Documentation ¶
Overview ¶
Package k8sprocessor allow automatic tagging of spans with k8s metadata.
The processor automatically discovers k8s resources (pods), extracts metadata from them and adds the extracted metadata to the relevant spans. The processor use the kubernetes API to discover all pods running in a cluster, keeps a record of their IP addresses and interesting metadata. Upon receiving spans, the processor tries to identify the source IP address of the service that sent the spans and matches it with the in memory data. If a match is found, the cached metadata is added to the spans as attributes.
RBAC ¶
TODO: mention the required RBAC rules.
Config ¶
TODO: example config.
Deployment scenarios ¶
The processor supports running both in agent and collector mode.
As an agent ¶
When running as an agent, the processor detects IP addresses of pods sending spans to the agent and uses this information to extract metadata from pods and add to spans. When running as an agent, it is important to apply a discovery filter so that the processor only discovers pods from the same host that it is running on. Not using such a filter can result in unnecessary resource usage especially on very large clusters. Once the filter is applied, each processor will only query the k8s API for pods running on it's own node.
Node filter can be applied by setting the `filter.node` config option to the name of a k8s node. While this works as expected, it cannot be used to automatically filter pods by the same node that the processor is running on in most cases as it is not know before hand which node a pod will be scheduled on. Luckily, kubernetes has a solution for this called the downward API. To automatically filter pods by the node the processor is running on, you'll need to complete the following steps:
1. Use the downward API to inject the node name as an environment variable. Add the following snippet under the pod env section of the OpenTelemetry container.
env: - name: KUBE_NODE_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName
This will inject a new environment variable to the OpenTelemetry container with the value as the name of the node the pod was scheduled to run on.
2. Set "filter.node_from_env_var" to the name of the environment variable holding the node name.
k8s_tagger: filter: node_from_env_var: KUBE_NODE_NAME # this should be same as the var name used in previous step
This will restrict each OpenTelemetry agent to query pods running on the same node only dramatically reducing resource requirements for very large clusters.
As a collector ¶
The processor can be deployed both as an agent or as a collector.
When running as a collector, the processor cannot correctly detect the IP address of the pods generating the spans when it receives the spans from an agent instead of receiving them directly from the pods. To workaround this issue, agents deployed with the k8s_tagger processor can be configured to detect the IP addresses and forward them along with the span resources. Collector can then match this IP address with k8s pods and enrich the spans with the metadata. In order to set this up, you'll need to complete the following steps:
1. Setup agents in passthrough mode Configure the agents' k8s_tagger processors to run in passthrough mode.
# k8s_tagger config for agent k8s_tagger: passthrough: true
This will ensure that the agents detect the IP address as add it as an attribute to all span resources. Agents will not make any k8s API calls, do any discovery of pods or extract any metadata.
2. Configure the collector as usual No special configuration changes are needed to be made on the collector. It'll automatically detect the IP address of spans sent by the agents as well as directly by other services/pods.
Caveats ¶
There are some edge-cases and scenarios where k8s_tagger will not work properly.
Host networking mode ¶
The processor cannot correct identify pods running in the host network mode and enriching spans generated by such pods is not supported at the moment.
As a sidecar ¶
The processor does not support detecting containers from the same pods when running as a sidecar. While this can be done, we think it is simpler to just use the kubernetes downward API to inject environment variables into the pods and directly use their values as tags.
Index ¶
- func NewTraceProcessor(logger *zap.Logger, nextConsumer consumer.TraceConsumer, ...) (component.TraceProcessor, error)
- type Config
- type ExtractConfig
- type Factory
- func (f *Factory) CreateDefaultConfig() configmodels.Processor
- func (f *Factory) CreateMetricsProcessor(ctx context.Context, params component.ProcessorCreateParams, ...) (component.MetricsProcessor, error)
- func (f *Factory) CreateTraceProcessor(ctx context.Context, params component.ProcessorCreateParams, ...) (component.TraceProcessor, error)
- func (f *Factory) Type() configmodels.Type
- type FieldExtractConfig
- type FieldFilterConfig
- type FilterConfig
- type Option
- func WithAPIConfig(cfg k8sconfig.APIConfig) Option
- func WithExtractAnnotations(annotations ...FieldExtractConfig) Option
- func WithExtractLabels(labels ...FieldExtractConfig) Option
- func WithExtractMetadata(fields ...string) Option
- func WithFilterFields(filters ...FieldFilterConfig) Option
- func WithFilterLabels(filters ...FieldFilterConfig) Option
- func WithFilterNamespace(ns string) Option
- func WithFilterNode(node, nodeFromEnvVar string) Option
- func WithPassthrough() Option
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func NewTraceProcessor ¶
func NewTraceProcessor( logger *zap.Logger, nextConsumer consumer.TraceConsumer, kubeClient kube.ClientProvider, options ...Option, ) (component.TraceProcessor, error)
NewTraceProcessor returns a component.TraceProcessorOld that adds the WithAttributeMap(attributes) to all spans passed to it.
Types ¶
type Config ¶
type Config struct { configmodels.ProcessorSettings `mapstructure:",squash"` k8sconfig.APIConfig `mapstructure:",squash"` // Passthrough mode only annotates resources with the pod IP and // does not try to extract any other metadata. It does not need // access to the K8S cluster API. Agent/Collector must receive spans // directly from services to be able to correctly detect the pod IPs. Passthrough bool `mapstructure:"passthrough"` // Extract section allows specifying extraction rules to extract // data from k8s pod specs Extract ExtractConfig `mapstructure:"extract"` // Filter section allows specifying filters to filter // pods by labels, fields, namespaces, nodes, etc. Filter FilterConfig `mapstructure:"filter"` }
Config defines configuration for k8s attributes processor.
type ExtractConfig ¶
type ExtractConfig struct { // Metadata allows to extract pod metadata from a list of metadata fields. // The field accepts a list of strings. // // Metadata fields supported right now are, // namespace, podName, podUID, deployment, cluster, node and startTime // // Specifying anything other than these values will result in an error. // By default all of the fields are extracted and added to spans. Metadata []string `mapstructure:"metadata"` // Annotations allows extracting data from pod annotations and record it // as resource attributes. // It is a list of FieldExtractConfig type. See FieldExtractConfig // documentation for more details. Annotations []FieldExtractConfig `mapstructure:"annotations"` // Annotations allows extracting data from pod labels and record it // as resource attributes. // It is a list of FieldExtractConfig type. See FieldExtractConfig // documentation for more details. Labels []FieldExtractConfig `mapstructure:"labels"` }
ExtractConfig section allows specifying extraction rules to extract data from k8s pod specs.
type Factory ¶
type Factory struct { // Factory dependencies that can be provided form outside. KubeClient kube.ClientProvider }
Factory is the factory for Attributes processor.
func (*Factory) CreateDefaultConfig ¶
func (f *Factory) CreateDefaultConfig() configmodels.Processor
CreateDefaultConfig creates the default configuration for processor.
func (*Factory) CreateMetricsProcessor ¶
func (f *Factory) CreateMetricsProcessor( ctx context.Context, params component.ProcessorCreateParams, nextConsumer consumer.MetricsConsumer, cfg configmodels.Processor, ) (component.MetricsProcessor, error)
CreateMetricsProcessor creates a metrics processor based on this config.
func (*Factory) CreateTraceProcessor ¶
func (f *Factory) CreateTraceProcessor( ctx context.Context, params component.ProcessorCreateParams, nextConsumer consumer.TraceConsumer, cfg configmodels.Processor, ) (component.TraceProcessor, error)
CreateTraceProcessor creates a trace processor based on this config.
func (*Factory) Type ¶
func (f *Factory) Type() configmodels.Type
Type gets the type of the config created by this factory.
type FieldExtractConfig ¶
type FieldExtractConfig struct { TagName string `mapstructure:"tag_name"` Key string `mapstructure:"key"` Regex string `mapstructure:"regex"` }
FieldExtractConfig allows specifying an extraction rule to extract a value from exactly one field.
The field accepts a list FilterExtractConfig map. The map accepts three keys
tag-name, key and regex
- tag-name represents the name of the tag that will be added to the span. When not specified a default tag name will be used of the format: k8s.<annotation>.<annotation key> For example, if tag-name is not specified and the key is git_sha, then the span name will be `k8s.annotation.deployment.git_sha`.
- key represents the annotation name. This must exactly match an annotation name.
regex is an optional field used to extract a sub-string from a complex field value. The supplied regular expression must contain one named parameter with the string "value" as the name. For example, if your pod spec contains the following annotation,
kubernetes.io/change-cause: 2019-08-28T18:34:33Z APP_NAME=my-app GIT_SHA=58a1e39 CI_BUILD=4120
and you'd like to extract the GIT_SHA and the CI_BUILD values as tags, then you must specify the following two extraction rules:
procesors: k8s-tagger: annotations:
name: git.sha key: kubernetes.io/change-cause regex: GIT_SHA=(?P<value>\w+)
name: ci.build key: kubernetes.io/change-cause regex: JENKINS=(?P<value>[\w]+)
this will add the `git.sha` and `ci.build` tags to the spans.
type FieldFilterConfig ¶
type FieldFilterConfig struct { // Key represents the key or name of the field or labels that a filter // can apply on. Key string `mapstructure:"key"` // Value represents the value associated with the key that a filter // operation specified by the `Op` field applies on. Value string `mapstructure:"value"` // Op represents the filter operation to apply on the given // Key: Value pair. The following operations are supported // equals, not-equals, exists, does-not-exist. Op string `mapstructure:"op"` }
FieldFilterConfig allows specifying exactly one filter by a field. It can be used to represent a label or generic field filter.
type FilterConfig ¶
type FilterConfig struct { // Node represents a k8s node or host. If specified, any pods not running // on the specified node will be ignored by the tagger. Node string `mapstructure:"node"` // NodeFromEnv can be used to extract the node name from an environment // variable. The value must be the name of the environment variable. // This is useful when the node a Otel agent will run on cannot be // predicted. In such cases, the Kubernetes downward API can be used to // add the node name to each pod as an environment variable. K8s tagger // can then read this value and filter pods by it. // // For example, node name can be passed to each agent with the downward API as follows // // env: // - name: K8S_NODE_NAME // valueFrom: // fieldRef: // fieldPath: spec.nodeName // // Then the NodeFromEnv field can be set to `K8S_NODE_NAME` to filter all pods by the node that // the agent is running on. // // More on downward API here: https://kubernetes.io/docs/tasks/inject-data-application/downward-api-volume-expose-pod-information/ NodeFromEnvVar string `mapstructure:"node_from_env_var"` // Namespace filters all pods by the provided namespace. All other pods are ignored. Namespace string `mapstructure:"namespace"` // Fields allows to filter pods by generic k8s fields. // Only the following operations are supported: // - equals // - not-equals // // Check FieldFilterConfig for more details. Fields []FieldFilterConfig `mapstructure:"fields"` // Labels allows to filter pods by generic k8s pod labels. // Only the following operations are supported: // - equals // - not-equals // - exists // - not-exists // // Check FieldFilterConfig for more details. Labels []FieldFilterConfig `mapstructure:"labels"` }
FilterConfig section allows specifying filters to filter pods by labels, fields, namespaces, nodes, etc.
type Option ¶
type Option func(*kubernetesprocessor) error
Option represents a configuration option that can be passes. to the k8s-tagger
func WithAPIConfig ¶
WithAPIConfig provides k8s API related configuration to the processor. It defaults the authentication method to in-cluster auth using service accounts.
func WithExtractAnnotations ¶
func WithExtractAnnotations(annotations ...FieldExtractConfig) Option
WithExtractAnnotations allows specifying options to control extraction of pod annotations tags.
func WithExtractLabels ¶
func WithExtractLabels(labels ...FieldExtractConfig) Option
WithExtractLabels allows specifying options to control extraction of pod labels.
func WithExtractMetadata ¶
WithExtractMetadata allows specifying options to control extraction of pod metadata. If no fields explicitly provided, all metadata extracted by default.
func WithFilterFields ¶
func WithFilterFields(filters ...FieldFilterConfig) Option
WithFilterFields allows specifying options to control filtering pods by pod fields.
func WithFilterLabels ¶
func WithFilterLabels(filters ...FieldFilterConfig) Option
WithFilterLabels allows specifying options to control filtering pods by pod labels.
func WithFilterNamespace ¶
WithFilterNamespace allows specifying options to control filtering pods by a namespace.
func WithFilterNode ¶
WithFilterNode allows specifying options to control filtering pods by a node/host.
func WithPassthrough ¶
func WithPassthrough() Option
WithPassthrough enables passthrough mode. In passthrough mode, the processor only detects and tags the pod IP and does not invoke any k8s APIs.