| 1 | { |
| 2 | "uid": "shithubd-actions", |
| 3 | "title": "shithubd - Actions", |
| 4 | "tags": ["shithubd", "actions"], |
| 5 | "timezone": "browser", |
| 6 | "schemaVersion": 39, |
| 7 | "version": 1, |
| 8 | "refresh": "30s", |
| 9 | "time": {"from": "now-6h", "to": "now"}, |
| 10 | "templating": { |
| 11 | "list": [ |
| 12 | { |
| 13 | "name": "instance", |
| 14 | "type": "query", |
| 15 | "datasource": "Prometheus", |
| 16 | "query": "label_values(up{job=\"shithubd-web\"}, instance)", |
| 17 | "includeAll": true, |
| 18 | "multi": true |
| 19 | } |
| 20 | ] |
| 21 | }, |
| 22 | "panels": [ |
| 23 | { |
| 24 | "id": 1, |
| 25 | "type": "stat", |
| 26 | "title": "Queued jobs", |
| 27 | "gridPos": {"x": 0, "y": 0, "w": 4, "h": 4}, |
| 28 | "targets": [{"expr": "sum(shithub_actions_queue_depth{resource=\"jobs\",instance=~\"$instance\"})", "refId": "A"}], |
| 29 | "fieldConfig": { |
| 30 | "defaults": { |
| 31 | "thresholds": { |
| 32 | "mode": "absolute", |
| 33 | "steps": [ |
| 34 | {"color": "green", "value": null}, |
| 35 | {"color": "yellow", "value": 50}, |
| 36 | {"color": "red", "value": 100} |
| 37 | ] |
| 38 | } |
| 39 | } |
| 40 | } |
| 41 | }, |
| 42 | { |
| 43 | "id": 2, |
| 44 | "type": "stat", |
| 45 | "title": "Running jobs", |
| 46 | "gridPos": {"x": 4, "y": 0, "w": 4, "h": 4}, |
| 47 | "targets": [{"expr": "sum(shithub_actions_active{resource=\"jobs\",instance=~\"$instance\"})", "refId": "A"}] |
| 48 | }, |
| 49 | { |
| 50 | "id": 3, |
| 51 | "type": "stat", |
| 52 | "title": "Stale runners", |
| 53 | "gridPos": {"x": 8, "y": 0, "w": 4, "h": 4}, |
| 54 | "targets": [{"expr": "count(shithub_actions_runner_heartbeat_age_seconds{status!=\"offline\",instance=~\"$instance\"} > 60)", "refId": "A"}], |
| 55 | "fieldConfig": { |
| 56 | "defaults": { |
| 57 | "thresholds": { |
| 58 | "mode": "absolute", |
| 59 | "steps": [ |
| 60 | {"color": "green", "value": null}, |
| 61 | {"color": "red", "value": 1} |
| 62 | ] |
| 63 | } |
| 64 | } |
| 65 | } |
| 66 | }, |
| 67 | { |
| 68 | "id": 4, |
| 69 | "type": "stat", |
| 70 | "title": "Log MB/day", |
| 71 | "gridPos": {"x": 12, "y": 0, "w": 4, "h": 4}, |
| 72 | "targets": [{"expr": "sum(increase(shithub_actions_log_chunk_bytes_total{instance=~\"$instance\"}[24h])) / 1024 / 1024", "refId": "A"}] |
| 73 | }, |
| 74 | { |
| 75 | "id": 5, |
| 76 | "type": "stat", |
| 77 | "title": "Run p99", |
| 78 | "gridPos": {"x": 16, "y": 0, "w": 4, "h": 4}, |
| 79 | "targets": [{"expr": "histogram_quantile(0.99, sum(rate(shithub_actions_run_duration_seconds_bucket{instance=~\"$instance\"}[30m])) by (le))", "refId": "A"}], |
| 80 | "fieldConfig": {"defaults": {"unit": "s"}} |
| 81 | }, |
| 82 | { |
| 83 | "id": 6, |
| 84 | "type": "stat", |
| 85 | "title": "Storage MB", |
| 86 | "gridPos": {"x": 20, "y": 0, "w": 4, "h": 4}, |
| 87 | "targets": [{"expr": "sum(shithub_actions_storage_bytes{instance=~\"$instance\"}) / 1024 / 1024", "refId": "A"}] |
| 88 | }, |
| 89 | { |
| 90 | "id": 7, |
| 91 | "type": "timeseries", |
| 92 | "title": "Queue depth", |
| 93 | "gridPos": {"x": 0, "y": 4, "w": 12, "h": 8}, |
| 94 | "targets": [ |
| 95 | { |
| 96 | "expr": "sum(shithub_actions_queue_depth{instance=~\"$instance\"}) by (resource)", |
| 97 | "legendFormat": "{{resource}} queued", |
| 98 | "refId": "A" |
| 99 | } |
| 100 | ] |
| 101 | }, |
| 102 | { |
| 103 | "id": 8, |
| 104 | "type": "timeseries", |
| 105 | "title": "Active runs and jobs", |
| 106 | "gridPos": {"x": 12, "y": 4, "w": 12, "h": 8}, |
| 107 | "targets": [ |
| 108 | { |
| 109 | "expr": "sum(shithub_actions_active{instance=~\"$instance\"}) by (resource)", |
| 110 | "legendFormat": "{{resource}} active", |
| 111 | "refId": "A" |
| 112 | } |
| 113 | ] |
| 114 | }, |
| 115 | { |
| 116 | "id": 9, |
| 117 | "type": "timeseries", |
| 118 | "title": "Run duration p95 and p99", |
| 119 | "gridPos": {"x": 0, "y": 12, "w": 12, "h": 8}, |
| 120 | "targets": [ |
| 121 | { |
| 122 | "expr": "histogram_quantile(0.95, sum(rate(shithub_actions_run_duration_seconds_bucket{instance=~\"$instance\"}[15m])) by (le, event))", |
| 123 | "legendFormat": "p95 {{event}}", |
| 124 | "refId": "A" |
| 125 | }, |
| 126 | { |
| 127 | "expr": "histogram_quantile(0.99, sum(rate(shithub_actions_run_duration_seconds_bucket{instance=~\"$instance\"}[15m])) by (le, event))", |
| 128 | "legendFormat": "p99 {{event}}", |
| 129 | "refId": "B" |
| 130 | } |
| 131 | ], |
| 132 | "fieldConfig": {"defaults": {"unit": "s"}} |
| 133 | }, |
| 134 | { |
| 135 | "id": 10, |
| 136 | "type": "timeseries", |
| 137 | "title": "Runner heartbeat age", |
| 138 | "gridPos": {"x": 12, "y": 12, "w": 12, "h": 8}, |
| 139 | "targets": [ |
| 140 | { |
| 141 | "expr": "shithub_actions_runner_heartbeat_age_seconds{instance=~\"$instance\"}", |
| 142 | "legendFormat": "{{runner}} {{status}}", |
| 143 | "refId": "A" |
| 144 | } |
| 145 | ], |
| 146 | "fieldConfig": {"defaults": {"unit": "s"}} |
| 147 | }, |
| 148 | { |
| 149 | "id": 11, |
| 150 | "type": "timeseries", |
| 151 | "title": "Runs per minute", |
| 152 | "gridPos": {"x": 0, "y": 20, "w": 12, "h": 8}, |
| 153 | "targets": [ |
| 154 | { |
| 155 | "expr": "sum(rate(shithub_actions_runs_enqueued_total{result=\"fresh\",instance=~\"$instance\"}[5m])) * 60", |
| 156 | "legendFormat": "enqueued", |
| 157 | "refId": "A" |
| 158 | }, |
| 159 | { |
| 160 | "expr": "sum(rate(shithub_actions_runs_completed_total{instance=~\"$instance\"}[5m])) * 60", |
| 161 | "legendFormat": "completed", |
| 162 | "refId": "B" |
| 163 | } |
| 164 | ] |
| 165 | }, |
| 166 | { |
| 167 | "id": 12, |
| 168 | "type": "timeseries", |
| 169 | "title": "Run conclusions", |
| 170 | "gridPos": {"x": 12, "y": 20, "w": 12, "h": 8}, |
| 171 | "targets": [ |
| 172 | { |
| 173 | "expr": "sum(rate(shithub_actions_runs_completed_total{instance=~\"$instance\"}[15m])) by (conclusion)", |
| 174 | "legendFormat": "{{conclusion}}", |
| 175 | "refId": "A" |
| 176 | } |
| 177 | ] |
| 178 | }, |
| 179 | { |
| 180 | "id": 13, |
| 181 | "type": "timeseries", |
| 182 | "title": "Step outcomes", |
| 183 | "gridPos": {"x": 0, "y": 28, "w": 12, "h": 8}, |
| 184 | "targets": [ |
| 185 | { |
| 186 | "expr": "sum(rate(shithub_actions_steps_completed_total{instance=~\"$instance\"}[15m])) by (step_type, conclusion)", |
| 187 | "legendFormat": "{{step_type}} {{conclusion}}", |
| 188 | "refId": "A" |
| 189 | } |
| 190 | ] |
| 191 | }, |
| 192 | { |
| 193 | "id": 14, |
| 194 | "type": "timeseries", |
| 195 | "title": "Log throughput", |
| 196 | "gridPos": {"x": 12, "y": 28, "w": 12, "h": 8}, |
| 197 | "targets": [ |
| 198 | { |
| 199 | "expr": "sum(rate(shithub_actions_log_chunk_bytes_total{instance=~\"$instance\"}[5m])) by (location)", |
| 200 | "legendFormat": "{{location}} bytes/sec", |
| 201 | "refId": "A" |
| 202 | }, |
| 203 | { |
| 204 | "expr": "sum(rate(shithub_actions_log_chunks_total{instance=~\"$instance\"}[5m])) by (location)", |
| 205 | "legendFormat": "{{location}} chunks/sec", |
| 206 | "refId": "B" |
| 207 | } |
| 208 | ] |
| 209 | }, |
| 210 | { |
| 211 | "id": 15, |
| 212 | "type": "timeseries", |
| 213 | "title": "Actions storage", |
| 214 | "gridPos": {"x": 0, "y": 36, "w": 12, "h": 8}, |
| 215 | "targets": [ |
| 216 | { |
| 217 | "expr": "sum(shithub_actions_storage_bytes{instance=~\"$instance\"}) by (kind)", |
| 218 | "legendFormat": "{{kind}} bytes", |
| 219 | "refId": "A" |
| 220 | }, |
| 221 | { |
| 222 | "expr": "sum(shithub_actions_storage_objects{instance=~\"$instance\"}) by (kind)", |
| 223 | "legendFormat": "{{kind}} objects", |
| 224 | "refId": "B" |
| 225 | } |
| 226 | ] |
| 227 | }, |
| 228 | { |
| 229 | "id": 16, |
| 230 | "type": "timeseries", |
| 231 | "title": "Cancellations and retention", |
| 232 | "gridPos": {"x": 12, "y": 36, "w": 12, "h": 8}, |
| 233 | "targets": [ |
| 234 | { |
| 235 | "expr": "sum(rate(shithub_actions_jobs_cancelled_total{instance=~\"$instance\"}[15m])) by (reason)", |
| 236 | "legendFormat": "cancel {{reason}}", |
| 237 | "refId": "A" |
| 238 | }, |
| 239 | { |
| 240 | "expr": "sum(rate(shithub_actions_runs_pruned_total{instance=~\"$instance\"}[1h])) by (kind)", |
| 241 | "legendFormat": "pruned {{kind}}", |
| 242 | "refId": "B" |
| 243 | } |
| 244 | ] |
| 245 | } |
| 246 | ] |
| 247 | } |