mirror of
https://github.com/coder/coder.git
synced 2026-06-02 20:48:20 +00:00
Jon Ayers
6035e45cb8
Adds coderd_template_workspace_build_duration_seconds histogram that tracks the full duration from workspace build creation to agent ready. This captures the complete user-perceived build time including provisioning and agent startup. The metric is emitted when the agent reports ready/error/timeout via the lifecycle API, ensuring each build is counted exactly once per replica.
98 lines
3.0 KiB
Go
98 lines
3.0 KiB
Go
package agentapi
|
|
|
|
import (
|
|
"context"
|
|
"strconv"
|
|
"time"
|
|
|
|
"github.com/google/uuid"
|
|
"github.com/prometheus/client_golang/prometheus"
|
|
|
|
"cdr.dev/slog/v3"
|
|
)
|
|
|
|
// BuildDurationMetricName is the short name for the end-to-end
|
|
// workspace build duration histogram. The full metric name is
|
|
// prefixed with the namespace "coderd_".
|
|
const BuildDurationMetricName = "template_workspace_build_duration_seconds"
|
|
|
|
// LifecycleMetrics contains Prometheus metrics for the lifecycle API.
|
|
type LifecycleMetrics struct {
|
|
BuildDuration *prometheus.HistogramVec
|
|
}
|
|
|
|
// NewLifecycleMetrics creates and registers all lifecycle-related
|
|
// Prometheus metrics.
|
|
//
|
|
// The build duration histogram tracks the end-to-end duration from
|
|
// workspace build creation to agent ready, by template. It is
|
|
// recorded by the coderd replica handling the agent's connection
|
|
// when the last agent reports ready. In multi-replica deployments,
|
|
// each replica only has observations for agents it handles.
|
|
//
|
|
// The "is_prebuild" label distinguishes prebuild creation (background,
|
|
// no user waiting) from user-initiated builds (regular workspace
|
|
// creation or prebuild claims).
|
|
func NewLifecycleMetrics(reg prometheus.Registerer) *LifecycleMetrics {
|
|
m := &LifecycleMetrics{
|
|
BuildDuration: prometheus.NewHistogramVec(prometheus.HistogramOpts{
|
|
Namespace: "coderd",
|
|
Name: BuildDurationMetricName,
|
|
Help: "Duration from workspace build creation to agent ready, by template.",
|
|
Buckets: []float64{
|
|
1, // 1s
|
|
10,
|
|
30,
|
|
60, // 1min
|
|
60 * 5,
|
|
60 * 10,
|
|
60 * 30, // 30min
|
|
60 * 60, // 1hr
|
|
},
|
|
NativeHistogramBucketFactor: 1.1,
|
|
NativeHistogramMaxBucketNumber: 100,
|
|
NativeHistogramMinResetDuration: time.Hour,
|
|
}, []string{"template_name", "organization_name", "transition", "status", "is_prebuild"}),
|
|
}
|
|
reg.MustRegister(m.BuildDuration)
|
|
return m
|
|
}
|
|
|
|
// emitBuildDurationMetric records the end-to-end workspace build
|
|
// duration from build creation to when all agents are ready.
|
|
func (a *LifecycleAPI) emitBuildDurationMetric(ctx context.Context, resourceID uuid.UUID) {
|
|
if a.Metrics == nil {
|
|
return
|
|
}
|
|
|
|
buildInfo, err := a.Database.GetWorkspaceBuildMetricsByResourceID(ctx, resourceID)
|
|
if err != nil {
|
|
a.Log.Warn(ctx, "failed to get build info for metrics", slog.Error(err))
|
|
return
|
|
}
|
|
|
|
// Wait until all agents have reached a terminal startup state.
|
|
if !buildInfo.AllAgentsReady {
|
|
return
|
|
}
|
|
|
|
// LastAgentReadyAt is the MAX(ready_at) across all agents. Since
|
|
// we only get here when AllAgentsReady is true, this should always
|
|
// be valid.
|
|
if buildInfo.LastAgentReadyAt.IsZero() {
|
|
a.Log.Warn(ctx, "last_agent_ready_at is unexpectedly zero",
|
|
slog.F("last_agent_ready_at", buildInfo.LastAgentReadyAt))
|
|
return
|
|
}
|
|
|
|
duration := buildInfo.LastAgentReadyAt.Sub(buildInfo.CreatedAt).Seconds()
|
|
|
|
a.Metrics.BuildDuration.WithLabelValues(
|
|
buildInfo.TemplateName,
|
|
buildInfo.OrganizationName,
|
|
string(buildInfo.Transition),
|
|
buildInfo.WorstStatus,
|
|
strconv.FormatBool(buildInfo.IsPrebuild),
|
|
).Observe(duration)
|
|
}
|