clusterforge/otel-lgtm-stack/OpenTelemetryCollector_otel-collector-metrics-rest.yaml
2025-10-06 09:34:03 +00:00

149 lines
4.8 KiB
YAML

---
apiVersion: opentelemetry.io/v1beta1
kind: OpenTelemetryCollector
metadata:
name: otel-collector-metrics-rest
namespace: otel-lgtm-stack
spec:
config:
exporters:
debug:
verbosity: detailed
otlp:
endpoint: http://lgtm-stack.otel-lgtm-stack.svc.cluster.local:4317
tls:
insecure: true
processors:
attributes:
actions:
- action: insert
key: k8s_cluster_name
value: cluster-name
batch:
send_batch_size: 2000
timeout: 10s
memory_limiter:
check_interval: 5s
limit_percentage: 80
spike_limit_percentage: 25
receivers:
prometheus:
config:
scrape_configs:
- job_name: otel-collector
scrape_interval: 30s
static_configs:
- targets:
- localhost:8888
- dns_sd_configs:
- names:
- opencost-prometheus-opencost-exporter.monitoring
port: 9003
type: A
honor_labels: true
job_name: opencost
metrics_path: /metrics
scheme: http
scrape_interval: 1m
scrape_timeout: 10s
- job_name: gpu-operator-metrics-exporter
kubernetes_sd_configs:
- role: node
metrics_path: /metrics
relabel_configs:
- action: keep
regex: true
source_labels:
- __meta_kubernetes_node_label_feature_node_kubernetes_io_amd_gpu
- regex: (.+)
replacement: $1:32500
source_labels:
- __meta_kubernetes_node_address_InternalIP
target_label: __address__
- source_labels:
- __meta_kubernetes_node_name
target_label: hostname
- job_name: minio-cluster-metrics
metrics_path: /minio/v2/metrics/cluster
scheme: http
static_configs:
- targets:
- minio.minio-tenant-default.svc.cluster.local
- job_name: minio-bucket-metrics
metrics_path: /minio/v2/metrics/bucket
scheme: http
static_configs:
- targets:
- minio.minio-tenant-default.svc.cluster.local
- job_name: minio-resource-metrics
metrics_path: /minio/v2/metrics/resource
scheme: http
static_configs:
- targets:
- minio.minio-tenant-default.svc.cluster.local
- job_name: argocd-controller
metrics_path: /metrics
scheme: http
static_configs:
- targets:
- argocd-metrics.argocd.svc.cluster.local:8082
- job_name: argocd-applicationset
metrics_path: /metrics
scheme: http
static_configs:
- targets:
- argocd-applicationset-controller.argocd.svc.cluster.local:8080
- job_name: argocd-repo-server
metrics_path: /metrics
scheme: http
static_configs:
- targets:
- argocd-repo-server.argocd.svc.cluster.local:8084
- job_name: longhorn
metrics_path: /metrics
scheme: http
static_configs:
- targets:
- longhorn-backend.longhorn.svc.cluster.local:9500
- job_name: chrony-exporter
kubernetes_sd_configs:
- role: pod
relabel_configs:
- action: keep
regex: chrony-exporter
source_labels:
- __meta_kubernetes_pod_label_app
- regex: (.*)
replacement: $1:9123
source_labels:
- __meta_kubernetes_pod_ip
target_label: __address__
- source_labels:
- __meta_kubernetes_pod_node_name
target_label: k8s_node_name
service:
pipelines:
metrics:
exporters:
- otlp
processors:
- memory_limiter
- batch
- attributes
receivers:
- prometheus
image: ghcr.io/open-telemetry/opentelemetry-collector-releases/opentelemetry-collector-contrib:0.113.0
mode: deployment
podAnnotations:
prometheus.io/port: "8888"
prometheus.io/scrape: "true"
replicas: 1
resources:
limits:
cpu: "2"
memory: 8Gi
requests:
cpu: 500m
memory: 1Gi
serviceAccount: otel-collector