|
| 1 | +package prebuilds |
| 2 | + |
| 3 | +import ( |
| 4 | + "context" |
| 5 | + "time" |
| 6 | + |
| 7 | + "cdr.dev/slog" |
| 8 | + |
| 9 | + "github.com/prometheus/client_golang/prometheus" |
| 10 | + |
| 11 | + "github.com/coder/coder/v2/coderd/database" |
| 12 | + "github.com/coder/coder/v2/coderd/database/dbauthz" |
| 13 | + "github.com/coder/coder/v2/coderd/prebuilds" |
| 14 | +) |
| 15 | + |
| 16 | +var ( |
| 17 | + labels = []string{"template_name", "preset_name", "organization_name"} |
| 18 | + createdPrebuildsDesc = prometheus.NewDesc( |
| 19 | + "coderd_prebuilt_workspaces_created_total", |
| 20 | + "Total number of prebuilt workspaces that have been created to meet the desired instance count of each "+ |
| 21 | + "template preset.", |
| 22 | + labels, |
| 23 | + nil, |
| 24 | + ) |
| 25 | + failedPrebuildsDesc = prometheus.NewDesc( |
| 26 | + "coderd_prebuilt_workspaces_failed_total", |
| 27 | + "Total number of prebuilt workspaces that failed to build.", |
| 28 | + labels, |
| 29 | + nil, |
| 30 | + ) |
| 31 | + claimedPrebuildsDesc = prometheus.NewDesc( |
| 32 | + "coderd_prebuilt_workspaces_claimed_total", |
| 33 | + "Total number of prebuilt workspaces which were claimed by users. Claiming refers to creating a workspace "+ |
| 34 | + "with a preset selected for which eligible prebuilt workspaces are available and one is reassigned to a user.", |
| 35 | + labels, |
| 36 | + nil, |
| 37 | + ) |
| 38 | + desiredPrebuildsDesc = prometheus.NewDesc( |
| 39 | + "coderd_prebuilt_workspaces_desired", |
| 40 | + "Target number of prebuilt workspaces that should be available for each template preset.", |
| 41 | + labels, |
| 42 | + nil, |
| 43 | + ) |
| 44 | + runningPrebuildsDesc = prometheus.NewDesc( |
| 45 | + "coderd_prebuilt_workspaces_running", |
| 46 | + "Current number of prebuilt workspaces that are in a running state. These workspaces have started "+ |
| 47 | + "successfully but may not yet be claimable by users (see coderd_prebuilt_workspaces_eligible).", |
| 48 | + labels, |
| 49 | + nil, |
| 50 | + ) |
| 51 | + eligiblePrebuildsDesc = prometheus.NewDesc( |
| 52 | + "coderd_prebuilt_workspaces_eligible", |
| 53 | + "Current number of prebuilt workspaces that are eligible to be claimed by users. These are workspaces that "+ |
| 54 | + "have completed their build process with their agent reporting 'ready' status.", |
| 55 | + labels, |
| 56 | + nil, |
| 57 | + ) |
| 58 | +) |
| 59 | + |
| 60 | +type MetricsCollector struct { |
| 61 | + database database.Store |
| 62 | + logger slog.Logger |
| 63 | + snapshotter prebuilds.StateSnapshotter |
| 64 | +} |
| 65 | + |
| 66 | +var _ prometheus.Collector = new(MetricsCollector) |
| 67 | + |
| 68 | +func NewMetricsCollector(db database.Store, logger slog.Logger, snapshotter prebuilds.StateSnapshotter) *MetricsCollector { |
| 69 | + return &MetricsCollector{ |
| 70 | + database: db, |
| 71 | + logger: logger.Named("prebuilds_metrics_collector"), |
| 72 | + snapshotter: snapshotter, |
| 73 | + } |
| 74 | +} |
| 75 | + |
| 76 | +func (*MetricsCollector) Describe(descCh chan<- *prometheus.Desc) { |
| 77 | + descCh <- createdPrebuildsDesc |
| 78 | + descCh <- failedPrebuildsDesc |
| 79 | + descCh <- claimedPrebuildsDesc |
| 80 | + descCh <- desiredPrebuildsDesc |
| 81 | + descCh <- runningPrebuildsDesc |
| 82 | + descCh <- eligiblePrebuildsDesc |
| 83 | +} |
| 84 | + |
| 85 | +func (mc *MetricsCollector) Collect(metricsCh chan<- prometheus.Metric) { |
| 86 | + // nolint:gocritic // We need to set an authz context to read metrics from the db. |
| 87 | + ctx, cancel := context.WithTimeout(dbauthz.AsPrebuildsOrchestrator(context.Background()), 10*time.Second) |
| 88 | + defer cancel() |
| 89 | + prebuildMetrics, err := mc.database.GetPrebuildMetrics(ctx) |
| 90 | + if err != nil { |
| 91 | + mc.logger.Error(ctx, "failed to get prebuild metrics", slog.Error(err)) |
| 92 | + return |
| 93 | + } |
| 94 | + |
| 95 | + for _, metric := range prebuildMetrics { |
| 96 | + metricsCh <- prometheus.MustNewConstMetric(createdPrebuildsDesc, prometheus.CounterValue, float64(metric.CreatedCount), metric.TemplateName, metric.PresetName, metric.OrganizationName) |
| 97 | + metricsCh <- prometheus.MustNewConstMetric(failedPrebuildsDesc, prometheus.CounterValue, float64(metric.FailedCount), metric.TemplateName, metric.PresetName, metric.OrganizationName) |
| 98 | + metricsCh <- prometheus.MustNewConstMetric(claimedPrebuildsDesc, prometheus.CounterValue, float64(metric.ClaimedCount), metric.TemplateName, metric.PresetName, metric.OrganizationName) |
| 99 | + } |
| 100 | + |
| 101 | + snapshot, err := mc.snapshotter.SnapshotState(ctx, mc.database) |
| 102 | + if err != nil { |
| 103 | + mc.logger.Error(ctx, "failed to get latest prebuild state", slog.Error(err)) |
| 104 | + return |
| 105 | + } |
| 106 | + |
| 107 | + for _, preset := range snapshot.Presets { |
| 108 | + if !preset.UsingActiveVersion { |
| 109 | + continue |
| 110 | + } |
| 111 | + |
| 112 | + presetSnapshot, err := snapshot.FilterByPreset(preset.ID) |
| 113 | + if err != nil { |
| 114 | + mc.logger.Error(ctx, "failed to filter by preset", slog.Error(err)) |
| 115 | + continue |
| 116 | + } |
| 117 | + state := presetSnapshot.CalculateState() |
| 118 | + |
| 119 | + metricsCh <- prometheus.MustNewConstMetric(desiredPrebuildsDesc, prometheus.GaugeValue, float64(state.Desired), preset.TemplateName, preset.Name, preset.OrganizationName) |
| 120 | + metricsCh <- prometheus.MustNewConstMetric(runningPrebuildsDesc, prometheus.GaugeValue, float64(state.Actual), preset.TemplateName, preset.Name, preset.OrganizationName) |
| 121 | + metricsCh <- prometheus.MustNewConstMetric(eligiblePrebuildsDesc, prometheus.GaugeValue, float64(state.Eligible), preset.TemplateName, preset.Name, preset.OrganizationName) |
| 122 | + } |
| 123 | +} |
0 commit comments