gitadmin/devops-infra-helm-charts-gcp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

added repo

Browse Source
This commit is contained in:
Your Name
2026-08-26 03:39:42 +05:30
parent 45c25a95af
commit b8575bb8b9
6889 changed files with 1217125 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-HTTP_4XX_Alert_Contour.yaml
+16
View File
@@ -0,0 +1,16 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_4XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-offer-platform-supply_prd-offer-platform-supply-primary_80",envoy_response_code=~"[4].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 2000
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-supply
env: prd
priority: p1
annotations:
summary: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}\n LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-HTTP_5XX_Alert_Contour.yaml
+16
View File
@@ -0,0 +1,16 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_5XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-offer-platform-supply_prd-offer-platform-supply-primary_80",envoy_response_code=~"[5].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 100
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-supply
env: prd
priority: p1
annotations:
summary: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}\n LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-P99_Latency_Alert_Contour.yaml
+16
View File
@@ -0,0 +1,16 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: P99_Latency_Alert_Contour_EKS
expr: histogram_quantile(0.99, sum(rate(envoy_cluster_upstream_rq_time_bucket{envoy_cluster_name=~"prd-offer-platform-supply_prd-offer-platform-supply-primary_80"}[1m])) by (cluster, envoy_cluster_name, le, job)) >= 200
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-supply
env: prd
priority: p1
annotations:
summary: "EKS P99_Latency_Alert_Contour on {{ $labels.service }}"
description: "EKS P99_Latency_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}\n LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-supply/custom_alerts/offers-offer-platform-supply-FY_CACHE_HIT_BELOW_THRESHOLD_ALERT_EKS.yaml
+16
View File
@@ -0,0 +1,16 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: FY_CACHE_HIT_BELOW_THRESHOLD_ALERT_EKS_EKS
expr: (sum(rate(op_offer_cache_hit_total{job="kubernetes-pods",service=~"offers-offer-platform-supply",cache_hit="redis"}[1m]))/(sum(rate(op_offer_cache_hit_total{job="kubernetes-pods",service=~"offers-offer-platform-supply",cache_hit="redis"}[1m])) + sum(rate(op_offer_cache_hit_total{job="kubernetes-pods",service=~"offers-offer-platform-supply",cache_hit="hbase"}[1m])))) * 100 < 90
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-supply
env: prd
priority: p0
annotations:
summary: "EKS FY_CACHE_HIT_BELOW_THRESHOLD_ALERT_EKS on {{ $labels.service }}"
description: "EKS FY_CACHE_HIT_BELOW_THRESHOLD_ALERT_EKS on {{ $labels.service }} VALUE = {{ $value }}\n LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-supply/custom_alerts/offers-offer-platform-supply-HBASE_AVG_TIME_EKS.yaml
+16
View File
@@ -0,0 +1,16 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HBASE_AVG_TIME_EKS_EKS
expr: (sum by (service,method,parent) (HBASE_execTime_sum{service=~"offers-offer-platform-supply"}))/ (sum by (service,parent,method) (HBASE_execTime_count{service=~"offers-offer-platform-supply"})) > 100
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-supply
env: prd
priority: p0
annotations:
summary: "EKS HBASE_AVG_TIME_EKS on {{ $labels.service }}"
description: "EKS HBASE_AVG_TIME_EKS on {{ $labels.service }} VALUE = {{ $value }}\n LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-worker-HTTP_4XX_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_4XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-offer-platform-worker-supply_prd-offer-platform-worker-supply_80",envoy_response_code=~"[4].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 50
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-worker-supply
env: prd
priority: p1
annotations:
summary: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-worker-HTTP_5XX_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_5XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-offer-platform-worker-supply_prd-offer-platform-worker-supply_80",envoy_response_code=~"[5].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 50
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-worker-supply
env: prd
priority: p1
annotations:
summary: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/offers-offer-platform-worker-P99_Latency_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: P99_Latency_Alert_Contour_EKS
expr: histogram_quantile(0.99, sum(rate(envoy_cluster_upstream_rq_time_bucket{envoy_cluster_name=~"prd-offer-platform-worker-supply_prd-offer-platform-worker-supply_80"}[1m])) by (cluster, envoy_cluster_name, le, job)) >= 2500
for: 5m
labels:
severity: critical
bu: supply
team: offers
service: offers-offer-platform-worker-supply
env: prd
priority: p1
annotations:
summary: "EKS P99_Latency_Alert_Contour on {{ $labels.service }}"
description: "EKS P99_Latency_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/supplier-platform-promotions-HTTP_4XX_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_4XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-supplier-platform-promotions_prd-supplier-platform-promotions_80",envoy_response_code=~"[4].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 100
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: supplier-platform-promotions
env: prd
priority: p1
annotations:
summary: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_4XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/supplier-platform-promotions-HTTP_5XX_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: HTTP_5XX_Alert_Contour_EKS
expr: sum(rate(envoy_cluster_upstream_rq{envoy_cluster_name=~"prd-supplier-platform-promotions_prd-supplier-platform-promotions_80",envoy_response_code=~"[5].*"}[1m])) by (cluster, envoy_cluster_name, job) * 60 >= 60
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: supplier-platform-promotions
env: prd
priority: p1
annotations:
summary: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }}"
description: "EKS HTTP_5XX_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"
helm-templates/victoria-metrics-alerts-config/configmap/supply/offers/supplier-platform-promotions-P99_Latency_Alert_Contour.yaml
+17
View File
@@ -0,0 +1,17 @@
groups:
- name: EKS-App-Alerts
rules:
- alert: P99_Latency_Alert_Contour_EKS
expr: histogram_quantile(0.99, sum(rate(envoy_cluster_upstream_rq_time_bucket{envoy_cluster_name=~"prd-supplier-platform-promotions_prd-supplier-platform-promotions_80"}[1m])) by (cluster, envoy_cluster_name, le, job)) >= 4000
for: 1m
labels:
severity: critical
bu: supply
team: offers
service: supplier-platform-promotions
env: prd
priority: p1
annotations:
summary: "EKS P99_Latency_Alert_Contour on {{ $labels.service }}"
description: "EKS P99_Latency_Alert_Contour on {{ $labels.service }} VALUE = {{ $value }}
LABELS: {{ $labels }}"