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

monitoring: restore top1 hottest stats

Browse Source
This commit is contained in:
Brad Stein 2025-11-17 21:20:19 -03:00
parent 53427cc8fa
commit 76d3dc6ae2
5 changed files with 12 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
scripts/render_dashboards.py
View File

@ -151,8 +151,8 @@ def topk_with_node(expr):
def node_net_expr(scope=""): def node_net_expr(scope=""):
base = ( base = (
'sum by (instance) (' 'sum by (instance) ('
'rate(container_network_receive_bytes_total{namespace!="" ,pod!=""}[5m]) ' 'rate(node_network_receive_bytes_total{device!~"lo"}[5m]) '
'+ rate(container_network_transmit_bytes_total{namespace!="" ,pod!=""}[5m]))' '+ rate(node_network_transmit_bytes_total{device!~"lo"}[5m]))'
) )
return scoped_node_expr(base, scope) return scoped_node_expr(base, scope)
@ -471,8 +471,8 @@ def build_overview():
) )
hottest = [ hottest = [
(7, "Hottest node: CPU", node_cpu_expr(), "percent"), (7, "Hottest node: CPU", topk_with_node(node_cpu_expr()), "percent"),
(8, "Hottest node: RAM", node_mem_expr(), "percent"), (8, "Hottest node: RAM", topk_with_node(node_mem_expr()), "percent"),
(9, "Hottest node: NET (rx+tx)", topk_with_node(node_net_expr()), "Bps"), (9, "Hottest node: NET (rx+tx)", topk_with_node(node_net_expr()), "Bps"),
(10, "Hottest node: I/O (r+w)", topk_with_node(node_io_expr()), "Bps"), (10, "Hottest node: I/O (r+w)", topk_with_node(node_io_expr()), "Bps"),
] ]

2
services/monitoring/dashboards/atlas-network.json
View File

@ -202,7 +202,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) ((sum by (instance) (rate(container_network_receive_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]) + rate(container_network_transmit_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "avg by (node) ((sum by (instance) (rate(node_network_receive_bytes_total{device!~\"lo\"}[5m]) + rate(node_network_transmit_bytes_total{device!~\"lo\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}" "legendFormat": "{{node}}"
} }

6
services/monitoring/dashboards/atlas-overview.json
View File

@ -438,7 +438,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "label_replace(topk(1, avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true
@ -511,7 +511,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "label_replace(topk(1, avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true
@ -584,7 +584,7 @@
}, },
"targets": [ "targets": [
{ {
"expr": "label_replace(topk(1, avg by (node) ((sum by (instance) (rate(container_network_receive_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]) + rate(container_network_transmit_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")", "expr": "label_replace(topk(1, avg by (node) ((sum by (instance) (rate(node_network_receive_bytes_total{device!~\"lo\"}[5m]) + rate(node_network_transmit_bytes_total{device!~\"lo\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true

2
services/monitoring/grafana-dashboard-network.yaml
View File

@ -211,7 +211,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) ((sum by (instance) (rate(container_network_receive_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]) + rate(container_network_transmit_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "avg by (node) ((sum by (instance) (rate(node_network_receive_bytes_total{device!~\"lo\"}[5m]) + rate(node_network_transmit_bytes_total{device!~\"lo\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}" "legendFormat": "{{node}}"
} }

6
services/monitoring/grafana-dashboard-overview.yaml
View File

@ -447,7 +447,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "label_replace(topk(1, avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true
@ -520,7 +520,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))", "expr": "label_replace(topk(1, avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true
@ -593,7 +593,7 @@ data:
}, },
"targets": [ "targets": [
{ {
"expr": "label_replace(topk(1, avg by (node) ((sum by (instance) (rate(container_network_receive_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]) + rate(container_network_transmit_bytes_total{namespace!=\"\" ,pod!=\"\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")", "expr": "label_replace(topk(1, avg by (node) ((sum by (instance) (rate(node_network_receive_bytes_total{device!~\"lo\"}[5m]) + rate(node_network_transmit_bytes_total{device!~\"lo\"}[5m]))) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))), \"__name__\", \"$1\", \"node\", \"(.*)\")",
"refId": "A", "refId": "A",
"legendFormat": "{{node}}", "legendFormat": "{{node}}",
"instant": true "instant": true