ACM Observability

Overview

Red Hat Advanced Cluster Management (ACM) provides centralized observability for multi-cluster OpenShift environments through its Observability Service. Metrics from all managed clusters (including the hub cluster itself) are collected and aggregated into a central Thanos-based storage system

This guide uses the multicluster observability add-on (MCOA) to aggregate mesh metrics on managed clusters (introduced in ACM 4.17), federate selected series, and remote-write them to the central Thanos-based storage system.

Kiali can query these aggregated metrics either through ACM’s external Observatorium API (using mTLS authentication) or directly through internal Thanos services. This guide explains both options, with detailed steps for the Observatorium API approach.

Architecture

Components

On the Hub Cluster:

• ACM Observability Service: Centralized observability platform
  • Observatorium API: External HTTPS endpoint with mTLS authentication
  • Thanos: Metrics storage and query engine (Query, Query Frontend, Receive, Store)

On Managed Clusters (Hub + Spokes):

• User Workload Monitoring (UWM): OpenShift’s Prometheus for user workloads
• PodMonitor/ServiceMonitor: Scrape Istio metrics from:
  • Sidecar proxies (in application namespaces)
  • Control plane (istiod in istio-system)
  • Ztunnel (in ztunnel namespace, for L4 metrics in Ambient mode)
  • Waypoint proxies (in application namespaces, for L7 metrics in Ambient mode)
• One MCOA PrometheusRule in a dedicated aggregation namespace: Aggregates raw per-proxy traffic series from all scraped namespaces into workload:istio_* series on UWM
• MCOA user-workload Prometheus Agent: Federates selected UWM series, restores standard istio_* metric names, and remote-writes them to hub Thanos

Kiali Deployment Location:

Kiali can be deployed on any cluster with network access to:

1. The hub cluster’s metrics backend (Observatorium API or internal Thanos services)
2. Each managed cluster’s Kubernetes API (for workload and configuration data)

Common deployment locations:

• Hub cluster (recommended): Co-located with ACM for lower latency metric queries and simplified networking. Can use internal Thanos services (HTTP) or external Observatorium API (HTTPS). Typically requires external deployment mode (ignore_home_cluster: true) since the hub usually doesn’t run mesh workloads or an Istio control plane.
• Spoke/managed cluster: Kiali deployed alongside the mesh workloads or the Istio control plane. Must use external Observatorium API route.
• Separate management cluster: Kiali deployed externally in dedicated “external deployment” mode (see External Kiali). Must use external Observatorium API route.

This guide assumes Kiali is deployed on the hub cluster in external deployment mode, but the configuration applies to any deployment location.

Metrics Flow

There are two independent flows:

Ingestion (managed cluster → hub):

1. Istio data plane components (sidecars, ztunnel, or waypoint proxies) expose metrics at :15020/stats/prometheus.
2. User Workload Monitoring Prometheus scrapes those metrics (typically every 30s).
3. UWM evaluates the cross-namespace recording rule in the dedicated mesh-observability namespace and produces workload:istio_* series.
4. The MCOA user-workload Prometheus Agent federates selected series through /federate, relabels workload:istio_* back to istio_*, and remote-writes them to the hub (every 5 minutes by default).
5. The hub stores them in Thanos Receive/Store and serves them through Thanos Query Frontend.

Query (Kiali → hub):

Kiali can query metrics through either of these paths:

Via Observatorium API Route (HTTPS with mTLS):

1. Kiali queries the external Observatorium API route.
2. Observatorium forwards the request to Thanos Query Frontend.
3. Thanos Query Frontend reads from Thanos Store/Receive and returns the result back through Observatorium to Kiali.

Via Internal Thanos Service (HTTP):

1. Kiali queries the internal Thanos Query Frontend service directly within the cluster, bypassing Observatorium.

Expected Latency: 5-6 minutes from traffic generation to visibility in Kiali due to the 5-minute (default) push interval.

Prerequisites

1. ACM Observability Service

ACM MultiClusterObservability must be installed on the hub cluster:

# Verify ACM Observability is running
oc get mco observability

# Check Observatorium API route
oc get route observatorium-api -n open-cluster-management-observability

2. User Workload Monitoring

User Workload Monitoring must be enabled on all clusters (hub and spokes):

# Enable UWM by editing cluster-monitoring-config
oc -n openshift-monitoring edit configmap cluster-monitoring-config

# Add:
# data:
#   config.yaml: |
#     enableUserWorkload: true

# Verify UWM pods are running
oc get pods -n openshift-user-workload-monitoring

See: Enabling monitoring for user-defined projects

3. Istio Metrics Collection

Create ServiceMonitor and PodMonitor resources to collect Istio metrics. The PodMonitor for sidecars must be created in each namespace with Istio sidecars because OpenShift monitoring ignores namespaceSelector in these resources. The ServiceMonitor for istiod is created once in istio-system.

ServiceMonitor for istiod (in istio-system):

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: istiod-monitor
  namespace: istio-system
spec:
  targetLabels:
  - app
  selector:
    matchLabels:
      istio: pilot
  endpoints:
  - port: http-monitoring
    interval: 30s

PodMonitor for Istio proxies (must be applied in every mesh namespace):

apiVersion: monitoring.coreos.com/v1
kind: PodMonitor
metadata:
  name: istio-proxies-monitor-<your-mesh-namespace>
  namespace: <your-mesh-namespace>
spec:
  selector:
    matchExpressions:
    - key: istio-prometheus-ignore
      operator: DoesNotExist
  podMetricsEndpoints:
  - path: /stats/prometheus
    interval: 30s
    relabelings:
    - action: keep
      sourceLabels: ["__meta_kubernetes_pod_container_name"]
      regex: "istio-proxy"
    - action: keep
      sourceLabels: ["__meta_kubernetes_pod_annotationpresent_prometheus_io_scrape"]
    - action: replace
      regex: (\d+);(([A-Fa-f0-9]{1,4}::?){1,7}[A-Fa-f0-9]{1,4})
      replacement: '[$2]:$1'
      sourceLabels: ["__meta_kubernetes_pod_annotation_prometheus_io_port","__meta_kubernetes_pod_ip"]
      targetLabel: "__address__"
    - action: replace
      regex: (\d+);((([0-9]+?)(\.|$)){4})
      replacement: '$2:$1'
      sourceLabels: ["__meta_kubernetes_pod_annotation_prometheus_io_port","__meta_kubernetes_pod_ip"]
      targetLabel: "__address__"
    - sourceLabels: ["__meta_kubernetes_pod_label_app_kubernetes_io_name","__meta_kubernetes_pod_label_app"]
      separator: ";"
      targetLabel: "app"
      action: replace
      regex: "(.+);.*|.*;(.+)"
      replacement: "${1}${2}"
    - sourceLabels: ["__meta_kubernetes_pod_label_app_kubernetes_io_version","__meta_kubernetes_pod_label_version"]
      separator: ";"
      targetLabel: "version"
      action: replace
      regex: "(.+);.*|.*;(.+)"
      replacement: "${1}${2}"
    - action: replace
      regex: "(.+)"
      replacement: "${1}"
      sourceLabels:
      - __meta_kubernetes_pod_label_app_kubernetes_io_name
      targetLabel: app_kubernetes_io_name
    - action: replace
      regex: "(.+)"
      replacement: "${1}"
      sourceLabels:
      - __meta_kubernetes_pod_label_app_kubernetes_io_version
      targetLabel: app_kubernetes_io_version
    - sourceLabels: ["__meta_kubernetes_namespace"]
      action: replace
      targetLabel: namespace
    - action: replace
      replacement: "<your-mesh-identification-string>"
      targetLabel: mesh_id

See: Configuring OpenShift Monitoring with Service Mesh

Ambient Mode Metrics

If you are using Istio’s Ambient mode instead of (or in addition to) sidecar mode, you need additional PodMonitors to collect metrics from the Ambient data plane components.

Understanding Ambient Mode Metrics

Ambient mode uses a layered architecture with two metric sources:

Ztunnel (L4 metrics only)

• Runs as a DaemonSet (namespace varies by installation)
• Handles all L4 traffic for pods enrolled in ambient mode
• Produces TCP-level metrics:
  • istio_tcp_sent_bytes_total
  • istio_tcp_received_bytes_total
  • istio_tcp_connections_opened_total
  • istio_tcp_connections_closed_total
• Does not produce HTTP metrics

Waypoint proxies (L7 metrics)

• Run as Deployments in application namespaces
• Optional L7 proxies deployed per-namespace or per-service
• Produce full HTTP metrics (same as sidecars):
  • istio_requests_total
  • istio_request_duration_milliseconds_*
  • istio_request_bytes_*
  • istio_response_bytes_*
  • Plus all TCP metrics listed above

If you only use ztunnel (no waypoints), Kiali will show TCP traffic but not HTTP-level details like response codes or latency histograms.

PodMonitor for Ztunnel

Create a PodMonitor in the namespace where ztunnel runs. Ztunnel pods expose metrics using the same interface as sidecars:

• Container name: istio-proxy
• Annotation: prometheus.io/scrape: "true"
• Metrics path: /stats/prometheus on port 15020

Because ztunnel uses the same metrics interface, you can use the same PodMonitor configuration shown in the Istio Metrics Collection section above, changing only the namespace field to match your ztunnel namespace.

PodMonitor for Waypoint Proxies

Create a PodMonitor in each namespace with a waypoint. Waypoint pods also expose metrics using the same interface as sidecars:

• Container name: istio-proxy
• Annotation: prometheus.io/scrape: "true"
• Metrics path: /stats/prometheus on port 15020

Because waypoints use the same metrics interface, you can use the same PodMonitor configuration shown in the Istio Metrics Collection section above.

4. Enable the ACM multicluster observability add-on

MCOA is disabled by default. It uses the OpenShift Cluster Observability Operator’s Prometheus Operator and Prometheus Agent to federate and remote-write metrics from managed clusters. Install the Cluster Observability Operator as required by ACM, then enable both platform and user-workload metrics on the hub:

apiVersion: observability.open-cluster-management.io/v1beta2
kind: MultiClusterObservability
metadata:
  name: observability
spec:
  capabilities:
    platform:
      metrics:
        default:
          enabled: true
    userWorkloads:
      metrics:
        default:
          enabled: true

Enabling these capabilities replaces the legacy metrics collectors with MCOA collectors. See the ACM 2.17 MCOA documentation.

5. Aggregate and federate Istio metrics

Use UWM as the Edge Prometheus described in Recording Rules and Federation. MCOA’s Prometheus Agent is the federation and remote-write layer, and ACM Observatorium/Thanos is the long-retention federated backend that Kiali queries.

Create the following resources on the hub in namespace open-cluster-management-observability: one PrometheusRule targeting the dedicated mesh-observability namespace, one shared user-workload ScrapeConfig, and one cluster-wide platform ScrapeConfig for container CPU and memory. Application namespaces still need their ServiceMonitor/PodMonitor resources, but they are not MCOA rule targets.

Before propagating the rule, exempt mesh-observability from UWM label enforcement on every managed cluster. This is required because UWM otherwise injects the rule namespace into selectors and recorded series, preventing one rule from aggregating metrics across application namespaces:

apiVersion: v1
kind: ConfigMap
metadata:
  name: user-workload-monitoring-config
  namespace: openshift-user-workload-monitoring
data:
  config.yaml: |
    namespacesWithoutLabelEnforcement:
    - mesh-observability    

Create one PrometheusRule named kiali-istio-aggregation. Its annotation tells MCOA to propagate it into mesh-observability; the label selects UWM Prometheus, rather than Thanos Ruler, to evaluate the rule on each managed cluster.

apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: kiali-istio-aggregation
  namespace: open-cluster-management-observability
  annotations:
    observability.open-cluster-management.io/target-namespace: mesh-observability
  labels:
    app.kubernetes.io/component: user-workload-metrics-collector
    openshift.io/prometheus-rule-evaluation-scope: leaf-prometheus
spec:
  groups:
  - name: istio.workload-aggregation
    interval: 30s
    rules:
    - record: workload:istio_requests_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_requests_total)
    - record: workload:istio_request_messages_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_messages_total)
    - record: workload:istio_response_messages_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_response_messages_total)
    - record: workload:istio_tcp_sent_bytes_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_tcp_sent_bytes_total)
    - record: workload:istio_tcp_received_bytes_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_tcp_received_bytes_total)
    - record: workload:istio_tcp_connections_opened_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_tcp_connections_opened_total)
    - record: workload:istio_tcp_connections_closed_total
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_tcp_connections_closed_total)
    - record: workload:istio_request_duration_milliseconds_bucket
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_duration_milliseconds_bucket)
    - record: workload:istio_request_duration_milliseconds_sum
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_duration_milliseconds_sum)
    - record: workload:istio_request_duration_milliseconds_count
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_duration_milliseconds_count)
    - record: workload:istio_request_bytes_bucket
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_bytes_bucket)
    - record: workload:istio_request_bytes_sum
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_bytes_sum)
    - record: workload:istio_request_bytes_count
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_request_bytes_count)
    - record: workload:istio_response_bytes_bucket
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_response_bytes_bucket)
    - record: workload:istio_response_bytes_sum
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_response_bytes_sum)
    - record: workload:istio_response_bytes_count
      expr: sum without (pod, pod_template_hash, instance, job, node) (istio_response_bytes_count)

Next, create this UWM ScrapeConfig. It federates the aggregated traffic metrics and the non-aggregated Istio, control-plane, process, and Envoy metrics that Kiali needs. The relabeling restores the original istio_* names before remote write. In effect, these are the metrics that will be stored in the federated Prometheus (i.e. the hub Thanos) that Kiali will see.

apiVersion: monitoring.rhobs/v1alpha1
kind: ScrapeConfig
metadata:
  name: kiali-istio-federation
  namespace: open-cluster-management-observability
  labels:
    app.kubernetes.io/component: user-workload-metrics-collector
spec:
  honorLabels: true
  jobName: kiali-istio-federation
  metricRelabelings:
  - action: replace
    regex: 'workload:(.*)'
    replacement: '${1}'
    sourceLabels: [__name__]
    targetLabel: __name__
  metricsPath: /federate
  params:
    match[]:
    - '{__name__=~"workload:istio_requests_total"}'
    - '{__name__=~"workload:istio_request_bytes_(bucket|count|sum)"}'
    - '{__name__=~"workload:istio_request_duration_milliseconds_(bucket|count|sum)"}'
    - '{__name__=~"workload:istio_request_messages_total"}'
    - '{__name__=~"workload:istio_response_bytes_(bucket|count|sum)"}'
    - '{__name__=~"workload:istio_response_messages_total"}'
    - '{__name__=~"workload:istio_tcp_connections_(opened|closed)_total"}'
    - '{__name__=~"workload:istio_tcp_(received|sent)_bytes_total"}'
    - '{__name__=~"istio_build|process_cpu_seconds_total|process_resident_memory_bytes"}'
    - '{__name__=~"pilot_info|pilot_proxy_convergence_time_(sum|count)|pilot_services|pilot_xds$|pilot_xds_pushes"}'
    - '{__name__=~"workload_manager_active_proxy_count"}'
    - '{__name__=~"envoy_cluster_upstream_cx_active|envoy_cluster_upstream_rq_total|envoy_listener_downstream_cx_active|envoy_listener_http_downstream_rq|envoy_server_memory_allocated|envoy_server_memory_heap_size|envoy_server_uptime"}'
  scrapeInterval: 5m

Now create this platform ScrapeConfig once per cluster. It collects CPU and memory metrics for pods in all namespaces; no namespace matcher is needed.

apiVersion: monitoring.rhobs/v1alpha1
kind: ScrapeConfig
metadata:
  name: kiali-istio-platform-federation
  namespace: open-cluster-management-observability
  labels:
    app.kubernetes.io/component: platform-metrics-collector
spec:
  jobName: kiali-istio-platform-federation
  metricsPath: /federate
  params:
    match[]:
    - '{__name__=~"container_cpu_usage_seconds_total|container_memory_working_set_bytes"}'
  scrapeInterval: 5m

Finally, add references for the single PrometheusRule and both ScrapeConfig objects to the selected placement in the existing ClusterManagementAddOn object named multicluster-observability-addon. The following is a fragment of that placement’s configs list; merge these entries with its existing entries rather than applying this as a replacement for the whole add-on object.

oc get clustermanagementaddon multicluster-observability-addon -o yaml

spec:
  installStrategy:
    placements:
    - name: <placement-name>
      namespace: <placement-namespace>
      configs:
      - group: monitoring.coreos.com
        resource: prometheusrules
        name: kiali-istio-aggregation
        namespace: open-cluster-management-observability
      - group: monitoring.rhobs
        resource: scrapeconfigs
        name: kiali-istio-federation
        namespace: open-cluster-management-observability
      - group: monitoring.rhobs
        resource: scrapeconfigs
        name: kiali-istio-platform-federation
        namespace: open-cluster-management-observability

The kiali-istio-federation user-workload ScrapeConfig collects the aggregated traffic metrics rather than raw per-proxy series. The cluster-wide platform ScrapeConfig collects CPU and memory separately because those metrics belong to OpenShift platform monitoring rather than UWM.

Configuring Kiali for ACM Observability

Choosing Between Observatorium API and Internal Thanos Services

You have two options for connecting Kiali to ACM metrics:

Option 1: Observatorium API Route (HTTPS with mTLS)

external_services:
  prometheus:
    url: "https://observatorium-api-<namespace>.<apps-domain>/api/metrics/v1/default"
    auth:
      type: none
      cert_file: "secret:acm-observability-certs:tls.crt"
      key_file: "secret:acm-observability-certs:tls.key"

Provides:

• HTTPS with mTLS authentication and encryption
• External access (can be accessed from outside the cluster if needed)
• RBAC enforcement via Observatorium
• Multi-tenant isolation
• Requires certificate setup

Option 2: Internal Thanos Service (HTTP)

external_services:
  prometheus:
    url: "http://observability-thanos-query-frontend.open-cluster-management-observability.svc:9090"
    auth:
      type: none

Provides:

• Simpler setup (no certificates required)
• Direct access to Thanos (potentially lower latency)
• Internal cluster networking only
• HTTP only (no encryption between Kiali and Thanos)

Recommendation: Use the Observatorium API for production environments where you want encrypted connections and proper authentication. Use internal services for development/testing environments where simplicity is preferred or where network security is already provided by the cluster infrastructure.

The rest of this guide focuses on the Observatorium API approach with mTLS authentication.

Step 1: Obtain mTLS Certificates from ACM

ACM automatically creates long-lived client certificates (1 year validity) for accessing the Observatorium API. Extract these from the hub cluster:

# Extract client certificate (for authentication)
oc get secret observability-grafana-certs \
  -n open-cluster-management-observability \
  -o jsonpath='{.data.tls\.crt}' | base64 -d > tls.crt

# Extract client key (for authentication)
oc get secret observability-grafana-certs \
  -n open-cluster-management-observability \
  -o jsonpath='{.data.tls\.key}' | base64 -d > tls.key

Note: These certificates are created automatically when ACM MultiClusterObservability is deployed and are already trusted by the Observatorium API.

oc get secrets -n open-cluster-management-observability | grep -i cert

Step 2: Extract Server CA Certificate

Extract the CA certificate that signed the Observatorium API server certificate. This is used by Kiali to validate the server’s TLS certificate.

First, identify which CA issued the server certificate:

# Get the Observatorium API route hostname
HOST=$(oc get route observatorium-api -n open-cluster-management-observability -o jsonpath='{.spec.host}')

# Check who issued the server certificate
echo | openssl s_client -connect "${HOST}:443" -servername "${HOST}" -showcerts 2>/dev/null | openssl x509 -noout -issuer

Example output:

issuer=C=US, O=Red Hat, Inc., CN=observability-server-ca-certificate

Then, extract the matching CA certificate based on the issuer CN:

If the issuer CN is observability-server-ca-certificate:

oc get secret observability-server-ca-certs \
  -n open-cluster-management-observability \
  -o jsonpath='{.data.ca\.crt}' | base64 -d > server-ca.crt

If the issuer CN is observability-client-ca-certificate:

oc get secret observability-client-ca-certs \
  -n open-cluster-management-observability \
  -o jsonpath='{.data.ca\.crt}' | base64 -d > server-ca.crt

Note: Both secrets are in the open-cluster-management-observability namespace. The exact CA used may vary depending on your ACM version and configuration.

Step 3: Create Kubernetes Resources

Create the mTLS certificate secret in Kiali’s namespace:

KIALI_NAMESPACE="istio-system"  # Replace with your Kiali namespace

oc create secret generic acm-observability-certs \
  -n ${KIALI_NAMESPACE} \
  --from-file=tls.crt=tls.crt \
  --from-file=tls.key=tls.key

Create the CA bundle ConfigMap in Kiali’s namespace:

oc create configmap kiali-cabundle \
  -n ${KIALI_NAMESPACE} \
  --from-file=additional-ca-bundle.pem=server-ca.crt

For more details about CA bundle configuration, see TLS Configuration.

Step 4: Get Observatorium API URL

Find the external Observatorium API route URL:

oc get route observatorium-api \
  -n open-cluster-management-observability \
  -o jsonpath='https://{.spec.host}/api/metrics/v1/default'

The URL format is: https://observatorium-api-<namespace>.<apps-domain>/api/metrics/v1/default

Step 5: Configure Kiali

Using Kiali Operator (Kiali CR):

spec:
  external_services:
    prometheus:
      # Use Observatorium API route
      url: "<observatorium-api-url>"

      auth:
        type: none  # mTLS authentication at TLS layer, no Authorization header
        cert_file: "secret:acm-observability-certs:tls.crt"
        key_file: "secret:acm-observability-certs:tls.key"

      # Enable Thanos proxy mode
      thanos_proxy:
        enabled: true
        retention_period: "365d"
        scrape_interval: "5m"

Using Server Helm Chart:

OBSERVATORIUM_API_URL="$(oc get route observatorium-api -n open-cluster-management-observability -o jsonpath='https://{.spec.host}/api/metrics/v1/default')"

helm install kiali kiali-server \
  --namespace ${KIALI_NAMESPACE} \
  --set external_services.prometheus.url="${OBSERVATORIUM_API_URL}" \
  --set external_services.prometheus.auth.type="none" \
  --set external_services.prometheus.auth.cert_file="secret:acm-observability-certs:tls.crt" \
  --set external_services.prometheus.auth.key_file="secret:acm-observability-certs:tls.key" \
  --set external_services.prometheus.thanos_proxy.enabled="true" \
  --set external_services.prometheus.thanos_proxy.retention_period="365d" \
  --set external_services.prometheus.thanos_proxy.scrape_interval="5m"

Important Configuration Notes

Scrape Intervals

This architecture has several independent intervals. They control different stages of the metrics pipeline and should not be treated as one global scrape period:

• Edge target scraping — ServiceMonitor.spec.endpoints[].interval and PodMonitor.spec.podMetricsEndpoints[].interval control how often the spoke’s UWM Prometheus scrapes Istiod, sidecars, ztunnel, and waypoint metrics. If omitted, UWM uses the prometheus.scrapeInterval default of 30s.
• Recording-rule evaluation — PrometheusRule.spec.groups[].interval is how often UWM evaluates the aggregation rules. This is an evaluation interval, not a target scrape interval. If omitted, UWM uses the prometheus.evaluationInterval default of 30s.
• UWM-to-hub federation — ScrapeConfig.spec.scrapeInterval controls how often the MCOA PrometheusAgent fetches metrics from the spoke UWM Prometheus /federate endpoint and remote-writes them to hub Thanos. If omitted, the MCOA PrometheusAgent default is 300s (5m).
• Kiali’s Thanos hint — external_services.prometheus.thanos_proxy.scrape_interval tells Kiali what interval to use when calculating rate windows and query steps. It does not control federation. If omitted, Kiali defaults it to 30s; for ACM, set it to the same effective interval as the MCOA federation job, whose default is 5m.

If a ServiceMonitor or PodMonitor omits its endpoint interval, OpenShift UWM uses the default from the user-workload-monitoring-config ConfigMap. If a PrometheusRule group omits its interval, UWM uses the ConfigMap’s rule evaluation default. These defaults can be configured without changing every monitor or rule:

apiVersion: v1
kind: ConfigMap
metadata:
  name: user-workload-monitoring-config
  namespace: openshift-user-workload-monitoring
data:
  config.yaml: |
    prometheus:
      evaluationInterval: 30s
      scrapeInterval: 30s    

The ConfigMap settings are defaults for omitted ServiceMonitor, PodMonitor, and PrometheusRule interval fields. They do not configure UWM-to-hub federation: ScrapeConfig.spec.scrapeInterval defaults to 5m, and Kiali’s thanos_proxy.scrape_interval defaults to 30s. For ACM, configure Kiali’s value to match the effective federation interval.

These intervals must be selected as a pipeline. The 30s edge scrape and rule evaluation intervals determine how quickly UWM discovers new samples and refreshes the aggregated recording-rule series; they do not make those samples available in hub Thanos every 30 seconds. The 5m federation interval determines how frequently new data points arrive in hub Thanos, so it determines both the normal metrics delay and the minimum useful range for PromQL rate() queries. A rate window should contain at least two federated samples, which means it should normally be at least twice the effective federation interval (10m for a 5m federation interval).

If the federation interval changes, update Kiali’s thanos_proxy.scrape_interval, dashboard and alert rate() windows, and any health-check duration settings together. If Kiali’s thanos_proxy.scrape_interval remains at its 30s default while federation remains at 5m, Kiali can calculate rate windows and query steps that are too short for the samples actually arriving in Thanos. This can produce empty or misleading rate results even though the underlying metrics exist. Explicit PromQL range windows, such as [5m], are separate settings and must also be long enough to contain at least two federated samples. Conversely, changing only the edge scrape interval affects freshness and UWM resource usage, but does not reduce the hub’s federation delay.

Metrics Latency

The MCOA PrometheusAgent scrapes its federation endpoint every 300 seconds by default and remote-writes the result to Thanos. This means there is normally a 5-6 minute delay before new metrics appear in Kiali.

To use a different interval for the Kiali federation jobs, set spec.scrapeInterval on the kiali-istio-federation and kiali-istio-platform-federation ScrapeConfig resources. For example, the following makes the shared user-workload job run every minute:

spec:
  jobName: kiali-istio-federation
  scrapeInterval: 1m
  # Other fields from the ScrapeConfig in step 5 remain unchanged.

Set Kiali’s external_services.prometheus.thanos_proxy.scrape_interval to the same interval. Do not change the generated MCOA PrometheusAgent merely to alter these Kiali jobs: that changes the default interval for the agent’s other jobs as well.

Initial warm-up period: After deploying a new application, it takes approximately twice the collection interval before data appears in Kiali’s graph and metrics tab. This is because Kiali uses PromQL rate() functions which require at least two data points to compute a result, and with ACM’s collection interval, two data points take at least two collection cycles to accumulate. For example, with the default 5-minute interval, expect a ~10-minute warm-up period. After this initial warm-up, all time ranges in Kiali should display data normally. However, keep in mind that the most recent data visible in Kiali will always be at least one collection interval old, since metrics must complete a full collection cycle before they appear in Thanos.

Thanos Proxy Mode

Enable thanos_proxy when using ACM/Thanos:

external_services:
  prometheus:
    thanos_proxy:
      enabled: true
      retention_period: "365d" # Match your actual ACM Thanos retention
      scrape_interval: "5m"   # Must match the MCOA federation interval

When enabled: true, Kiali uses the configured scrape_interval and retention_period values directly, rather than querying Prometheus’s /api/v1/status/config and /api/v1/status/runtimeinfo endpoints to discover them. This is necessary because Thanos does not expose these Prometheus configuration endpoints.

Why these values matter:

• scrape_interval: Kiali’s UI uses this value to compute PromQL rate() intervals and query step sizes. The rate interval must be large enough to contain at least two data points for rate() to produce results. With ACM, data points arrive in Thanos at the ACM collection interval (default 5 minutes), not at the local Prometheus scrape interval (typically 15-30 seconds). If scrape_interval is set too low (e.g., “30s”), the computed rate windows will be too narrow to capture two ACM data points, causing Kiali’s metrics tab to show empty charts even though data exists in Thanos.

• retention_period: Used to limit time range queries to available data. ACM defaults to 365d retention when spec.advanced.retentionConfig is not explicitly configured in the MultiClusterObservability CR. If using the default, set retention_period to “365d”. If configuring custom retention, use at least 10d minimum (a Thanos requirement for downsampling to function). Always match retention_period to your actual ACM retention configuration. Use 14d only when your MultiClusterObservability retention policy is configured for 14 days.

Multi-Cluster Setup

For multi-cluster service mesh deployments with ACM:

1. Metrics Aggregation (Handled by ACM)

ACM automatically aggregates metrics from all managed clusters. Each cluster’s metrics include a cluster label with the cluster name (the metadata.name of the ManagedCluster resource). To get a list of all the clusters managed by ACM, run oc get managedcluster on the hub cluster.

Kiali can filter metrics by cluster using query_scope. The query_scope configuration adds label filters to every Prometheus query:

external_services:
  prometheus:
    # Example 1: Filter to a single cluster
    query_scope:
      cluster: "east-cluster"

    # Example 2: Filter by mesh_id and cluster
    query_scope:
      mesh_id: "mesh-1"
      cluster: "east-cluster"

Each key-value pair in query_scope is added as key="value" to every query. For example, cluster: "east-cluster" adds cluster="east-cluster" to all PromQL queries.

2. Remote Cluster Access (For Workload/Config Data)

While metrics come from ACM’s central Thanos, Kiali still needs direct API access to each cluster for:

• Workload and service discovery
• Istio configuration validation
• Kubernetes resource details

Create remote cluster secrets as described in the multi-cluster setup guide. When the external Kiali deployment uses auth.strategy: openshift, also follow the OpenShift multi-cluster authentication guidance. In particular, a remote cluster that provides only remote_cluster_resources_only: true needs its Kiali OAuthClient callback URI to point to the external Kiali route and end in /api/auth/callback/<remote-cluster-name>. This remote API authentication is independent of ACM/MCOA metrics federation.

3. External Deployment Model

For multi-cluster with ACM, if you deploy Kiali on the hub cluster (or on a separate management cluster), you will typically want to run Kiali in external deployment mode:

clustering:
  ignore_home_cluster: true  # Kiali is external to mesh

kubernetes_config:
  cluster_name: "<management-cluster-name>"  # Unique name for the cluster where Kiali runs

This is Kiali’s own home-cluster identity. It is not required to match the ACM ManagedCluster name or an Istio clusterName; choose a value unique among the clusters configured in Kiali. In this external deployment model, set clustering.ignore_home_cluster: true so Kiali does not attempt to treat the management cluster as a mesh cluster.

See the External Kiali guide for complete external deployment instructions.

Certificate Management

Automatic Rotation

ACM-issued certificates (stored in the observability-grafana-certs secret in the ACM observability namespace) have 1-year validity and are automatically rotated by ACM before expiration. When certificates are rotated:

1. ACM updates the observability-grafana-certs secret in open-cluster-management-observability namespace
2. You must update the acm-observability-certs secret in Kiali’s namespace with the new certificate data. Options include:
   • Re-run the extraction commands from Step 1: Obtain mTLS Certificates from ACM manually
   • Use an ACM ConfigurationPolicy with hub cluster templating to automatically distribute and update the secret to the cluster where Kiali runs (see ACM Governance documentation for details)
3. Kubernetes updates the mounted files in Kiali pod (within 60 seconds after the secret update)
4. Kiali automatically uses new certificates on next connection (no pod restart needed)

Using Custom Certificates

If you prefer to use your own certificate infrastructure instead of ACM’s certificates:

1. Generate/obtain certificates signed by a CA trusted by ACM Observatorium API
2. Configure ACM to trust your CA (consult ACM documentation)
3. Create the acm-observability-certs secret with your certificates

Verification

Check Certificate Configuration

# Verify secret exists
oc get secret acm-observability-certs -n ${KIALI_NAMESPACE}

# Check certificate expiration
oc get secret acm-observability-certs -n ${KIALI_NAMESPACE} \
  -o jsonpath='{.data.tls\.crt}' | base64 -d | \
  openssl x509 -noout -enddate

# Verify CA bundle
oc get configmap kiali-cabundle -n ${KIALI_NAMESPACE} \
  -o jsonpath='{.data.additional-ca-bundle\.pem}' | \
  openssl x509 -noout -subject

Check Kiali Logs

Verify certificates are loaded successfully:

oc logs -n ${KIALI_NAMESPACE} deployment/kiali | grep -i "credential\|certificate"

# Expected output (at "info" log level):
# INF Loaded [1] valid CA certificate(s) from [/kiali-cabundle/additional-ca-bundle.pem]
#
# Additional output (at "debug" log level):
# DBG Credential file path configured: [/kiali-override-secrets/prometheus-cert/tls.crt]
# DBG Credential file path configured: [/kiali-override-secrets/prometheus-key/tls.key]

Test Metrics

1. Generate mesh traffic in one of your managed clusters
2. Wait for the initial warm-up period (approximately twice the ACM collection interval; default ~10 minutes) for metrics to propagate to Thanos and for enough data points to accumulate for rate calculations. The graph may appear sooner (after ~5 minutes).
3. Access Kiali UI and navigate to a workload
4. Verify metrics appear in the Metrics tab and traffic graph

Verify Metrics in Thanos Directly

Validate each stage of the federation pipeline: raw Istio metrics in the managed cluster’s edge UWM Prometheus, workload:* recording-rule output in that same Prometheus, and the relabeled metrics in hub Thanos. Set the three values first:

HUB_CONTEXT="<hub-kubecontext>"
MANAGED_CONTEXT="<managed-cluster-kubecontext>"
MANAGED_CLUSTER_NAME="<acm-managed-cluster-name>" # List names with: oc --context="${HUB_CONTEXT}" get managedcluster -o jsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}'

PROM_POD=$(oc --context="${MANAGED_CONTEXT}" \
  -n openshift-user-workload-monitoring \
  get pods -l app.kubernetes.io/name=prometheus \
  -o jsonpath='{.items[0].metadata.name}')

# Edge UWM: raw counters scraped from Istio proxies
oc --context="${MANAGED_CONTEXT}" \
  -n openshift-user-workload-monitoring \
  exec -c prometheus "${PROM_POD}" -- \
  wget -qO- \
  'http://localhost:9090/api/v1/query?query=sum%28istio_requests_total%29' \
  | jq '.data.result'

# Edge UWM: aggregates produced by the recording rules
oc --context="${MANAGED_CONTEXT}" \
  -n openshift-user-workload-monitoring \
  exec -c prometheus "${PROM_POD}" -- \
  wget -qO- \
  'http://localhost:9090/api/v1/query?query=sum%28workload%3Aistio_requests_total%29' \
  | jq '.data.result'

# Hub Thanos: federated aggregates, relabeled back to istio_requests_total
oc --context="${HUB_CONTEXT}" get --raw \
  "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/api/v1/query?query=sum%28istio_requests_total%7Bcluster%3D%22${MANAGED_CLUSTER_NAME}%22%7D%29" \
  | jq '.data.result'

Generate traffic before querying and allow at least one MCOA collection interval. Each query should return a non-empty result. The workload: prefix exists only on the edge recording-rule series; MCOA removes it before writing the metric to hub Thanos. Because the hub sample can lag the edge by the federation interval, compare presence, labels, and approximately corresponding counter values rather than expecting exact point-in-time equality.

List the Istio and Envoy metric names in edge UWM. This output includes the workload:* series produced by the recording rules:

oc --context="${MANAGED_CONTEXT}" \
  -n openshift-user-workload-monitoring \
  exec -c prometheus "${PROM_POD}" -- \
  wget -qO- \
  'http://localhost:9090/api/v1/label/__name__/values' \
  | jq -r '
  [.data[] | select(test("istio|envoy"; "i"))] | unique
  | .[] ,
  "===\nTOTAL COUNT OF istio_ AND envoy_ METRICS: \(length)"'

List the metrics stored in hub Thanos. The result does not contain workload:* names because federation relabels those series, and it should have far fewer metric names than the edge:

oc --context="${HUB_CONTEXT}" get --raw \
  "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/api/v1/label/__name__/values" \
  | jq -r '
  [.data[] | select(test("istio|envoy|pilot"; "i"))] | unique
  | .[] ,
  "===\nTOTAL COUNT OF istio_, envoy_, AND pilot_ METRICS: \(length)"'

The following additional hub queries inspect the backend metric datastore used by ACM.

# List available metric names (Kiali uses istio_*, pilot_*, and envoy_* metrics)
oc --context="${HUB_CONTEXT}" get --raw "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/api/v1/label/__name__/values" | jq -r '.data[] | select(startswith("istio_") or startswith("pilot_") or startswith("envoy_"))'

# Count timeseries for key Istio metrics (shows which metrics have data and how many unique timeseries)
oc --context="${HUB_CONTEXT}" get --raw "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/api/v1/query?query=count%20by%20(__name__)%20({__name__=~%22istio_requests_total|istio_tcp.*total%22})" | jq -r '.data.result[] | "\(.metric.__name__): \(.value[1])"'

# Query Istio request metrics with full details (limited to first result to show structure)
oc --context="${HUB_CONTEXT}" get --raw "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/api/v1/query?query=istio_requests_total" | jq '.data.result |= .[0:1]'

Troubleshooting

Empty Graph or No Metrics

Symptom: Kiali shows an empty graph, “No metrics” in the metrics tab, or both.

Causes and Solutions:

1. scrape_interval too low: If thanos_proxy.scrape_interval is set lower than the ACM collection interval (e.g., “30s” instead of “5m”), Kiali’s rate calculations will use windows too narrow to capture enough data points from Thanos

   • Solution: Set thanos_proxy.scrape_interval to match the ACM collection interval (default “5m”). See Thanos Proxy Mode for details

2. Still in warm-up period: After deploying a new application, it takes approximately twice the ACM collection interval (~10 minutes by default) before enough data points exist for rate calculations

   • Solution: Wait for the warm-up period to elapse

3. MCOA federation resources missing: The recording rules or federation job were not propagated to the managed cluster

   • Solution: On the hub, verify the correctness of the kiali-istio-federation and kiali-istio-platform-federation ScrapeConfig resources and the kiali-istio-aggregation PrometheusRule. Then verify their references under the selected ClusterManagementAddOn placement. On a managed cluster, run oc get prometheusagent,scrapeconfig -A to find MCOA’s configured agent namespace (the default is open-cluster-management-agent-addon) and run oc get prometheusrule -n mesh-observability to verify the propagated rule.

4. PodMonitor missing: Prometheus not scraping Istio data plane components

   • Solution: Create an istio-proxies-monitor-<namespace> PodMonitor in each mesh namespace (including the ztunnel namespace and namespaces with waypoint proxies if using Ambient mode)

5. UWM not enabled: User Workload Monitoring not configured

   • Solution: Enable enableUserWorkload: true in cluster-monitoring-config ConfigMap in openshift-monitoring namespace

6. Missing source/destination labels: The graph builds its topology from workload and namespace labels in the metrics. Verify Istio metrics have proper labels

7. Namespace not selected: Ensure the namespace is selected in the graph’s namespace dropdown

8. Query scope mismatch: Check query_scope cluster names match actual cluster label values

See also the Why is my graph empty? FAQ for additional troubleshooting information.

TLS/Certificate Errors

Symptom: Kiali logs show “x509: certificate signed by unknown authority” or “tls: bad certificate”

Solutions:

1. Verify CA bundle: Ensure kiali-cabundle ConfigMap has the correct CA

   oc get configmap kiali-cabundle -n ${KIALI_NAMESPACE} -o yaml
   

2. Check certificate chain: Verify client cert is signed by expected CA

   oc get secret acm-observability-certs -n ${KIALI_NAMESPACE} \
     -o jsonpath='{.data.tls\.crt}' | base64 -d | \
     openssl x509 -noout -issuer
   

3. Verify projected volume: Check both ConfigMaps are mounted

   oc exec -n ${KIALI_NAMESPACE} deploy/kiali -- ls -la /kiali-cabundle/
   # Should show: additional-ca-bundle.pem, service-ca.crt
   

Connection Refused / Timeout

Symptom: Kiali cannot reach Observatorium API

Solutions:

1. Verify route exists:

   oc get route observatorium-api -n open-cluster-management-observability
   

2. Check ACM is ready (should return “True”):

   oc get mco observability -o jsonpath='{.status.conditions[?(@.type=="Ready")].status}{"\n"}'
   

3. Test connectivity (should return “OK”):

   oc get --raw "/api/v1/namespaces/open-cluster-management-observability/services/http:observability-thanos-query-frontend:9090/proxy/-/ready"
   

4. Check NetworkPolicies: Ensure no policies block egress from Kiali’s namespace

Ambient Mode: No HTTP Metrics

Symptom: Ambient mode workloads show TCP traffic in Kiali but no HTTP metrics (response codes, latency)

Possible causes:

1. No waypoint deployed: Ztunnel only provides L4 (TCP) metrics. Deploy a waypoint proxy for L7 (HTTP) visibility.

2. Missing waypoint PodMonitor: Even with a waypoint, metrics won’t be collected without a PodMonitor:

   • Verify waypoint pod exists: oc get pods -n <namespace> -l gateway.networking.k8s.io/gateway-class-name=istio-waypoint
   • Create PodMonitor in the waypoint’s namespace (same config as sidecar PodMonitor)

3. Missing recording rule for waypoint metrics: Verify that the shared kiali-istio-aggregation rule is present in mesh-observability and that the waypoint namespace has a PodMonitor. The platform ScrapeConfig is cluster-wide and does not need a namespace-specific copy.

Ambient Mode: No Ztunnel Metrics

Symptom: Ambient mode workloads show no traffic at all in Kiali

Possible causes:

1. Missing ztunnel PodMonitor: Create an istio-proxies-monitor-<ztunnel-namespace> PodMonitor in the ztunnel namespace
2. Wrong ztunnel namespace: Verify ztunnel location: oc get pods -l app=ztunnel -A
3. Missing recording rule: Verify that the shared kiali-istio-aggregation rule is present in mesh-observability and that the ztunnel namespace has a PodMonitor. The platform ScrapeConfig is cluster-wide and does not need a namespace-specific copy.

Reference

This example represents a fully configured Kiali installation using ACM Observability via the Observatorium API with mTLS:

apiVersion: kiali.io/v1alpha1
kind: Kiali
metadata:
  name: kiali
  namespace: <kiali-namespace>
spec:
  clustering:
    ignore_home_cluster: true  # External deployment

  kubernetes_config:
    cluster_name: "<management-cluster-name>"

  external_services:
    prometheus:
      url: "<observatorium-api-url>"

      auth:
        type: none
        cert_file: "secret:acm-observability-certs:tls.crt"
        key_file: "secret:acm-observability-certs:tls.key"

      thanos_proxy:
        enabled: true
        retention_period: "365d"
        scrape_interval: "5m"

Required Kubernetes resources:

---
# mTLS client certificates (from ACM)
# Data extracted from Secret observability-grafana-certs in namespace open-cluster-management-observability
apiVersion: v1
kind: Secret
metadata:
  name: acm-observability-certs
  namespace: <kiali-namespace>
type: Opaque
data:
  tls.crt: <base64-encoded-certificate>  # From observability-grafana-certs secret, tls.crt key
  tls.key: <base64-encoded-key>          # From observability-grafana-certs secret, tls.key key

---
# Server CA trust (from ACM)
# Data extracted from Secret observability-client-ca-certs (or observability-server-ca-certs) in namespace open-cluster-management-observability
apiVersion: v1
kind: ConfigMap
metadata:
  name: kiali-cabundle
  namespace: <kiali-namespace>
data:
  additional-ca-bundle.pem: |
    -----BEGIN CERTIFICATE-----
    <ACM Observability CA certificate>  # From ca.crt or tls.crt key (see Step 2 for extraction commands)
    -----END CERTIFICATE-----    

Additional Resources

• Red Hat ACM 2.17 Observability Documentation
• Configuring User Workload Monitoring
• OpenShift Service Mesh Observability
• Istio Standard Metrics
• Troubleshoot Ztunnel Connectivity (Istio Ambient Mode)
• Connecting Grafana to ACM Observability (Red Hat Blog)
• Kiali Multi-cluster Setup
• External Kiali Deployment
• TLS Configuration