flow in ambient mode?
The Istio CNI plugin redirects pod traffic to the node's ztunnel. ztunnel wraps it in an HBONE tunnel (HTTP CONNECT over mTLS on port 15008) to the destination node's ztunnel. If the destination uses a waypoint, traffic goes through the waypoint first, which applies L7 policy and routing.
60. What is a waypoint proxy?
An Envoy deployment, usually one per namespace or per service account, that handles L7 processing for ambient workloads. Create it with istioctl waypoint apply and attach workloads using the istio.io/use-waypoint label. You pay for L7 only where you need it.
61. When would you choose sidecar mode over ambient?
When you rely on features that ambient supports differently or not yet, such as some EnvoyFilter uses, per-pod L7 processing, or existing tooling built around sidecars. Ambient suits large meshes that want lower resource use and no pod restarts. Both modes can run in the same cluster.
62. What is the resource cost of the mesh, and how do you reduce it?
Each sidecar uses CPU and memory and adds latency per hop. Reduce it by scoping configuration with the Sidecar resource, discovery selectors in MeshConfig, tuned proxy resource requests, and ambient mode where L7 is not needed. Measure with your own load tests, since published numbers depend on workload.
63. What are discovery selectors?
A MeshConfig setting (discoverySelectors) that limits which namespaces istiod watches. It reduces control plane memory and push work in clusters where only some namespaces belong to the mesh.
64. How do you upgrade Istio with minimal risk?
Use a revision-based canary upgrade: install the new control plane with a revision name, move namespaces to it by changing labels (or a revision tag such as prod-stable), restart workloads in waves, check metrics, then remove the old revision. Read the upgrade notes for each minor version first.
65. What are revision tags?
Stable aliases such as prod-stable that point at a control plane revision. Namespaces reference the tag, so you change one tag with istioctl tag set instead of relabelling every namespace during an upgrade.
66. What multicluster models does Istio support?
Multi-primary (a control plane in every cluster) and primary-remote (remote clusters use a control plane in another cluster), each on a single network or across multiple networks. Cross-network traffic goes through east-west gateways. All clusters need a shared root of trust.
67. How does locality load balancing work?
Envoy prefers endpoints in the same region, zone and subzone as the caller, using Kubernetes topology labels. With outlier detection enabled, traffic fails over to the next locality when local endpoints become unhealthy. You can also set explicit distribution weights.
68. How would you integrate Istio with an external authorization service?
Register the service as an extension provider in MeshConfig (envoyExtAuthzHttp or envoyExtAuthzGrpc), then create an AuthorizationPolicy with action: CUSTOM naming that provider for selected paths. Open Policy Agent is a common choice.
69. How do you protect the control plane?
Run several istiod replicas with a PodDisruptionBudget, restrict who can create Istio resources with RBAC, keep plaintext port 15010 closed, use admission validation, and monitor push errors and latency (pilot_xds_pushes, pilot_proxy_convergence_time).
70. What happens to traffic if istiod goes down?
Existing proxies keep their last configuration and keep serving traffic. New pods cannot get configuration or certificates, configuration changes stop applying, and certificates eventually expire if istiod stays down longer than their lifetime.
71. How do you scale configuration in a mesh with thousands of services?
Limit what each proxy receives (Sidecar resource, exportTo on services and rules), use discovery selectors, avoid mesh-wide EnvoyFilters, keep VirtualServices per host small, and watch convergence time after pushes.
72. How do Istio and Gateway API work together for mesh routing?
Besides ingress, Gateway API's HTTPRoute can attach to a Service as a parent (the GAMMA pattern) to configure mesh traffic. In ambient mode, waypoints are Gateway API Gateways, so new L7 routing often uses HTTPRoute instead of VirtualService.
73. How do you approach zero-trust networking with Istio?
Turn on strict mTLS mesh-wide, apply default-deny AuthorizationPolicies per namespace, allow only named service accounts, require JWTs at the edge, lock down egress with REGISTRY_ONLY, and still keep Kubernetes NetworkPolicies as a second layer.
74. How would you compare Istio with Linkerd or Cilium service mesh?
Linkerd uses its own lightweight Rust proxy and aims for simplicity. Cilium can provide mesh features using eBPF with an Envoy for L7. Istio has the widest feature set and offers both sidecar and ambient modes. Choose based on required features, team skills and operating cost, and test with your own traffic.
Scenario-based Istio interview questions
75. After enabling injection, pods have only one container. What happened?
Existing pods were not restarted, the namespace label is wrong or conflicts with a revision label, the pod has sidecar.istio.io/inject: "false", or the webhook failed. Check kubectl get ns --show-labels, restart the deployment and read kubectl describe events.
76. Calls between services return 503 with flag UH. What do you check?
UH means no healthy upstream. Check that the destination has ready endpoints (istioctl proxy-config endpoint), that subset labels in the DestinationRule match pod labels, and whether outlier detection ejected every endpoint.
77. Requests fail with NR after a VirtualService change. Why?
NR means no route matched. The VirtualService host may not match the requested host, the match rules may not cover the request, or the gateway binding is wrong. Run istioctl analyze and inspect routes with istioctl proxy-config route.
78. A service broke after the namespace was set to STRICT mTLS. Who is affected?
Callers without sidecars, such as pods in non-mesh namespaces, legacy VMs or kubelet health checks on custom ports. Find them in metrics or logs, add them to the mesh, or allow plaintext on a specific port with a port-level PeerAuthentication while you migrate.
79. A canary at 10% shows nearly 50% of requests going to v2. Why?
Traffic might bypass the VirtualService, for example clients calling a pod IP, a headless service or a different host name. Requests from outside the mesh do not go through a client sidecar. Check the source in access logs and make sure callers use the service host.
80. A new AuthorizationPolicy blocks everything with RBAC: access denied. What happened?
An ALLOW policy now selects the workload, so anything not matching its rules is denied. Check the principals, namespaces and paths in the rule, and remember health checks and metric scrapers may need rules too. Use action: AUDIT or dry-run annotation to test new policies first.
81. Latency rose after adding Istio. How do you investigate?
Compare istio_request_duration_milliseconds with application latency to isolate proxy overhead, check proxy CPU throttling and limits, look for retries hiding errors, and check large configuration pushes. Scope configuration with the Sidecar resource and right-size proxy resources.
82. Calls to an external API fail after you set REGISTRY_ONLY. What is the fix?
Add a ServiceEntry for the external host with the correct ports and protocol, plus a DestinationRule if TLS origination is needed. Confirm with access logs that the destination shows as the ServiceEntry host, not BlackHoleCluster.
83. Some proxies show STALE in istioctl proxy-status. What do you do?
Check istiod logs and metrics for push errors, CPU or memory pressure, and rejected configuration (pilot_total_xds_rejects). Large or invalid configuration often causes it. Fix the bad resource, scale istiod if needed, and restart affected pods only as a last step.
84. An ambient workload ignores your L7 routing rule. Why?
L7 rules need a waypoint. Without one, ztunnel applies only L4 policy. Deploy a waypoint with istioctl waypoint apply, label the namespace or service with istio.io/use-waypoint, and attach the route to the waypoint or service.
85. gRPC streams drop during deployments. How do you fix it?
Set a drain duration so proxies finish connections before exit (terminationDrainDuration), add a preStop hook in the app, make sure the Service port is named or uses appProtocol: grpc so Istio treats it as HTTP/2, and configure client retries for unavailable.
86. The security team wants proof that all traffic is encrypted. How do you show it?
Query istio_requests_total grouped by connection_security_policy and show only mutual_tls, show STRICT PeerAuthentication across namespaces, and use Kiali's security badges. For ambient, ztunnel logs and metrics confirm HBONE mTLS connections.
How to prepare for an Istio interview
- Install Istio on kind or minikube with the demo profile and deploy the Bookinfo sample.
- Do a canary, fault injection and mirroring exercise, and inspect each change with
istioctl proxy-config. - Switch a namespace from PERMISSIVE to STRICT, then add default-deny and one ALLOW policy.
- Install ambient mode in a second cluster, add a waypoint and compare resource use.
- The Linux Foundation offers an Istio Certified Associate (ICA) exam; check its page for current details.
Istio sits on top of Kubernetes, so solid cluster skills come first. Our CKA training covers the networking and workload basics, in the Kharadi, Pune classroom and live online. Pair this guide with the Kubernetes scenario-based questions, the Kubernetes gateways comparison and the DevOps interview questions hub.
Frequently asked questions
Is Istio still worth learning in 2027?
Yes, if you target platform or Kubernetes roles. Istio is a CNCF graduated project, and ambient mode has lowered its resource cost. Interviewers also use it to test networking and security basics.
What is the difference between sidecar mode and ambient mode?
Sidecar mode puts an Envoy proxy in every pod. Ambient mode uses a shared ztunnel per node for mTLS and L4 policy, plus optional waypoint proxies for L7 features, and pods join without a restart.
Do I need to know Envoy to work with Istio?
You do not need to write Envoy configuration, but you should read access logs, response flags and istioctl proxy-config output, since that is how most Istio problems are debugged.
Should I use VirtualService or Gateway API?
Both are supported. Istio's documentation recommends Gateway API for new setups, and waypoints in ambient mode use it. Many existing meshes still run VirtualService, so learn both.
Does Istio replace Kubernetes NetworkPolicy?
No. NetworkPolicy works at L3 and L4 in the CNI, while Istio authorization works on identities and L7 attributes. Using both gives defence in depth.
Is there an Istio certification?
Yes. The Linux Foundation offers the Istio Certified Associate (ICA). Check the Linux Foundation Training page for the current domains and price.
References
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0