Skip to content

Commit

Permalink
Sherif akoush/demo fixes (#312)
Browse files Browse the repository at this point in the history 
* add k6 changes

* update dashboard

* update doc dashboard

* fix size
  • Loading branch information
@sakoush
sakoush authored Jun 24, 2022
1 parent 18cf74f commit 6df366a
Show file tree
Hide file tree
Showing 5 changed files with 198 additions and 49 deletions.
Binary file modified docs/source/contents/metrics/dashboard.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
197 changes: 164 additions & 33 deletions prometheus/dashboards/Seldon Core Model Mesh Monitoring.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -704,7 +704,7 @@
},
"gridPos": {
"h": 8,
"w": 8,
"w": 5,
"x": 0,
"y": 12
},
Expand All @@ -728,7 +728,7 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_bytes_gauge)",
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_bytes_gauge{server=\"triton\"})",
"hide": false,
"interval": "",
"legendFormat": "Capacity",
Expand All @@ -740,7 +740,7 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge)",
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge{server=\"triton\"})",
"hide": false,
"interval": "",
"legendFormat": "Used",
Expand All @@ -752,7 +752,7 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_overcommit_bytes_gauge)",
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_overcommit_bytes_gauge{server=\"triton\"})",
"hide": false,
"interval": "",
"legendFormat": "Capacity with Over-commit",
Expand All @@ -764,14 +764,14 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge) + sum(seldon_mesh_seldon_evicted_model_memory_bytes_gauge)",
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge{server=\"triton\"}) + sum(seldon_mesh_seldon_evicted_model_memory_bytes_gauge{server=\"triton\"})",
"hide": false,
"interval": "",
"legendFormat": "Used with Over-commit",
"refId": "D"
}
],
"title": "Memory Slots",
"title": "Memory Slots (triton)",
"transformations": [],
"type": "timeseries"
},
Expand All @@ -791,24 +791,21 @@
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"fillOpacity": 10,
"gradientMode": "none",
"hideFrom": {
"legend": false,
"tooltip": false,
"viz": false
},
"lineInterpolation": "linear",
"lineStyle": {
"fill": "solid"
},
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"showPoints": "never",
"spanNulls": true,
"stacking": {
"group": "A",
"mode": "none"
Expand All @@ -818,6 +815,7 @@
}
},
"mappings": [],
"min": 0,
"thresholds": {
"mode": "absolute",
"steps": [
Expand All @@ -831,25 +829,25 @@
}
]
},
"unit": "s"
"unit": "decbytes"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 8,
"x": 8,
"w": 5,
"x": 5,
"y": 12
},
"id": 15,
"id": 21,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "single",
"mode": "multi",
"sort": "none"
}
},
Expand All @@ -861,14 +859,50 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "avg((rate(seldon_mesh_seldon_aggregate_infer_seconds_total{container=\"agent\"}[1m]) / rate(seldon_mesh_seldon_aggregate_infer_total{container=\"agent\"}[1m])) > 0 ) by (server, method_type)",
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_bytes_gauge{server=\"mlserver\"})",
"hide": false,
"interval": "",
"legendFormat": "{{server}}_{{method_type}}_avg",
"legendFormat": "Capacity",
"refId": "B"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge{server=\"mlserver\"})",
"hide": false,
"interval": "",
"legendFormat": "Used",
"refId": "C"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_server_replica_memory_capacity_overcommit_bytes_gauge{server=\"mlserver\"})",
"hide": false,
"interval": "",
"legendFormat": "Capacity with Over-commit",
"refId": "A"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "sum(seldon_mesh_seldon_loaded_model_memory_bytes_gauge{server=\"mlserver\"}) + sum(seldon_mesh_seldon_evicted_model_memory_bytes_gauge{server=\"mlserver\"})",
"hide": false,
"interval": "",
"legendFormat": "Used with Over-commit",
"refId": "D"
}
],
"title": "Infer Latency [1m]",
"title": "Memory Slots (mlserver)",
"transformations": [],
"type": "timeseries"
},
Expand Down Expand Up @@ -929,10 +963,10 @@
"overrides": []
},
"gridPos": {
"h": 9,
"w": 8,
"x": 0,
"y": 20
"h": 8,
"w": 6,
"x": 10,
"y": 12
},
"id": 19,
"options": {
Expand All @@ -942,7 +976,7 @@
"placement": "bottom"
},
"tooltip": {
"mode": "single",
"mode": "multi",
"sort": "none"
}
},
Expand All @@ -953,9 +987,9 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "container_memory_working_set_bytes{container=\"mlserver\"}",
"expr": "sum(container_memory_working_set_bytes{container=\"mlserver\"}) ",
"interval": "10s",
"legendFormat": "{{pod}}",
"legendFormat": "mlserver",
"refId": "A"
},
{
Expand All @@ -964,16 +998,113 @@
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "container_memory_working_set_bytes{container=\"triton\"}",
"expr": "sum(container_memory_working_set_bytes{container=\"triton\"})",
"hide": false,
"interval": "10s",
"legendFormat": "{{pod}}",
"legendFormat": "triton",
"refId": "B"
}
],
"title": "Memory Used",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"custom": {
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"hideFrom": {
"legend": false,
"tooltip": false,
"viz": false
},
"lineInterpolation": "linear",
"lineStyle": {
"fill": "solid"
},
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": null
},
{
"color": "red",
"value": 80
}
]
},
"unit": "s"
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 8,
"x": 0,
"y": 20
},
"id": 15,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "single",
"sort": "none"
}
},
"pluginVersion": "8.4.6",
"targets": [
{
"datasource": {
"type": "prometheus",
"uid": "P1809F7CD0C75ACF3"
},
"exemplar": true,
"expr": "avg((rate(seldon_mesh_seldon_aggregate_infer_seconds_total{container=\"agent\"}[1m]) / rate(seldon_mesh_seldon_aggregate_infer_total{container=\"agent\"}[1m])) > 0 ) by (server, method_type)",
"hide": false,
"interval": "",
"legendFormat": "{{server}}_{{method_type}}_avg",
"refId": "A"
}
],
"title": "Infer Latency [1m]",
"transformations": [],
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
Expand Down Expand Up @@ -1030,7 +1161,7 @@
"overrides": []
},
"gridPos": {
"h": 9,
"h": 8,
"w": 8,
"x": 8,
"y": 20
Expand Down Expand Up @@ -1076,21 +1207,21 @@
"type": "timeseries"
}
],
"refresh": "5s",
"refresh": "30s",
"schemaVersion": 35,
"style": "dark",
"tags": [],
"templating": {
"list": []
},
"time": {
"from": "now-15m",
"from": "now-10m",
"to": "now"
},
"timepicker": {},
"timezone": "",
"title": "Seldon Core Model Mesh Monitoring",
"uid": "MHloCP_7z",
"version": 22,
"version": 7,
"weekStart": ""
}
8 changes: 8 additions & 0 deletions tests/k6/components/settings.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,13 @@ function inferType() {
return "REST"
}

function doWarmup() {
if (__ENV.WARMUP) {
return (__ENV.WARMUP === "true")
}
return true
}

export function getConfig() {
return {
"schedulerEndpoint": schedulerEndpoint(),
Expand All @@ -192,5 +199,6 @@ export function getConfig() {
"modelEndIdx" : modelEndIdx(),
"modelName" : modelName(),
"inferType" : inferType(),
"doWarmup": doWarmup(),
}
}
Loading

0 comments on commit 6df366a

Please sign in to comment.