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

feature/atlas-monitoring #3

Merged
bstein merged 71 commits from feature/atlas-monitoring into main 2025-12-02 20:52:36 +00:00
Conversation 0 Commits 71 Files Changed 23 +8 -12
3 changed files with 8 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit cc62f497e9 - Show all commits

12
scripts/render_dashboards.py
View File

@ -166,17 +166,13 @@ def node_io_expr(scope=""):
def namespace_cpu_share_expr(): def namespace_cpu_share_expr():
return ( selected = f"( {NAMESPACE_CPU_RAW} ) and on(namespace) ( {NAMESPACE_COMBINED_FILTER} )"
f"(100 * ( {NAMESPACE_CPU_RAW} ) / sum( {NAMESPACE_CPU_RAW} )) * on(namespace) group_left() " return f"100 * ( {selected} ) / sum( {selected} )"
f"( {NAMESPACE_COMBINED_FILTER} )"
)
def namespace_ram_share_expr(): def namespace_ram_share_expr():
return ( selected = f"( {NAMESPACE_RAM_RAW} ) and on(namespace) ( {NAMESPACE_COMBINED_FILTER} )"
f"(100 * ( {NAMESPACE_RAM_RAW} ) / sum( {NAMESPACE_RAM_RAW} )) * on(namespace) group_left() " return f"100 * ( {selected} ) / sum( {selected} )"
f"( {NAMESPACE_COMBINED_FILTER} )"
)
PROBLEM_PODS_EXPR = 'sum(max by (namespace,pod) (kube_pod_status_phase{phase!~"Running|Succeeded"}))' PROBLEM_PODS_EXPR = 'sum(max by (namespace,pod) (kube_pod_status_phase{phase!~"Running|Succeeded"}))'

4
services/monitoring/dashboards/atlas-overview.json
View File

@ -722,7 +722,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "(100 * ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) / sum( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) )) * on(namespace) group_left() ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) )", "expr": "100 * ( ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) ) / sum( ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) )",
"refId": "A", "refId": "A",
"legendFormat": "{{namespace}}" "legendFormat": "{{namespace}}"
} }
@ -764,7 +764,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "(100 * ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) / sum( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) )) * on(namespace) group_left() ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) )", "expr": "100 * ( ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) ) / sum( ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) )",
"refId": "A", "refId": "A",
"legendFormat": "{{namespace}}" "legendFormat": "{{namespace}}"
} }

4
services/monitoring/grafana-dashboard-overview.yaml
View File

@ -731,7 +731,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "(100 * ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) / sum( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) )) * on(namespace) group_left() ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) )", "expr": "100 * ( ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) ) / sum( ( sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) )",
"refId": "A", "refId": "A",
"legendFormat": "{{namespace}}" "legendFormat": "{{namespace}}"
} }
@ -773,7 +773,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "(100 * ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) / sum( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) )) * on(namespace) group_left() ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) )", "expr": "100 * ( ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) ) / sum( ( sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) ) and on(namespace) ( topk(10, (sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace)) + (sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace) / 1e9)) ) )",
"refId": "A", "refId": "A",
"legendFormat": "{{namespace}}" "legendFormat": "{{namespace}}"
} }