bstein/titan-iac
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

atlas overview: fix availability scaling

Browse Source
This commit is contained in:
Brad Stein 2025-12-12 16:36:47 -03:00
parent 87fbba0d3e
commit 0a0966db78
3 changed files with 14 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
scripts/dashboards_render_atlas.py
View File

@ -214,7 +214,7 @@ UPTIME_AVAIL_EXPR = (
f"min(({CONTROL_READY_FRACTION_EXPR}), ({TRAEFIK_READY_EXPR}))"
)
UPTIME_AVG_EXPR = f"avg_over_time(({UPTIME_AVAIL_EXPR})[{UPTIME_WINDOW}:5m])"
UPTIME_PERCENT_EXPR = f"({UPTIME_AVG_EXPR}) * 100"
UPTIME_PERCENT_EXPR = UPTIME_AVG_EXPR
UPTIME_NINES_EXPR = f"-log10(1 - clamp_max({UPTIME_AVG_EXPR}, 0.999999999))"
UPTIME_THRESHOLDS = {
"mode": "absolute",
@ -229,9 +229,9 @@ UPTIME_PERCENT_THRESHOLDS = {
"mode": "absolute",
"steps": [
{"color": "red", "value": None},
{"color": "orange", "value": 99},
{"color": "yellow", "value": 99.9},
{"color": "green", "value": 99.99},
{"color": "orange", "value": 0.99},
{"color": "yellow", "value": 0.999},
{"color": "green", "value": 0.9999},
],
}
PROBLEM_TABLE_EXPR = (
@ -641,7 +641,6 @@ def build_overview():
"expr": UPTIME_PERCENT_EXPR,
"kind": "stat",
"thresholds": UPTIME_PERCENT_THRESHOLDS,
"value_suffix": "%",
"unit": "percentunit",
"decimals": 3,
"text_mode": "value",

11
services/monitoring/dashboards/atlas-overview.json
View File

@ -226,7 +226,7 @@
},
"targets": [
{
"expr": "(avg_over_time((min(((sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-0a|titan-0b|titan-0c\"}) / 3)), ((sum(kube_deployment_status_replicas_available{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}) / clamp_min(sum(kube_deployment_spec_replicas{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}), 1)))))[30d:5m])) * 100",
"expr": "avg_over_time((min(((sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-0a|titan-0b|titan-0c\"}) / 3)), ((sum(kube_deployment_status_replicas_available{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}) / clamp_min(sum(kube_deployment_spec_replicas{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}), 1)))))[30d:5m])",
"refId": "A"
}
],
@ -245,22 +245,21 @@
},
{
"color": "orange",
"value": 99
"value": 0.99
},
{
"color": "yellow",
"value": 99.9
"value": 0.999
},
{
"color": "green",
"value": 99.99
"value": 0.9999
}
]
},
"unit": "percentunit",
"custom": {
"displayMode": "auto",
"valueSuffix": "%"
"displayMode": "auto"
},
"decimals": 3
},

11
services/monitoring/grafana-dashboard-overview.yaml
View File

@ -235,7 +235,7 @@ data:
},
"targets": [
{
"expr": "(avg_over_time((min(((sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-0a|titan-0b|titan-0c\"}) / 3)), ((sum(kube_deployment_status_replicas_available{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}) / clamp_min(sum(kube_deployment_spec_replicas{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}), 1)))))[30d:5m])) * 100",
"expr": "avg_over_time((min(((sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-0a|titan-0b|titan-0c\"}) / 3)), ((sum(kube_deployment_status_replicas_available{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}) / clamp_min(sum(kube_deployment_spec_replicas{namespace=~\"traefik|kube-system\",deployment=\"traefik\"}), 1)))))[30d:5m])",
"refId": "A"
}
],
@ -254,22 +254,21 @@ data:
},
{
"color": "orange",
"value": 99
"value": 0.99
},
{
"color": "yellow",
"value": 99.9
"value": 0.999
},
{
"color": "green",
"value": 99.99
"value": 0.9999
}
]
},
"unit": "percentunit",
"custom": {
"displayMode": "auto",
"valueSuffix": "%"
"displayMode": "auto"
},
"decimals": 3
},