Skip to content

make component loader event group formation deterministic for post-processing #517

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 15, 2025
Merged

make component loader event group formation deterministic for post-processing #517

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3c27b7e
make arm event group formation deterministic for post-processing
harp-intel Oct 15, 2025
fa2ccb7
Update cmd/metrics/loader_component.go
harp-intel Oct 15, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
50 changes: 20 additions & 30 deletions cmd/metrics/loader_component.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ package metrics
import (
"encoding/json"
"fmt"
"io/fs"
"log/slog"
"os"
"path/filepath"
Expand All @@ -26,10 +27,6 @@ func (l *ComponentLoader) Load(loaderConfig LoaderConfig) ([]MetricDefinition, [
if err != nil {
return nil, nil, fmt.Errorf("failed to load event definitions: %w", err)
}
metricDefinitions, err = l.filterUncollectableMetrics(metricDefinitions, eventDefinitions, loaderConfig.Metadata)
if err != nil {
return nil, nil, fmt.Errorf("failed to filter uncollectable metrics: %w", err)
}
groupDefinitions, err := l.formEventGroups(metricDefinitions, eventDefinitions, loaderConfig.Metadata)
if err != nil {
return nil, nil, fmt.Errorf("failed to form event groups: %w", err)
Expand Down Expand Up @@ -180,6 +177,10 @@ func (l *ComponentLoader) loadEventDefinitions(metadata Metadata) (events []Comp
if err != nil {
return
}
// sort for deterministic processing order
slices.SortFunc(dirEntries, func(a, b fs.DirEntry) int {
return strings.Compare(a.Name(), b.Name())
})
for _, entry := range dirEntries {
if entry.IsDir() || entry.Name() == "metrics.json" {
continue
Expand All @@ -198,30 +199,6 @@ func (l *ComponentLoader) loadEventDefinitions(metadata Metadata) (events []Comp
return events, nil
}

func (l *ComponentLoader) filterUncollectableMetrics(metrics []MetricDefinition, events []ComponentEvent, metadata Metadata) (filteredMetrics []MetricDefinition, err error) {
uncollectableEvents, err := l.identifyUncollectableEvents(events, metadata)
if err != nil {
return
}
for _, metric := range metrics {
for variable := range metric.Variables {
if slices.Contains(uncollectableEvents, variable) {
slog.Info("Excluding metric due to uncollectable event", slog.String("metric", metric.Name), slog.String("event", variable))
goto nextMetric
}
}
filteredMetrics = append(filteredMetrics, metric)
nextMetric:
}
return filteredMetrics, nil
}

func (l *ComponentLoader) identifyUncollectableEvents(events []ComponentEvent, metadata Metadata) (uncollectableEvents []string, err error) {
// TODO:
// For now, assume all events are collectable
return uncollectableEvents, nil
}

func (l *ComponentLoader) formEventGroups(metrics []MetricDefinition, events []ComponentEvent, metadata Metadata) (groups []GroupDefinition, err error) {
numGPCounters := metadata.NumGeneralPurposeCounters // groups can have at most this many events (plus fixed counters)
eventNames := make(map[string]bool)
Expand All @@ -242,7 +219,14 @@ func (l *ComponentLoader) formEventGroups(metrics []MetricDefinition, events []C
var currentGroup GroupDefinition
var currentGPCount int // Track the current number of GP counters used

// Get variable names and sort them for deterministic order
var variables []string
for variable := range metric.Variables {
variables = append(variables, variable)
}
slices.Sort(variables)

for _, variable := range variables {
// confirm variable is a valid event
if _, exists := eventNames[variable]; !exists {
slog.Warn("Metric variable does not correspond to a known event, skipping variable", slog.String("metric", metric.Name), slog.String("variable", variable))
Expand Down Expand Up @@ -290,12 +274,18 @@ func mergeSmallGroups(groups []GroupDefinition, numGPCounters int) []GroupDefini
}

// Sort groups by size for efficient merging (smallest first)
// Important that this is a deterministic sort
slices.SortFunc(groups, func(a, b GroupDefinition) int {
if len(a) == 0 || len(b) == 0 {
panic("empty group encountered during sorting in mergeSmallGroups")
}
aGPCount := countGPEvents(a)
bGPCount := countGPEvents(b)
return aGPCount - bGPCount
if aGPCount != bGPCount {
return aGPCount - bGPCount
}
return 1 // arbitrary but consistent
})

var mergedGroups []GroupDefinition
processed := make([]bool, len(groups))

Expand Down