Telemetry for backend

The OAP backend cluster itself is a distributed streaming process system. To assist the Ops team, we provide the telemetry for the OAP backend itself, also known as self-observability (so11y)

By default, the telemetry is disabled by setting selector to none, like this:

telemetry:
  selector: ${SW_TELEMETRY:none}
  none:
  prometheus:
    host: ${SW_TELEMETRY_PROMETHEUS_HOST:0.0.0.0}
    port: ${SW_TELEMETRY_PROMETHEUS_PORT:1234}
    sslEnabled: ${SW_TELEMETRY_PROMETHEUS_SSL_ENABLED:false}
    sslKeyPath: ${SW_TELEMETRY_PROMETHEUS_SSL_KEY_PATH:""}
    sslCertChainPath: ${SW_TELEMETRY_PROMETHEUS_SSL_CERT_CHAIN_PATH:""}

You may also set Prometheus to enable them. For more information, refer to the details below.

Self Observability

Static IP or hostname

SkyWalking supports collecting telemetry data into the OAP backend directly. Users could check them out through UI or GraphQL API.

Add the following configuration to enable self-observability-related modules.

Set up prometheus telemetry.

telemetry:
  selector: ${SW_TELEMETRY:prometheus}
  prometheus:
    host: 127.0.0.1
    port: 1543

Set up Prometheus fetcher.

prometheus-fetcher:
  selector: ${SW_PROMETHEUS_FETCHER:default}
  default:
    enabledRules: ${SW_PROMETHEUS_FETCHER_ENABLED_RULES:"self"}

Make sure config/fetcher-prom-rules/self.yaml exists.

Once you deploy an OAP server cluster, the target host should be replaced with a dedicated IP or hostname. For instance, if there are three OAP servers in your cluster, their hosts are service1, service2, and service3, respectively. You should update each self.yaml to switch the target host.

service1:

fetcherInterval: PT15S
fetcherTimeout: PT10S
metricsPath: /metrics
staticConfig:
  # targets will be labeled as "instance"
  targets:
    - service1:1234
  labels:
    service: oap-server
...

service2:

fetcherInterval: PT15S
fetcherTimeout: PT10S
metricsPath: /metrics
staticConfig:
  # targets will be labeled as "instance"
  targets:
    - service2:1234
  labels:
    service: oap-server
...

service3:

fetcherInterval: PT15S
fetcherTimeout: PT10S
metricsPath: /metrics
staticConfig:
  # targets will be labeled as "instance"
  targets:
    - service3:1234
  labels:
    service: oap-server
...

Service discovery on Kubernetes

If you deploy an OAP server cluster on Kubernetes, the oap-server instance (pod) would not have a static IP or hostname. We can leverage OpenTelemetry Collector to discover the oap-server instance, and scrape & transfer the metrics to OAP OpenTelemetry receiver.

On how to install SkyWalking on k8s, you can refer to Apache SkyWalking Kubernetes.

Set this up following these steps:

Set up oap-server.

Set the metrics port.
```
prometheus-port: 1234
```

Set environment variables.

SW_TELEMETRY=prometheus
SW_OTEL_RECEIVER=default
SW_OTEL_RECEIVER_ENABLED_OTEL_RULES=oap

Here is an example to install by Apache SkyWalking Kubernetes:

helm -n istio-system install skywalking skywalking \
             --set elasticsearch.replicas=1 \
             --set elasticsearch.minimumMasterNodes=1 \
             --set elasticsearch.imageTag=7.5.1 \
             --set oap.replicas=2 \
             --set ui.image.repository=$HUB/skywalking-ui \
             --set ui.image.tag=$TAG \
             --set oap.image.tag=$TAG \
             --set oap.image.repository=$HUB/skywalking-oap \
             --set oap.storageType=elasticsearch \
             --set oap.ports.prometheus-port=1234 \ # <<< Expose self observability metrics port
             --set oap.env.SW_TELEMETRY=prometheus \
             --set oap.env.SW_OTEL_RECEIVER=default \ # <<< Enable Otel receiver
             --set oap.env.SW_OTEL_RECEIVER_ENABLED_OTEL_RULES=oap # <<< Add oap analyzer for Otel metrics

Set up OpenTelemetry Collector and config a scrape job:

- job_name: 'skywalking-so11y' # make sure to use this in the so11y.yaml to filter only so11y metrics
  metrics_path: '/metrics'
  kubernetes_sd_configs:
  - role: pod
  relabel_configs:
  - source_labels: [__meta_kubernetes_pod_container_name, __meta_kubernetes_pod_container_port_name]
    action: keep
    regex: oap;prometheus-port
  - source_labels: []
    target_label: service
    replacement: oap-server
  - source_labels: [__meta_kubernetes_pod_name]
    target_label: host_name
    regex: (.+)
    replacement: $$1

For the full example for OpenTelemetry Collector configuration and recommended version, you can refer to showcase.

NOTE: Since Apr 21, 2021, the Grafana project has been relicensed to AGPL-v3, and is no longer licensed for Apache 2.0. Check the LICENSE details. The following Prometheus + Grafana solution is optional rather than recommended.

Prometheus

Prometheus is supported as a telemetry implementor, which collects metrics from SkyWalking’s backend.

Set prometheus to provider. The endpoint opens at http://0.0.0.0:1234/ and http://0.0.0.0:1234/metrics.

telemetry:
  selector: ${SW_TELEMETRY:prometheus}
  prometheus:

Set host and port if needed.

telemetry:
  selector: ${SW_TELEMETRY:prometheus}
  prometheus:
    host: 127.0.0.1
    port: 1543

Set relevant SSL settings to expose a secure endpoint. Note that the private key file and cert chain file could be uploaded once changes are applied to them.

telemetry:
  selector: ${SW_TELEMETRY:prometheus}
  prometheus:
    host: 127.0.0.1
    port: 1543
    sslEnabled: true
    sslKeyPath: /etc/ssl/key.pem
    sslCertChainPath: /etc/ssl/cert-chain.pem

Grafana Visualization

Provide the Grafana dashboard settings. Check SkyWalking OAP Cluster Monitor Dashboard config and SkyWalking OAP Instance Monitor Dashboard config.