Kubernetes (K8s) monitoring from Rover

SkyWalking uses the Cilium Fetcher to gather traffic data between services from Cilium Hubble via the Observe API. It then leverages the OAL System for metrics and entity analysis.

Data flow

SkyWalking fetches Cilium Node and Observability Data from gRPC API, analysis to generate entity and using OAL to generating metrics.

API Requirements

Peers API: Listen the hubble node in the cluster, OAP would communicate with Hubble node to obtain Observe data.
Observe API: Fetch the Flow data from Hubble node.

Setup

Please following the Setup Hubble Observability documentation to setting the Hubble for provided API.
To activate Cilium receiver module, set selector=default in the YAML or set SW_CILIUM_FETCHER=default through the system environment variable.

cilium-fetcher:
    selector: ${SW_CILIUM_FETCHER:default}
    default:
        # Host name and port of Hubble peer component
        peerHost: ${SW_CILIUM_FETCHER_PEER_HOST:hubble-peer.kube-system.svc.cluster.local}
        peerPort: ${SW_CILIUM_FETCHER_PEER_PORT:80}
        fetchFailureRetrySecond: ${SW_CILIUM_FETCHER_FETCH_FAILURE_RETRY_SECOND:10}
        sslConnection: ${SW_CILIUM_FETCHER_SSL_CONNECTION:false}
        sslPrivateKeyFile: ${SW_CILIUM_FETCHER_PRIVATE_KEY_FILE_PATH:}
        sslCertChainFile: ${SW_CILIUM_FETCHER_CERT_CHAIN_FILE_PATH:}
        sslCaFile: ${SW_CILIUM_FETCHER_CA_FILE_PATH:}
        convertClientAsServerTraffic: ${SW_CILIUM_FETCHER_CONVERT_CLIENT_AS_SERVER_TRAFFIC:true}

If enabled the TLS certificate within the Hubble, please update these few configurations.
1. peerPort: usually should be updated to the 443.
2. sslConnection: should be set to true.
3. sslPrivateKeyFile: the path of the private key file.
4. sslCertChainFile: the path of the certificate chain file.
5. sslCaFile: the path of the CA file.
Configure the cilium rules please configure the following configuration:
1. cilium-rules/exclude.yaml: Configure the which endpoint should be excluded from the monitoring, Please read exclude rules selection for more detail.
2. cilium-rules/metadata-service-mapping.yaml: Configure the service name and endpoint mapping.

Exclude Rules

The exclude configuration in Cilium rules is used to specify which Cilium Endpoints would be excluded from being added to the topology map or from the generation of metrics and other data.

namespaces:   # define with traffic from which namespace should be excluded
   - kube-system

labels:       # define with traffic from which endpoint labels should be excluded, if matches any labels, the traffic would be excluded.
   - k8s:io.cilium.k8s.namespace.labels.istio-injection: "enabled" # Each labels is a key-value pair, the key is the label key, the value is the label value.
     k8s:security.istio.io/tlsMode: istio

By default, all the traffic from kube-system and traffic management by istio mesh would be excluded.

NOTE: Only the endpoint in both source and destination matches the exclude rules would be excluded. Otherwise, the traffic would be still included.

Generated Entities

SkyWalking fetch the flow from Cilium, analyzes the source and destination endpoint to parse out the following corresponding entities:

Service
Service Instance
Service Endpoint
Service Relation
Service Instance Relation
Service Endpoint Relation

Generate Metrics

For each of the above-mentioned entities, metrics such as L4 and L7 protocols can be analyzed.

L4 Metrics

Record the relevant metrics for every service read/write packages with other services.

Name	Unit	Description
Read Package CPM	Count	Total Read Package from other Service counts per minutes.
Write Package CPM	Count	Total Write Package from other Service counts per minutes.
Drop Package CPM	Count	Total Drop Package from other Service counts per minutes.
Drop Package Reason Count	Labeled Count	Total Read Package reason(labeled) from other Service counts per minutes.

Protocol

Based on each transfer data analysis, extract the information of the 7-layer network protocol.

NOTE: By default, Cilium only reports L4 metrics. If you need L7 metrics, they must be explicitly specified in each service’s CiliumNetworkPolicy. For details please refer to this document.

HTTP

Name	Unit	Description
CPM	Count	HTTP Request calls per minutes.
Duration	Nanoseconds	Total HTTP Response use duration.
Success CPM	Count	Total HTTP Response success(status < 500) count.
Status 1/2/3/4/5xx	Count	HTTP Response status code group by 1xx/2xx/3xx/4xx/5xx.

DNS

Name	Unit	Description
CPM	Count	DNS Request calls per minutes.
Duration	Nanoseconds	Total DNS Response use duration.
Success CPM	Count	Total DNS Response success(code == 0) count.
Error Count	Label Count	DNS Response error count with error description label.

Kafka

Name	Unit	Description
CPM	Count	Kafka Request calls per minutes.
Duration	Nanoseconds	Total Kafka Response use duration.
Success CPM	Count	Total Kafka Response success(errorCode == 0) count.
Error Count	Label Count	Kafka Response error count with error description label.

Load Balance for Cilium Fetcher with OAP cluster

The Cilium Fetcher module relies on the Cluster module, when the Cilium Fetcher module starts up, it obtains information about all Cilium nodes and node information in the OAP cluster through Peers API on each OAP node.

Additionally, it averagely distributes collected Cilium nodes to every OAP node. Moreover, it ensures that a single Cilium node is not monitored by multiple OAP nodes.

Customizations

You can customize your own metrics/dashboard panel. The metrics definition and expression rules are found in /config/oal/cilium.oal, please refer the Scope Declaration Documentation. The Cilium dashboard panel configurations are found in /config/ui-initialized-templates/cilium_service.

Edit this page

SkyWalking