Skip to content

Commit

Permalink
[dashboards] sync grafana dashboards (aptos-labs#4303)
Browse files Browse the repository at this point in the history 
Co-authored-by: rustielin <[email protected]>
  • Loading branch information
@github-actions @rustielin
github-actions[bot] and rustielin authored Sep 20, 2022
1 parent 31bdfdf commit 2cb091e
Show file tree
Hide file tree
Showing 6 changed files with 1,810 additions and 2,476 deletions.
226 changes: 211 additions & 15 deletions dashboards/consensus.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -896,7 +896,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_proposal', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_proposal', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{op='process_proposal', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", chain_name=~\"$chain_name\", namespace=~\"$namespace\"}[$interval]) / rate(consensus_duration_count{op='process_proposal', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", chain_name=~\"$chain_name\", namespace=~\"$namespace\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -976,7 +976,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_vote', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_vote', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{op='process_vote', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", chain_name=~\"$chain_name\", namespace=~\"$namespace\"}[$interval]) / rate(consensus_duration_count{op='process_vote', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", chain_name=~\"$chain_name\", namespace=~\"$namespace\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1056,7 +1056,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_sync_info', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_sync_info', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_sync_info', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_sync_info', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1136,7 +1136,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_local_timeout', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_local_timeout', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_local_timeout', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_local_timeout', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1216,7 +1216,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='verify_message', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='verify_message', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='verify_message', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='verify_message', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1296,7 +1296,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_different_epoch_consensus_msg', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_different_epoch_consensus_msg', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_different_epoch_consensus_msg', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_different_epoch_consensus_msg', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1376,7 +1376,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_epoch_proof', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_epoch_proof', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_epoch_proof', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_epoch_proof', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1536,7 +1536,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='reconfig', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='reconfig', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='reconfig', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='reconfig', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1616,7 +1616,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(consensus_duration_sum{op='process_block_retrieval', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{op='process_block_retrieval', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"expr": "rate(consensus_duration_sum{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_block_retrieval', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval]) / rate(consensus_duration_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", op='process_block_retrieval', kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
Expand Down Expand Up @@ -1732,6 +1732,86 @@
}
}
},
{
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"editable": false,
"error": false,
"gridPos": { "h": 8, "w": 8, "x": 16, "y": 75 },
"id": 129,
"isNew": false,
"renderer": "flot",
"span": 0,
"title": "Buffer Manager Retry count",
"description": "A sampled binary number indicating whether the node is currently in a function. (Useful when the node is stuck somewhere)",
"transparent": false,
"type": "graph",
"aliasColors": {},
"bars": false,
"dashLength": 10,
"dashes": false,
"fill": 0,
"legend": {
"alignAsTable": false,
"avg": false,
"current": false,
"hideEmpty": false,
"hideZero": false,
"max": false,
"min": false,
"rightSide": false,
"show": false,
"total": false,
"values": false
},
"lines": true,
"linewidth": 1,
"nullPointMode": "null",
"percentage": false,
"pointradius": 2,
"points": false,
"spaceLength": 10,
"stack": false,
"steppedLine": false,
"targets": [
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "aptos_consensus_buffer_manager_retry_count{chain_name=~\"$chain_name\", namespace=~\"$namespace\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}: {{op}}",
"format": "time_series"
}
],
"tooltip": { "shared": true, "value_type": "individual" },
"xaxis": { "format": "", "logBase": 0, "show": true },
"yaxes": [
{ "format": "none", "logBase": 1, "show": true },
{ "format": "short", "logBase": 1, "show": true }
],
"fieldConfig": {
"defaults": {
"unit": "",
"color": { "mode": "" },
"thresholds": { "mode": "", "steps": null },
"custom": {
"axisPlacement": "",
"barAlignment": 0,
"drawStyle": "",
"fillOpacity": 0,
"gradientMode": "",
"lineInterpolation": "",
"lineWidth": 0,
"pointSize": 0,
"showPoints": "",
"spanNulls": false,
"hideFrom": { "legend": false, "tooltip": false, "viz": false },
"lineStyle": { "fill": "" },
"scaleDistribution": { "type": "" },
"stacking": { "group": "", "mode": "" },
"thresholdsStyle": { "mode": "" }
}
}
}
},
{
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"editable": false,
Expand Down Expand Up @@ -2407,6 +2487,122 @@
}
}
},
{
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"editable": false,
"error": false,
"gridPos": { "h": 8, "w": 8, "x": 8, "y": 118 },
"id": 130,
"isNew": false,
"span": 0,
"title": "Back pressure",
"description": "",
"transparent": false,
"type": "timeseries",
"targets": [
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "consensus_gauge{chain_name=~\"$chain_name\", namespace=~\"$namespace\", kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", op=\"back_pressure\"} \u003e 10",
"intervalFactor": 1,
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
],
"options": {
"legend": { "calcs": [], "displayMode": "list", "placement": "bottom" },
"tooltip": { "mode": "multi" }
},
"fieldConfig": {
"defaults": {
"unit": "short",
"color": { "mode": "palette-classic" },
"thresholds": {
"mode": "absolute",
"steps": [
{ "color": "green", "value": null },
{ "color": "red", "value": 80 }
]
},
"custom": {
"axisLabel": "# blocks",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"showPoints": "never",
"spanNulls": false,
"hideFrom": { "legend": false, "tooltip": false, "viz": false },
"lineStyle": { "fill": "" },
"scaleDistribution": { "type": "linear" },
"stacking": { "group": "A", "mode": "none" },
"thresholdsStyle": { "mode": "off" }
}
}
}
},
{
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"editable": false,
"error": false,
"gridPos": { "h": 8, "w": 8, "x": 16, "y": 118 },
"id": 131,
"isNew": false,
"span": 0,
"title": "Sync only",
"description": "",
"transparent": false,
"type": "timeseries",
"targets": [
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "consensus_gauge{chain_name=~\"$chain_name\", namespace=~\"$namespace\", kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\", op=\"sync_only\"}",
"intervalFactor": 1,
"legendFormat": "{{kubernetes_pod_name}}-{{role}}",
"format": "time_series"
}
],
"options": {
"legend": { "calcs": [], "displayMode": "list", "placement": "bottom" },
"tooltip": { "mode": "multi" }
},
"fieldConfig": {
"defaults": {
"unit": "bool",
"color": { "mode": "palette-classic" },
"thresholds": {
"mode": "absolute",
"steps": [
{ "color": "green", "value": null },
{ "color": "red", "value": 80 }
]
},
"custom": {
"axisLabel": "# blocks",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"showPoints": "never",
"spanNulls": false,
"hideFrom": { "legend": false, "tooltip": false, "viz": false },
"lineStyle": { "fill": "" },
"scaleDistribution": { "type": "linear" },
"stacking": { "group": "A", "mode": "none" },
"thresholdsStyle": { "mode": "off" }
}
}
}
},
{
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"editable": false,
Expand Down Expand Up @@ -3481,7 +3677,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(aptos_consensus_channel_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"enqueued\", kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$__rate_interval])",
"expr": "rate(aptos_consensus_channel_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"enqueued\", kubernetes_pod_name=~\"$kubernetes_pod_name\", role=~\"$role\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}"
}
],
Expand Down Expand Up @@ -3536,7 +3732,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(aptos_consensus_channel_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$__rate_interval])",
"expr": "rate(aptos_consensus_channel_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}"
}
],
Expand Down Expand Up @@ -3647,7 +3843,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(aptos_consensus_round_manager_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"enqueued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$__rate_interval])",
"expr": "rate(aptos_consensus_round_manager_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"enqueued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}"
}
],
Expand Down Expand Up @@ -3702,7 +3898,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(aptos_consensus_round_manager_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$__rate_interval])",
"expr": "rate(aptos_consensus_round_manager_msgs_count{chain_name=~\"$chain_name\", namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}"
}
],
Expand Down Expand Up @@ -3868,7 +4064,7 @@
{
"refId": "A",
"datasource": { "type": "prometheus", "uid": "${Datasource}" },
"expr": "rate(aptos_consensus_buffer_manager_msgs_count{chain_name=~\"$chain_name\",namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$__rate_interval])",
"expr": "rate(aptos_consensus_buffer_manager_msgs_count{chain_name=~\"$chain_name\",namespace=\"$namespace\", state=\"dequeued\", kubernetes_pod_name=~\"$kubernetes_pod_name\"}[$interval])",
"legendFormat": "{{kubernetes_pod_name}}-{{role}}"
}
],
Expand Down Expand Up @@ -4414,7 +4610,7 @@
},
"refresh": false,
"schemaVersion": 37,
"version": 33,
"version": 45,
"links": [
{
"title": "Other Dashboards",
Expand Down
Binary file modified dashboards/consensus.json.gz
View file
Binary file not shown.
Loading

0 comments on commit 2cb091e

Please sign in to comment.