Documentation ¶
Overview ¶
Package metrics holds prometheus metrics objects and related utility functions. It does not abstract away the prometheus client but the caller rarely needs to refer to prometheus directly.
Index ¶
- Constants
- Variables
- func BoolToFloat64(v bool) float64
- func DefaultMetrics() map[string]struct{}
- func DumpMetrics() ([]*models.Metric, error)
- func Enable(addr string) <-chan error
- func Errno2Outcome(errno unix.Errno) string
- func Error2Outcome(err error) string
- func GetCounterValue(m prometheus.Counter) float64
- func GetGaugeValue(m prometheus.Gauge) float64
- func MustRegister(c ...prometheus.Collector)
- func Register(c prometheus.Collector) error
- func RegisterList(list []prometheus.Collector) error
- func ResetMetrics()
- func Unregister(c prometheus.Collector) bool
- type APIEventTSHelper
- type Configuration
- type CounterVec
- type GaugeVec
- type GaugeWithThreshold
- type LoggingHook
Constants ¶
const ( // ErrorTimeout is the value used to notify timeout errors. ErrorTimeout = "timeout" // ErrorProxy is the value used to notify errors on Proxy. ErrorProxy = "proxy" // L7DNS is the value used to report DNS label on metrics L7DNS = "dns" // SubsystemBPF is the subsystem to scope metrics related to the bpf syscalls. SubsystemBPF = "bpf" // SubsystemDatapath is the subsystem to scope metrics related to management of // the datapath. It is prepended to metric names and separated with a '_'. SubsystemDatapath = "datapath" // SubsystemAgent is the subsystem to scope metrics related to the cilium agent itself. SubsystemAgent = "agent" // SubsystemFQDN is the subsystem to scope metrics related to the FQDN proxy. SubsystemIPCache = "ipcache" // SubsystemK8s is the subsystem to scope metrics related to Kubernetes SubsystemK8s = "k8s" // SubsystemK8sClient is the subsystem to scope metrics related to the kubernetes client. SubsystemK8sClient = "k8s_client" // SubsystemKVStore is the subsystem to scope metrics related to the kvstore. SubsystemKVStore = "kvstore" // SubsystemFQDN is the subsystem to scope metrics related to the FQDN proxy. SubsystemFQDN = "fqdn" // SubsystemNodes is the subsystem to scope metrics related to the node manager. SubsystemNodes = "nodes" // SubsystemTriggers is the subsystem to scope metrics related to the trigger package. SubsystemTriggers = "triggers" // SubsystemAPILimiter is the subsystem to scope metrics related to the API limiter package. SubsystemAPILimiter = "api_limiter" // SubsystemNodeNeigh is the subsystem to scope metrics related to management of node neighbor. SubsystemNodeNeigh = "node_neigh" // Namespace is used to scope metrics from cilium. It is prepended to metric // names and separated with a '_' Namespace = "cilium" // LabelError indicates the type of error (string) LabelError = "error" // LabelOutcome indicates whether the outcome of the operation was successful or not LabelOutcome = "outcome" // LabelAttempts is the number of attempts it took to complete the operation LabelAttempts = "attempts" // LabelValueOutcomeSuccess is used as a successful outcome of an operation LabelValueOutcomeSuccess = "success" // LabelValueOutcomeFail is used as an unsuccessful outcome of an operation LabelValueOutcomeFail = "fail" // LabelEventSourceAPI marks event-related metrics that come from the API LabelEventSourceAPI = "api" // LabelEventSourceK8s marks event-related metrics that come from k8s LabelEventSourceK8s = "k8s" // LabelEventSourceFQDN marks event-related metrics that come from pkg/fqdn LabelEventSourceFQDN = "fqdn" // LabelEventSourceContainerd marks event-related metrics that come from docker LabelEventSourceContainerd = "docker" // LabelDatapathArea marks which area the metrics are related to (eg, which BPF map) LabelDatapathArea = "area" // LabelDatapathName marks a unique identifier for this metric. // The name should be defined once for a given type of error. LabelDatapathName = "name" // LabelDatapathFamily marks which protocol family (IPv4, IPV6) the metric is related to. LabelDatapathFamily = "family" // LabelProtocol marks the L4 protocol (TCP, ANY) for the metric. LabelProtocol = "protocol" // LabelSignalType marks the signal name LabelSignalType = "signal" // LabelSignalData marks the signal data LabelSignalData = "data" // LabelStatus the label from completed task LabelStatus = "status" // LabelPolicyEnforcement is the label used to see the enforcement status LabelPolicyEnforcement = "enforcement" // LabelPolicySource is the label used to see the enforcement status LabelPolicySource = "source" // LabelScope is the label used to defined multiples scopes in the same // metric. For example, one counter may measure a metric over the scope of // the entire event (scope=global), or just part of an event // (scope=slow_path) LabelScope = "scope" // LabelProtocolL7 is the label used when working with layer 7 protocols. LabelProtocolL7 = "protocol_l7" // LabelBuildState is the state a build queue entry is in LabelBuildState = "state" // LabelBuildQueueName is the name of the build queue LabelBuildQueueName = "name" // LabelAction is the label used to defined what kind of action was performed in a metric LabelAction = "action" // LabelSubsystem is the label used to refer to any of the child process // started by cilium (Envoy, monitor, etc..) LabelSubsystem = "subsystem" // LabelKind is the kind of a label LabelKind = "kind" // LabelEventSource is the source of a label for event metrics // i.e. k8s, containerd, api. LabelEventSource = "source" // LabelPath is the label for the API path LabelPath = "path" // LabelMethod is the label for the HTTP method LabelMethod = "method" // LabelAPIReturnCode is the HTTP code returned for that API path LabelAPIReturnCode = "return_code" // LabelOperation is the label for BPF maps operations LabelOperation = "operation" // LabelMapName is the label for the BPF map name LabelMapName = "map_name" // LabelVersion is the label for the version number LabelVersion = "version" // LabelDirection is the label for traffic direction LabelDirection = "direction" // LabelSourceCluster is the label for source cluster name LabelSourceCluster = "source_cluster" // LabelSourceNodeName is the label for source node name LabelSourceNodeName = "source_node_name" // LabelTargetCluster is the label for target cluster name LabelTargetCluster = "target_cluster" // LabelTargetNodeIP is the label for target node IP LabelTargetNodeIP = "target_node_ip" // LabelTargetNodeName is the label for target node name LabelTargetNodeName = "target_node_name" // LabelTargetNodeType is the label for target node type (local_node, remote_intra_cluster, vs remote_inter_cluster) LabelTargetNodeType = "target_node_type" LabelLocationLocalNode = "local_node" LabelLocationRemoteIntraCluster = "remote_intra_cluster" LabelLocationRemoteInterCluster = "remote_inter_cluster" // LabelType is the label for type in general (e.g. endpoint, node) LabelType = "type" LabelPeerEndpoint = "endpoint" LabelPeerNode = "node" LabelTrafficHTTP = "http" LabelTrafficICMP = "icmp" LabelAddressType = "address_type" LabelAddressTypePrimary = "primary" LabelAddressTypeSecondary = "secondary" )
Variables ¶
var ( NoOpMetric prometheus.Metric = &metric{} NoOpCollector prometheus.Collector = &collector{} NoOpCounter prometheus.Counter = &counter{NoOpMetric, NoOpCollector} NoOpCounterVec CounterVec = &counterVec{NoOpCollector} NoOpObserver prometheus.Observer = &observer{} NoOpObserverVec prometheus.ObserverVec = &observerVec{NoOpCollector} NoOpGauge prometheus.Gauge = &gauge{NoOpMetric, NoOpCollector} NoOpGaugeVec GaugeVec = &gaugeVec{NoOpCollector} )
var ( // APIInteractions is the total time taken to process an API call made // to the cilium-agent APIInteractions = NoOpObserverVec // NodeConnectivityStatus is the connectivity status between local node to // other node intra or inter cluster. NodeConnectivityStatus = NoOpGaugeVec // NodeConnectivityLatency is the connectivity latency between local node to // other node intra or inter cluster. NodeConnectivityLatency = NoOpGaugeVec // Endpoint is a function used to collect this metric. // It must be thread-safe. Endpoint prometheus.GaugeFunc // EndpointRegenerationTotal is a count of the number of times any endpoint // has been regenerated and success/fail outcome EndpointRegenerationTotal = NoOpCounterVec // EndpointStateCount is the total count of the endpoints in various states. EndpointStateCount = NoOpGaugeVec // EndpointRegenerationTimeStats is the total time taken to regenerate // endpoints, labeled by span name and status ("success" or "failure") EndpointRegenerationTimeStats = NoOpObserverVec // EndpointPropagationDelay is the delay between creation of local CiliumEndpoint // and update for that CiliumEndpoint received through CiliumEndpointSlice. // Measure of local CEP roundtrip time with CiliumEndpointSlice feature enabled. EndpointPropagationDelay = NoOpObserverVec // Policy // Policy is the number of policies loaded into the agent Policy = NoOpGauge // PolicyRegenerationCount is the total number of successful policy // regenerations. PolicyRegenerationCount = NoOpCounter // PolicyRegenerationTimeStats is the total time taken to generate policies PolicyRegenerationTimeStats = NoOpObserverVec // PolicyRevision is the current policy revision number for this agent PolicyRevision = NoOpGauge // PolicyImportErrorsTotal is a count of failed policy imports PolicyImportErrorsTotal = NoOpCounter // PolicyEndpointStatus is the number of endpoints with policy labeled by enforcement type PolicyEndpointStatus = NoOpGaugeVec // PolicyImplementationDelay is a distribution of times taken from adding a // policy (and incrementing the policy revision) to seeing it in the datapath // per Endpoint. This reflects the actual delay perceived by traffic flowing // through the datapath. The longest times will roughly correlate with the // time taken to fully deploy an endpoint. PolicyImplementationDelay = NoOpObserverVec // Identity is the number of identities currently in use on the node by type Identity = NoOpGaugeVec // EventTS is the timestamp of k8s resource events. EventTS = NoOpGaugeVec // EventLagK8s is the lag calculation for k8s Pod events. EventLagK8s = NoOpGauge // ProxyRedirects is the number of redirects labeled by protocol ProxyRedirects = NoOpGaugeVec // ProxyPolicyL7Total is a count of all l7 requests handled by proxy ProxyPolicyL7Total = NoOpCounterVec // ProxyParseErrors is a count of failed parse errors on proxy // Deprecated: in favor of ProxyPolicyL7Total ProxyParseErrors = NoOpCounter // ProxyForwarded is a count of all forwarded requests by proxy // Deprecated: in favor of ProxyPolicyL7Total ProxyForwarded = NoOpCounter // ProxyDenied is a count of all denied requests by policy by the proxy // Deprecated: in favor of ProxyPolicyL7Total ProxyDenied = NoOpCounter // ProxyReceived is a count of all received requests by the proxy // Deprecated: in favor of ProxyPolicyL7Total ProxyReceived = NoOpCounter // ProxyUpstreamTime is how long the upstream server took to reply labeled // by error, protocol and span time ProxyUpstreamTime = NoOpObserverVec // ProxyDatapathUpdateTimeout is a count of all the timeouts encountered while // updating the datapath due to an FQDN IP update ProxyDatapathUpdateTimeout = NoOpCounter // DropCount is the total drop requests, // tagged by drop reason and direction(ingress/egress) DropCount = NoOpCounterVec // DropBytes is the total dropped bytes, // tagged by drop reason and direction(ingress/egress) DropBytes = NoOpCounterVec // ForwardCount is the total forwarded packets, // tagged by ingress/egress direction ForwardCount = NoOpCounterVec // ForwardBytes is the total forwarded bytes, // tagged by ingress/egress direction ForwardBytes = NoOpCounterVec // ConntrackGCRuns is the number of times that the conntrack GC // process was run. ConntrackGCRuns = NoOpCounterVec // ConntrackGCKeyFallbacks number of times that the conntrack key fallback was invalid. ConntrackGCKeyFallbacks = NoOpCounterVec // ConntrackGCSize the number of entries in the conntrack table ConntrackGCSize = NoOpGaugeVec // NatGCSize the number of entries in the nat table NatGCSize = NoOpGaugeVec // ConntrackGCDuration the duration of the conntrack GC process in milliseconds. ConntrackGCDuration = NoOpObserverVec // ConntrackDumpReset marks the count for conntrack dump resets ConntrackDumpResets = NoOpCounterVec // SignalsHandled is the number of signals received. SignalsHandled = NoOpCounterVec // ServicesCount number of services ServicesCount = NoOpCounterVec // ErrorsWarnings is the number of errors and warnings in cilium-agent instances ErrorsWarnings = NoOpCounterVec // ControllerRuns is the number of times that a controller process runs. ControllerRuns = NoOpCounterVec // ControllerRunsDuration the duration of the controller process in seconds ControllerRunsDuration = NoOpObserverVec // subprocess, labeled by Subsystem SubprocessStart = NoOpCounterVec // KubernetesEventProcessed is the number of Kubernetes events // processed labeled by scope, action and execution result KubernetesEventProcessed = NoOpCounterVec // KubernetesEventReceived is the number of Kubernetes events received // labeled by scope, action, valid data and equalness. KubernetesEventReceived = NoOpCounterVec // KubernetesAPIInteractions is the total time taken to process an API call made // to the kube-apiserver KubernetesAPIInteractions = NoOpObserverVec // KubernetesAPICallsTotal is the counter for all API calls made to // kube-apiserver. KubernetesAPICallsTotal = NoOpCounterVec // KubernetesCNPStatusCompletion is the number of seconds it takes to // complete a CNP status update KubernetesCNPStatusCompletion = NoOpObserverVec // TerminatingEndpointsEvents is the number of terminating endpoint events received from kubernetes. TerminatingEndpointsEvents = NoOpCounter // IpamEvent is the number of IPAM events received labeled by action and // datapath family type IpamEvent = NoOpCounterVec // KVStoreOperationsDuration records the duration of kvstore operations KVStoreOperationsDuration = NoOpObserverVec // KVStoreEventsQueueDuration records the duration in seconds of time // received event was blocked before it could be queued KVStoreEventsQueueDuration = NoOpObserverVec // KVStoreQuorumErrors records the number of kvstore quorum errors KVStoreQuorumErrors = NoOpCounterVec // FQDNGarbageCollectorCleanedTotal is the number of domains cleaned by the // GC job. FQDNGarbageCollectorCleanedTotal = NoOpCounter // FQDNActiveNames is the number of domains inside the DNS cache that have // not expired (by TTL), per endpoint. FQDNActiveNames = NoOpGaugeVec // FQDNActiveIPs is the number of IPs inside the DNS cache associated with // a domain that has not expired (by TTL) and are currently active, per // endpoint. FQDNActiveIPs = NoOpGaugeVec // FQDNAliveZombieConnections is the number IPs associated with domains // that have expired (by TTL) yet still associated with an active // connection (aka zombie), per endpoint. FQDNAliveZombieConnections = NoOpGaugeVec // FQDNSemaphoreRejectedTotal is the total number of DNS requests rejected // by the DNS proxy because too many requests were in flight, as enforced by // the admission semaphore. FQDNSemaphoreRejectedTotal = NoOpCounter // IPCacheErrorsTotal is the total number of IPCache events handled in // the IPCache subsystem that resulted in errors. IPCacheErrorsTotal = NoOpCounterVec // IPCacheEventsTotal is the total number of IPCache events handled in // the IPCache subsystem. IPCacheEventsTotal = NoOpCounterVec // BPFSyscallDuration is the metric for bpf syscalls duration. BPFSyscallDuration = NoOpObserverVec // BPFMapOps is the metric to measure the number of operations done to a // bpf map. BPFMapOps = NoOpCounterVec // TriggerPolicyUpdateTotal is the metric to count total number of // policy update triggers TriggerPolicyUpdateTotal = NoOpCounterVec // TriggerPolicyUpdateFolds is the current level folding that is // happening when running policy update triggers TriggerPolicyUpdateFolds = NoOpGauge // TriggerPolicyUpdateCallDuration measures the latency and call // duration of policy update triggers TriggerPolicyUpdateCallDuration = NoOpObserverVec // VersionMetric labelled by Cilium version VersionMetric = NoOpGaugeVec // APILimiterWaitHistoryDuration is a histogram that measures the // individual wait durations of API limiters APILimiterWaitHistoryDuration = NoOpObserverVec // APILimiterWaitDuration is the gauge of the current mean, min, and // max wait duration APILimiterWaitDuration = NoOpGaugeVec // APILimiterProcessingDuration is the gauge of the mean and estimated // processing duration APILimiterProcessingDuration = NoOpGaugeVec // APILimiterRequestsInFlight is the gauge of the current and max // requests in flight APILimiterRequestsInFlight = NoOpGaugeVec // APILimiterRateLimit is the gauge of the current rate limiting // configuration including limit and burst APILimiterRateLimit = NoOpGaugeVec // APILimiterAdjustmentFactor is the gauge representing the latest // adjustment factor that was applied APILimiterAdjustmentFactor = NoOpGaugeVec // APILimiterProcessedRequests is the counter of the number of // processed (successful and failed) requests APILimiterProcessedRequests = NoOpCounterVec // ArpingRequestsTotal is the counter of the number of sent // (successful and failed) arping requests ArpingRequestsTotal = NoOpCounterVec )
Functions ¶
func BoolToFloat64 ¶
func DefaultMetrics ¶ added in v1.5.1
func DefaultMetrics() map[string]struct{}
func DumpMetrics ¶
DumpMetrics gets the current Cilium metrics and dumps all into a models.Metrics structure.If metrics cannot be retrieved, returns an error
func Enable ¶ added in v1.5.0
Enable begins serving prometheus metrics on the address passed in. Addresses of the form ":8080" will bind the port on all interfaces.
func Errno2Outcome ¶
Errno2Outcome converts a unix.Errno to LabelOutcome
func Error2Outcome ¶
Error2Outcome converts an error to LabelOutcome
func GetCounterValue ¶
func GetCounterValue(m prometheus.Counter) float64
GetCounterValue returns the current value stored for the counter
func GetGaugeValue ¶
func GetGaugeValue(m prometheus.Gauge) float64
GetGaugeValue returns the current value stored for the gauge. This function is useful in tests.
func MustRegister ¶ added in v1.5.0
func MustRegister(c ...prometheus.Collector)
MustRegister adds the collector to the registry, exposing this metric to prometheus scrapes. It will panic on error.
func RegisterList ¶
func RegisterList(list []prometheus.Collector) error
RegisterList registers a list of collectors. If registration of one collector fails, no collector is registered.
func ResetMetrics ¶ added in v1.13.0
func ResetMetrics()
Types ¶
type APIEventTSHelper ¶
type APIEventTSHelper struct { Next http.Handler TSGauge GaugeVec Histogram prometheus.ObserverVec }
APIEventTSHelper is intended to be a global middleware to track metrics around API calls. It records the timestamp of an API call in the provided gauge.
func (*APIEventTSHelper) ServeHTTP ¶
func (m *APIEventTSHelper) ServeHTTP(r http.ResponseWriter, req *http.Request)
ServeHTTP implements the http.Handler interface. It records the timestamp this API call began at, then chains to the next handler.
type Configuration ¶ added in v1.5.1
type Configuration struct { APIInteractionsEnabled bool NodeConnectivityStatusEnabled bool NodeConnectivityLatencyEnabled bool EndpointRegenerationCountEnabled bool EndpointStateCountEnabled bool EndpointRegenerationTimeStatsEnabled bool EndpointPropagationDelayEnabled bool PolicyCountEnabled bool PolicyRegenerationCountEnabled bool PolicyRegenerationTimeStatsEnabled bool PolicyRevisionEnabled bool PolicyImportErrorsEnabled bool PolicyEndpointStatusEnabled bool PolicyImplementationDelayEnabled bool IdentityCountEnabled bool EventTSEnabled bool EventLagK8sEnabled bool EventTSContainerdEnabled bool EventTSAPIEnabled bool ProxyRedirectsEnabled bool ProxyPolicyL7Enabled bool ProxyParseErrorsEnabled bool ProxyForwardedEnabled bool ProxyDeniedEnabled bool ProxyReceivedEnabled bool ProxyDatapathUpdateTimeoutEnabled bool NoOpObserverVecEnabled bool DropCountEnabled bool DropBytesEnabled bool NoOpCounterVecEnabled bool ForwardBytesEnabled bool ConntrackGCRunsEnabled bool ConntrackGCKeyFallbacksEnabled bool ConntrackGCSizeEnabled bool ConntrackGCDurationEnabled bool ConntrackDumpResetsEnabled bool SignalsHandledEnabled bool ServicesCountEnabled bool ErrorsWarningsEnabled bool ControllerRunsEnabled bool ControllerRunsDurationEnabled bool SubprocessStartEnabled bool KubernetesEventProcessedEnabled bool KubernetesEventReceivedEnabled bool KubernetesTimeBetweenEventsEnabled bool KubernetesAPIInteractionsEnabled bool KubernetesAPICallsEnabled bool KubernetesCNPStatusCompletionEnabled bool KubernetesTerminatingEndpointsEnabled bool IpamEventEnabled bool IPCacheErrorsTotalEnabled bool IPCacheEventsTotalEnabled bool KVStoreOperationsDurationEnabled bool KVStoreEventsQueueDurationEnabled bool KVStoreQuorumErrorsEnabled bool FQDNGarbageCollectorCleanedTotalEnabled bool FQDNActiveNames bool FQDNActiveIPs bool FQDNActiveZombiesConnections bool FQDNSemaphoreRejectedTotal bool BPFSyscallDurationEnabled bool BPFMapOps bool BPFMapPressure bool TriggerPolicyUpdateTotal bool TriggerPolicyUpdateFolds bool TriggerPolicyUpdateCallDuration bool VersionMetric bool APILimiterWaitHistoryDuration bool APILimiterWaitDuration bool APILimiterProcessingDuration bool APILimiterRequestsInFlight bool APILimiterRateLimit bool APILimiterAdjustmentFactor bool APILimiterProcessedRequests bool ArpingRequestsTotalEnabled bool }
func CreateConfiguration ¶ added in v1.5.1
func CreateConfiguration(metricsEnabled []string) (Configuration, []prometheus.Collector)
CreateConfiguration returns a Configuration with all metrics that are considered enabled from the given slice of metricsEnabled as well as a slice of prometheus.Collectors that must be registered in the prometheus default register.
type CounterVec ¶ added in v1.5.1
type CounterVec interface { WithLabelValues(lvls ...string) prometheus.Counter GetMetricWithLabelValues(lvs ...string) (prometheus.Counter, error) With(labels prometheus.Labels) prometheus.Counter prometheus.Collector }
type GaugeVec ¶ added in v1.5.1
type GaugeVec interface { WithLabelValues(lvls ...string) prometheus.Gauge prometheus.Collector }
type GaugeWithThreshold ¶
type GaugeWithThreshold struct {
// contains filtered or unexported fields
}
GaugeWithThreshold is a prometheus gauge that registers itself with prometheus if over a threshold value and unregisters when under.
func NewBPFMapPressureGauge ¶
func NewBPFMapPressureGauge(mapname string, threshold float64) *GaugeWithThreshold
NewBPFMapPressureGauge creates a new GaugeWithThreshold for the cilium_bpf_map_pressure metric with the map name as constant label.
func NewGaugeWithThreshold ¶
func NewGaugeWithThreshold(name string, subsystem string, desc string, labels map[string]string, threshold float64) *GaugeWithThreshold
NewGaugeWithThreshold creates a new GaugeWithThreshold.
func (*GaugeWithThreshold) Set ¶
func (gwt *GaugeWithThreshold) Set(value float64)
Set the value of the GaugeWithThreshold.
type LoggingHook ¶
type LoggingHook struct {
// contains filtered or unexported fields
}
LoggingHook is a hook for logrus which counts error and warning messages as a Prometheus metric.
func NewLoggingHook ¶
func NewLoggingHook(component string) *LoggingHook
NewLoggingHook returns a new instance of LoggingHook for the given Cilium component.
func (*LoggingHook) Fire ¶
func (h *LoggingHook) Fire(entry *logrus.Entry) error
Fire is the main method which is called every time when logger has an error or warning message.
func (*LoggingHook) Levels ¶
func (h *LoggingHook) Levels() []logrus.Level
Levels returns the list of logging levels on which the hook is triggered.