85742ab165
Deploy Documentation / deploy (push) Has been cancelled
CPU Test / Test (Utilities, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (LLM proxy, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (Others, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (Store, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (Utilities, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (Weave, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (AgentOps, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (LLM proxy, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (Others, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (Weave, latest, Python 3.13) (push) Has been cancelled
Dashboard / Chromatic (push) Has been cancelled
CPU Test / Lint - fast (push) Has been cancelled
CPU Test / Lint - next (push) Has been cancelled
CPU Test / Lint - slow (push) Has been cancelled
CPU Test / Lint - JavaScript (push) Has been cancelled
CPU Test / Build documentation (push) Has been cancelled
CPU Test / Test (AgentOps, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (LLM proxy, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (Others, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (Store, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (Weave, legacy, Python 3.10) (push) Has been cancelled
CPU Test / Test (AgentOps, stable, Python 3.11) (push) Has been cancelled
CPU Test / Test (Store, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (Utilities, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (Weave, stable, Python 3.12) (push) Has been cancelled
CPU Test / Test (AgentOps, latest, Python 3.13) (push) Has been cancelled
CPU Test / Test (LLM proxy, latest, Python 3.13) (push) Has been cancelled
CPU Test / Test (Others, latest, Python 3.13) (push) Has been cancelled
CPU Test / Test (Store, latest, Python 3.13) (push) Has been cancelled
CPU Test / Test (Utilities, latest, Python 3.13) (push) Has been cancelled
CPU Test / Test (JavaScript) (push) Has been cancelled
531 lines
11 KiB
JSON
531 lines
11 KiB
JSON
{
|
|
"annotations": {
|
|
"list": []
|
|
},
|
|
"editable": true,
|
|
"gnetId": null,
|
|
"graphTooltip": 0,
|
|
"id": null,
|
|
"links": [],
|
|
"panels": [
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 0
|
|
},
|
|
"id": 100,
|
|
"panels": [],
|
|
"title": "LightningStore HTTP API",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 0,
|
|
"y": 1
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (method, path) (rate(agl_http_latency_count[15s]))",
|
|
"legendFormat": "{{method}} {{path}}"
|
|
}
|
|
],
|
|
"title": "HTTP Requests / Sec (by Path)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 8,
|
|
"y": 1
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.50, sum by (le) (rate(agl_http_latency_bucket[15s])))",
|
|
"legendFormat": "P50"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le) (rate(agl_http_latency_bucket[15s])))",
|
|
"legendFormat": "P95"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.99, sum by (le) (rate(agl_http_latency_bucket[15s])))",
|
|
"legendFormat": "P99"
|
|
}
|
|
],
|
|
"title": "HTTP Latency (P50, P95, P99)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 16,
|
|
"y": 1
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (method, path, status) (rate(agl_http_latency_count{status!~\"2..\"}[15s]))",
|
|
"legendFormat": "{{method}} {{path}} {{status}}"
|
|
}
|
|
],
|
|
"title": "HTTP Errors / Sec (non-2xx)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 9
|
|
},
|
|
"id": 200,
|
|
"panels": [],
|
|
"title": "LightningStore RPC Metrics",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 0,
|
|
"y": 10
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (method) (rate(agl_store_latency_count[15s]))",
|
|
"legendFormat": "{{method}}"
|
|
}
|
|
],
|
|
"title": "Store Calls / Sec (by Method)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 8,
|
|
"y": 10
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.50, sum by (le) (rate(agl_store_latency_bucket[15s])))",
|
|
"legendFormat": "P50"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le) (rate(agl_store_latency_bucket[15s])))",
|
|
"legendFormat": "P95"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.99, sum by (le) (rate(agl_store_latency_bucket[15s])))",
|
|
"legendFormat": "P99"
|
|
}
|
|
],
|
|
"title": "Store Latency (P50, P95, P99)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 8,
|
|
"x": 16,
|
|
"y": 10
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (method, status) (rate(agl_store_latency_count{status!=\"OK\"}[15s]))",
|
|
"legendFormat": "{{method}} {{status}}"
|
|
}
|
|
],
|
|
"title": "Store Errors / Sec (by Method)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 18
|
|
},
|
|
"id": 300,
|
|
"panels": [],
|
|
"title": "Collection Metrics",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 19
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (store_pubmeth, operation) (rate(agl_collections_latency_count[15s]))",
|
|
"legendFormat": "{{store_pubmeth}} {{operation}}"
|
|
}
|
|
],
|
|
"title": "Collection Ops / Sec (by Store Method)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 19
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (store_pubmeth, collection, status) (rate(agl_collections_latency_count{status!=\"OK\"}[15s]))",
|
|
"legendFormat": "{{store_pubmeth}} {{collection}} {{status}}"
|
|
}
|
|
],
|
|
"title": "Collection Errors / Sec",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 27
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.50, sum by (le) (rate(agl_collections_latency_bucket[15s])))",
|
|
"legendFormat": "P50"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le) (rate(agl_collections_latency_bucket[15s])))",
|
|
"legendFormat": "P95"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.99, sum by (le) (rate(agl_collections_latency_bucket[15s])))",
|
|
"legendFormat": "P99"
|
|
}
|
|
],
|
|
"title": "Collection Latency (P50, P95, P99)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 27
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le, store_pubmeth) (rate(agl_collections_latency_bucket[15s])))",
|
|
"legendFormat": "{{store_pubmeth}}"
|
|
}
|
|
],
|
|
"title": "Collection Latency P95 (by Store Method)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 35
|
|
},
|
|
"id": 400,
|
|
"panels": [],
|
|
"title": "Rollout Outcomes",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 36
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (status) (rate(agl_rollouts_duration_count[15s]))",
|
|
"legendFormat": "{{status}}"
|
|
}
|
|
],
|
|
"title": "Rollout Completion Rate (by Status)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "ops/s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 36
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by (mode) (rate(agl_rollouts_duration_count[15s]))",
|
|
"legendFormat": "{{mode}}"
|
|
}
|
|
],
|
|
"title": "Rollout Completion Rate (by Mode)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 44
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.50, sum by (le) (rate(agl_rollouts_duration_bucket[15s])))",
|
|
"legendFormat": "P50"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le) (rate(agl_rollouts_duration_bucket[15s])))",
|
|
"legendFormat": "P95"
|
|
},
|
|
{
|
|
"expr": "histogram_quantile(0.99, sum by (le) (rate(agl_rollouts_duration_bucket[15s])))",
|
|
"legendFormat": "P99"
|
|
}
|
|
],
|
|
"title": "Rollout Duration (P50, P95, P99)",
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s"
|
|
}
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 44
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "single"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "histogram_quantile(0.95, sum by (le, status) (rate(agl_rollouts_duration_bucket[15s])))",
|
|
"legendFormat": "{{status}}"
|
|
}
|
|
],
|
|
"title": "Rollout Duration P95 (by Status)",
|
|
"type": "timeseries"
|
|
}
|
|
],
|
|
"schemaVersion": 36,
|
|
"style": "dark",
|
|
"tags": [
|
|
"agentlightning"
|
|
],
|
|
"templating": {
|
|
"list": []
|
|
},
|
|
"time": {
|
|
"from": "now-10m",
|
|
"to": "now"
|
|
},
|
|
"title": "Agent Lightning Metrics",
|
|
"version": 2
|
|
}
|