import "github.com/apache/beam/sdks/go/pkg/beam/core/metrics"
Package metrics implements the Beam metrics API, described at http://s.apache.org/beam-metrics-api
Metrics in the Beam model are uniquely identified by a namespace, a name, and the PTransform context in which they are used. Further, they are reported as a delta against the bundle being processed, so that overcounting doesn't occur if a bundle needs to be retried. Each metric is scoped to their bundle, and ptransform.
Cells (or metric cells) are defined for each Beam model metric type, and the serve as concurrency safe storage of a given metric's values. Proxys are exported values representing the metric, for use in user ptransform code. They don't retain their cells, since they don't have the context to be able to store them for export back to the pipeline runner.
Metric cells aren't initialized until their first mutation, which follows from the Beam model design, where metrics are only sent for a bundle if they have changed. This is particularly convenient for distributions which means their min and max fields can be set to the first value on creation rather than have some marker of uninitialized state, which would otherwise need to be checked for on every update.
Metric values are implemented as lightweight proxies of the user provided namespace and name. This allows them to be declared globally, and used in any ParDo. Further, as per the design, they can be declared dynamically at runtime.
To handle reporting deltas on the metrics by bundle, metrics are keyed by bundleID,PTransformID,namespace, and name, so metrics that are identical except for bundles are treated as distinct, effectively providing per bundle deltas, since a new value cell is used per bundle.
DumpToLog is a debugging function that outputs all metrics available locally to beam.Log.
DumpToLogFromStore dumps the metrics in the provided Store to beam.Log.
DumpToOutFromContext is a debugging function that outputs all metrics available locally to std out, extracting the metric store from the context.
DumpToOutFromStore is a debugging function that outputs all metrics available locally to std out directly from the store.
SetBundleID sets the id of the current Bundle, and populates the store.
SetPTransformID sets the id of the current PTransform. Must only be called on a context returned by SetBundleID.
type Counter struct {
// contains filtered or unexported fields
}
Counter is a simple counter for incrementing and decrementing a value.
NewCounter returns the Counter with the given namespace and name.
Dec decrements the counter within the given PTransform context by v.
Inc increments the counter within the given PTransform context by v.
CounterResult is an attempted and a commited value of a counter metric plus key.
MergeCounters combines counter metrics that share a common key.
func (r CounterResult) Result() int64
Result returns committed metrics. Falls back to attempted metrics if committed are not populated (e.g. due to not being supported on a given runner).
type Distribution struct {
// contains filtered or unexported fields
}
Distribution is a simple distribution of values.
func NewDistribution(ns, n string) *Distribution
NewDistribution returns the Distribution with the given namespace and name.
func (m *Distribution) String() string
func (m *Distribution) Update(ctx context.Context, v int64)
Update updates the distribution within the given PTransform context with v.
type DistributionResult struct { Attempted, Committed DistributionValue Key StepKey }
DistributionResult is an attempted and a commited value of a distribution metric plus key.
func MergeDistributions( attempted map[StepKey]DistributionValue, committed map[StepKey]DistributionValue) []DistributionResult
MergeDistributions combines distribution metrics that share a common key.
func (r DistributionResult) Result() DistributionValue
Result returns committed metrics. Falls back to attempted metrics if committed are not populated (e.g. due to not being supported on a given runner).
DistributionValue is the value of a Distribution metric.
type Extractor struct { // SumInt64 extracts data from Sum Int64 counters. SumInt64 func(labels Labels, v int64) // DistributionInt64 extracts data from Distribution Int64 counters. DistributionInt64 func(labels Labels, count, sum, min, max int64) // GaugeInt64 extracts data from Gauge Int64 counters. GaugeInt64 func(labels Labels, v int64, t time.Time) }
Extractor allows users to access metrics programatically after pipeline completion. Users assign functions to fields that interest them, and that function is called for each metric of the associated kind.
ExtractFrom the given metrics Store all the metrics for populated function fields. Returns an error if no fields were set.
type Gauge struct {
// contains filtered or unexported fields
}
Gauge is a time, value pair metric.
NewGauge returns the Gauge with the given namespace and name.
Set sets the gauge to the given value, and associates it with the current time on the clock.
type GaugeResult struct { Attempted, Committed GaugeValue Key StepKey }
GaugeResult is an attempted and a commited value of a gauge metric plus key.
func MergeGauges( attempted map[StepKey]GaugeValue, committed map[StepKey]GaugeValue) []GaugeResult
MergeGauges combines gauge metrics that share a common key.
func (r GaugeResult) Result() GaugeValue
Result returns committed metrics. Falls back to attempted metrics if committed are not populated (e.g. due to not being supported on a given runner).
GaugeValue is the value of a Gauge metric.
type Labels struct {
// contains filtered or unexported fields
}
Labels provide the context for the given metric.
PCollectionLabels builds a Labels for pcollection metrics. Intended for framework use.
PTransformLabels builds a Labels for transform metrics. Intended for framework use.
UserLabels builds a Labels for user metrics. Intended for framework use.
Name returns the name for this metric.
Namespace returns the namespace context for this metric.
Transform returns the transform context for this metric, if available.
type QueryResults struct {
// contains filtered or unexported fields
}
QueryResults is the result of a query. Allows accessing all of the metrics that matched the filter.
func (qr QueryResults) Counters() []CounterResult
Counters returns a slice of counter metrics.
func (qr QueryResults) Distributions() []DistributionResult
Distributions returns a slice of distribution metrics.
func (qr QueryResults) Gauges() []GaugeResult
Gauges returns a slice of gauge metrics.
type Results struct {
// contains filtered or unexported fields
}
Results represents all metrics gathered during the job's execution. It allows for querying metrics using a provided filter.
func NewResults( counters []CounterResult, distributions []DistributionResult, gauges []GaugeResult) *Results
NewResults creates a new Results.
func (mr Results) AllMetrics() QueryResults
AllMetrics returns all metrics from a Results instance.
StepKey uniquely identifies a metric within a pipeline graph.
type Store struct {
// contains filtered or unexported fields
}
Store retains per transform countersets, intended for per bundle use.
GetStore extracts the metrics Store for the given context for a bundle.
Returns nil if the context doesn't contain a metric Store.
Package metrics imports 9 packages (graph) and is imported by 8 packages. Updated 2020-12-18. Refresh now. Tools for package owners.