1054 lines
26 KiB
Plaintext
1054 lines
26 KiB
Plaintext
// gce node exporter
|
|
discovery.gce "node_exporter_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 9100
|
|
}
|
|
discovery.gce "node_exporter_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 9100
|
|
}
|
|
discovery.gce "node_exporter_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 9100
|
|
}
|
|
discovery.relabel "node_exporter_gce" {
|
|
targets = concat(
|
|
discovery.gce.node_exporter_a.targets,
|
|
discovery.gce.node_exporter_b.targets,
|
|
discovery.gce.node_exporter_c.targets,
|
|
)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_status"]
|
|
regex = "RUNNING"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_goog-k8s-cluster-name"]
|
|
regex = "k8s"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_monitoring"]
|
|
regex = "no"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_name"]
|
|
regex = "gke-k8s-.*"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
target_label = "project"
|
|
replacement = "meesho-datascience-prd-0622"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_name"]
|
|
target_label = "name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_id"]
|
|
target_label = "instance_id"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_status"]
|
|
target_label = "status"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_private_ip"]
|
|
target_label = "private_ip"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_public_ip"]
|
|
target_label = "public_ip"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_component"]
|
|
target_label = "component"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_service"]
|
|
target_label = "service"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_type"]
|
|
target_label = "type"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_alerts"]
|
|
target_label = "alerts"
|
|
}
|
|
}
|
|
prometheus.scrape "node_exporter_gce" {
|
|
targets = discovery.relabel.node_exporter_gce.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "node_exporter_gce"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// kube state metrics
|
|
discovery.relabel "kube_state_metrics_pod" {
|
|
targets = [{__address__ = "kube-state-metrics-datascience-prd.kube-system.svc.cluster.local:8080",}]
|
|
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "kube_state_metrics_pod" {
|
|
targets = discovery.relabel.kube_state_metrics_pod.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "kube_state_metrics_pod"
|
|
scrape_interval = "15s"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// cadvisor metrics
|
|
discovery.kubernetes "kubernetes_nodes" {
|
|
role = "node"
|
|
}
|
|
discovery.relabel "cadvisor_kubernetes_nodes" {
|
|
targets = discovery.kubernetes.kubernetes_nodes.targets
|
|
|
|
rule {
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
target_label = "cluster"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
replacement = "kubernetes.default.svc:443"
|
|
target_label = "__address__"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_node_name"]
|
|
regex = "(.+)"
|
|
target_label = "__metrics_path__"
|
|
replacement = "/api/v1/nodes/$1/proxy/metrics/cadvisor"
|
|
}
|
|
}
|
|
prometheus.scrape "cadvisor_kubernetes_nodes" {
|
|
targets = discovery.relabel.cadvisor_kubernetes_nodes.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "cadvisor_kubernetes_nodes"
|
|
scheme = "https"
|
|
scrape_interval = "15s"
|
|
authorization {
|
|
type = "Bearer"
|
|
credentials_file = "/var/run/secrets/kubernetes.io/serviceaccount/token"
|
|
}
|
|
tls_config {
|
|
ca_file = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
|
|
insecure_skip_verify = true
|
|
}
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// kubelet
|
|
discovery.relabel "kubelet_kubernetes_nodes" {
|
|
targets = discovery.kubernetes.kubernetes_nodes.targets
|
|
|
|
rule {
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
target_label = "cluster"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
target_label = "__address__"
|
|
replacement = "kubernetes.default.svc:443"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_node_name"]
|
|
regex = "(.+)"
|
|
target_label = "__metrics_path__"
|
|
replacement = "/api/v1/nodes/$1/proxy/metrics"
|
|
}
|
|
}
|
|
prometheus.scrape "kubelet_kubernetes_nodes" {
|
|
targets = discovery.relabel.kubelet_kubernetes_nodes.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "kubelet_kubernetes_nodes"
|
|
scheme = "https"
|
|
scrape_interval = "15s"
|
|
authorization {
|
|
type = "Bearer"
|
|
credentials_file = "/var/run/secrets/kubernetes.io/serviceaccount/token"
|
|
}
|
|
tls_config {
|
|
ca_file = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
|
|
insecure_skip_verify = true
|
|
}
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// prometheus_kubernetes_endpoints
|
|
discovery.kubernetes "kubernetes_endpoints" {
|
|
role = "endpoints"
|
|
}
|
|
discovery.relabel "prometheus_kubernetes_endpoints" {
|
|
targets = discovery.kubernetes.kubernetes_endpoints.targets
|
|
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_container_init"]
|
|
regex = "true"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_service_annotation_prometheus_io_port"]
|
|
target_label = "__meta_kubernetes_pod_container_port_number"
|
|
action = "keepequal"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_service_annotation_prometheus_io_scrape"]
|
|
regex = "true"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_service_annotation_prometheus_io_path"]
|
|
regex = "(.+)"
|
|
target_label = "__metrics_path__"
|
|
}
|
|
rule {
|
|
source_labels = ["__address__", "__meta_kubernetes_service_annotation_prometheus_io_port"]
|
|
regex = "([^:]+)(?::\\d+)?;(\\d+)"
|
|
target_label = "__address__"
|
|
replacement = "$1:$2"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_service_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_namespace"]
|
|
target_label = "kubernetes_namespace"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_endpoint_node_name"]
|
|
target_label = "kubernetes_node"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_service_name"]
|
|
target_label = "kubernetes_service_name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_endpoints_name"]
|
|
target_label = "kubernetes_endpoints_name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_name"]
|
|
target_label = "kubernetes_pod_name"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "prometheus_kubernetes_endpoints" {
|
|
targets = discovery.relabel.prometheus_kubernetes_endpoints.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "prometheus_kubernetes_endpoints"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// prometheus_kubernetes_pods
|
|
discovery.kubernetes "kubernetes_pods" {
|
|
role = "pod"
|
|
}
|
|
discovery.relabel "prometheus_kubernetes_pods" {
|
|
targets = discovery.kubernetes.kubernetes_pods.targets
|
|
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_container_init"]
|
|
regex = "true"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_prometheus_io_port"]
|
|
target_label = "__meta_kubernetes_pod_container_port_number"
|
|
action = "keepequal"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_prometheus_io_scrape"]
|
|
regex = "true"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_prometheus_io_path"]
|
|
regex = "(.+)"
|
|
target_label = "__metrics_path__"
|
|
}
|
|
rule {
|
|
source_labels = ["__address__", "__meta_kubernetes_pod_annotation_prometheus_io_port"]
|
|
regex = "([^:]+)(?::\\d+)?;(\\d+)"
|
|
target_label = "__address__"
|
|
replacement = "$1:$2"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_pod_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_namespace"]
|
|
target_label = "kubernetes_namespace"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_node_name"]
|
|
target_label = "kubernetes_node"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_name"]
|
|
target_label = "kubernetes_pod_name"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "prometheus_kubernetes_pods" {
|
|
targets = discovery.relabel.prometheus_kubernetes_pods.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "prometheus_kubernetes_pods"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// jmx_kubernetes_pods
|
|
discovery.relabel "jmx_kubernetes_pods" {
|
|
targets = discovery.kubernetes.kubernetes_pods.targets
|
|
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_container_init"]
|
|
regex = "true"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_jmx_io_port"]
|
|
target_label = "__meta_kubernetes_pod_container_port_number"
|
|
action = "keepequal"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_jmx_io_scrape"]
|
|
regex = "true"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_annotation_jmx_io_path"]
|
|
regex = "(.+)"
|
|
target_label = "__metrics_path__"
|
|
}
|
|
rule {
|
|
source_labels = ["__address__", "__meta_kubernetes_pod_annotation_jmx_io_port"]
|
|
regex = "([^:]+)(?::\\d+)?;(\\d+)"
|
|
target_label = "__address__"
|
|
replacement = "$1:$2"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_pod_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_namespace"]
|
|
target_label = "kubernetes_namespace"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_node_name"]
|
|
target_label = "kubernetes_node"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_name"]
|
|
target_label = "kubernetes_pod_name"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "jmx_kubernetes_pods" {
|
|
targets = discovery.relabel.jmx_kubernetes_pods.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "jmx_kubernetes_pods"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// pm2_metrics
|
|
discovery.relabel "pm2_kubernetes_pods" {
|
|
targets = discovery.kubernetes.kubernetes_pods.targets
|
|
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_container_init"]
|
|
regex = "true"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__address__"]
|
|
regex = "(.*9209)"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_pod_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_namespace"]
|
|
target_label = "kubernetes_namespace"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_node_name"]
|
|
target_label = "kubernetes_node"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_name"]
|
|
target_label = "kubernetes_pod_name"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "pm2_kubernetes_pods" {
|
|
targets = discovery.relabel.pm2_kubernetes_pods.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "pm2_kubernetes_pods"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// scylladb_manager_exporter
|
|
discovery.gce "scylladb_manager_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 5090
|
|
}
|
|
discovery.gce "scylladb_manager_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 5090
|
|
}
|
|
discovery.gce "scylladb_manager_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 5090
|
|
}
|
|
discovery.relabel "scylladb_manager_gce" {
|
|
targets = concat(
|
|
discovery.gce.scylladb_manager_a.targets,
|
|
discovery.gce.scylladb_manager_b.targets,
|
|
discovery.gce.scylladb_manager_c.targets,
|
|
)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_label_component"]
|
|
regex = "scyl"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_service"]
|
|
target_label = "service"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_type"]
|
|
target_label = "type"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_cluster"]
|
|
target_label = "cluster"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_name"]
|
|
target_label = "name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_priority"]
|
|
target_label = "priority"
|
|
}
|
|
}
|
|
prometheus.scrape "scylladb_manager_gce" {
|
|
targets = discovery.relabel.scylladb_manager_gce.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "scylladb_manager_gce"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// scylladb_exporter
|
|
discovery.gce "scylladb_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 9180
|
|
}
|
|
discovery.gce "scylladb_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 9180
|
|
}
|
|
discovery.gce "scylladb_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 9180
|
|
}
|
|
discovery.relabel "scylladb_gce" {
|
|
targets = concat(
|
|
discovery.gce.scylladb_a.targets,
|
|
discovery.gce.scylladb_b.targets,
|
|
discovery.gce.scylladb_c.targets,
|
|
)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_label_component"]
|
|
regex = "scyl"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_service"]
|
|
target_label = "service"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_type"]
|
|
target_label = "type"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_cluster"]
|
|
target_label = "cluster"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_name"]
|
|
target_label = "name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_priority"]
|
|
target_label = "priority"
|
|
}
|
|
}
|
|
prometheus.scrape "scylladb_gce" {
|
|
targets = discovery.relabel.scylladb_gce.output
|
|
forward_to = [prometheus.relabel.scylladb_gce.receiver]
|
|
job_name = "scylladb_gce"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
prometheus.relabel "scylladb_gce" {
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
|
|
rule {
|
|
source_labels = ["__name__", "scheduling_group_name"]
|
|
regex = "(scylla_storage_proxy_coordinator_.*_bucket;)(atexit|gossip|mem_compaction|memtable|streaming|background_reclaim|compaction|main|memtable_to_cache)"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "(.+)"
|
|
target_label = "CPU"
|
|
replacement = "cpu"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "(.+)"
|
|
target_label = "CQL"
|
|
replacement = "cql"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "(.+)"
|
|
target_label = "OS"
|
|
replacement = "os"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "(.+)"
|
|
target_label = "IO"
|
|
replacement = "io"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "(.+)"
|
|
target_label = "Errors"
|
|
replacement = "errors"
|
|
}
|
|
rule {
|
|
regex = "help|exported_instance"
|
|
action = "labeldrop"
|
|
}
|
|
rule {
|
|
source_labels = ["version"]
|
|
regex = "([0-9]+\\.[0-9]+)(\\.?[0-9]*).*"
|
|
target_label = "svr"
|
|
replacement = "$1$2"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_write_latency_summary;0.990*)"
|
|
target_label = "__name__"
|
|
replacement = "wlatencyp99"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_write_latency_summary;0.950*)"
|
|
target_label = "__name__"
|
|
replacement = "wlatencyp95"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_write_latency_summary;0.50*)"
|
|
target_label = "__name__"
|
|
replacement = "wlatencya"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_read_latency_summary;0.990*)"
|
|
target_label = "__name__"
|
|
replacement = "rlatencyp99"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_read_latency_summary;0.950*)"
|
|
target_label = "__name__"
|
|
replacement = "rlatencyp95"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_read_latency_summary;0.50*)"
|
|
target_label = "__name__"
|
|
replacement = "rlatencya"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_write_latency_summary;0.950*)"
|
|
target_label = "__name__"
|
|
replacement = "caswlatencyp95"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_write_latency_summary;0.990*)"
|
|
target_label = "__name__"
|
|
replacement = "caswlatencyp99"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_write_latency_summary;0.50*)"
|
|
target_label = "__name__"
|
|
replacement = "caswlatencya"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_read_latency_summary;0.950*)"
|
|
target_label = "__name__"
|
|
replacement = "casrlatencyp95"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_read_latency_summary;0.990*)"
|
|
target_label = "__name__"
|
|
replacement = "casrlatencyp99"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__", "quantile"]
|
|
regex = "(scylla_storage_proxy_coordinator_cas_read_latency_summary;0.50*)"
|
|
target_label = "__name__"
|
|
replacement = "casrlatencya"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__"]
|
|
regex = "(.latency..?.?|cas.latency..?.?)"
|
|
target_label = "by"
|
|
replacement = "instance,shard"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__"]
|
|
regex = "(scylla_storage_proxy_coordinator_read_timeouts|scylla_reactor_utilization|scylla_storage_proxy_coordinator_read_timeouts|scylla_storage_proxy_coordinator_read_unavailable|scylla_storage_proxy_coordinator_write_timeouts|scylla_storage_proxy_coordinator_write_unavailable|.latency..?.?)"
|
|
target_label = "level"
|
|
replacement = "2"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__"]
|
|
regex = "(scylla_storage_proxy_coordinator_read_timeouts|scylla_reactor_utilization|scylla_storage_proxy_coordinator_read_timeouts|scylla_storage_proxy_coordinator_read_unavailable|scylla_storage_proxy_coordinator_write_timeouts|scylla_storage_proxy_coordinator_write_unavailable|.latency..?.?)"
|
|
target_label = "dd"
|
|
replacement = "2"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__"]
|
|
regex = "(scylla_node_operation_mode)"
|
|
target_label = "level"
|
|
replacement = "1"
|
|
}
|
|
rule {
|
|
source_labels = ["__name__"]
|
|
regex = "(scylla_node_operation_mode)"
|
|
target_label = "dd"
|
|
replacement = "1"
|
|
}
|
|
rule {
|
|
source_labels = ["scheduling_group_name"]
|
|
regex = "(atexit|gossip|mem_compaction|memtable|streaming|background_reclaim|compaction|main|memtable_to_cache)"
|
|
target_label = "level"
|
|
replacement = ""
|
|
}
|
|
rule {
|
|
source_labels = ["scheduling_group_name"]
|
|
regex = "(atexit|gossip|mem_compaction|memtable|streaming|background_reclaim|compaction|main|memtable_to_cache)"
|
|
target_label = "dd"
|
|
replacement = ""
|
|
}
|
|
}
|
|
|
|
// dragonfly mcache
|
|
discovery.gce "dragonfly_mcache_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 6379
|
|
}
|
|
discovery.gce "dragonfly_mcache_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 6379
|
|
}
|
|
discovery.gce "dragonfly_mcache_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 6379
|
|
}
|
|
discovery.relabel "dragonfly_mcache_gce" {
|
|
targets = concat(
|
|
discovery.gce.dragonfly_mcache_a.targets,
|
|
discovery.gce.dragonfly_mcache_b.targets,
|
|
discovery.gce.dragonfly_mcache_c.targets,)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_label_monitoring"]
|
|
regex = "no"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
regex = "(DR|perf)"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_pod"]
|
|
regex = "(devops)"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_mcache"]
|
|
regex = "(dragonfly)"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_name"]
|
|
target_label = "name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_mcache"]
|
|
target_label = "mcache"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
}
|
|
prometheus.scrape "dragonfly_mcache_gce" {
|
|
targets = discovery.relabel.dragonfly_mcache_gce.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "dragonfly_mcache_gce"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// zookeeper
|
|
discovery.gce "zookeeper_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 7000
|
|
}
|
|
discovery.gce "zookeeper_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 7000
|
|
}
|
|
discovery.gce "zookeeper_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 7000
|
|
}
|
|
discovery.relabel "zookeeper_gce" {
|
|
targets = concat(
|
|
discovery.gce.zookeeper_a.targets,
|
|
discovery.gce.zookeeper_b.targets,
|
|
discovery.gce.zookeeper_c.targets,
|
|
)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_label_service"]
|
|
regex = "(comms-zookeeper)"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
target_label = "component"
|
|
replacement = "zookeeper"
|
|
}
|
|
rule {
|
|
target_label = "priority"
|
|
replacement = "p0"
|
|
}
|
|
rule {
|
|
target_label = "project"
|
|
replacement = "meesho-datascience-prd-0622"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_service"]
|
|
target_label = "app"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_type"]
|
|
target_label = "type"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_instance_name"]
|
|
target_label = "name"
|
|
}
|
|
}
|
|
prometheus.scrape "zookeeper_gce" {
|
|
targets = discovery.relabel.zookeeper_gce.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "zookeeper_gce"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// redis memcahe
|
|
discovery.gce "redis_memcache_a" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-a"
|
|
port = 9121
|
|
}
|
|
discovery.gce "redis_memcache_b" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-b"
|
|
port = 9121
|
|
}
|
|
discovery.gce "redis_memcache_c" {
|
|
project = "meesho-datascience-prd-0622"
|
|
zone = "asia-southeast1-c"
|
|
port = 9121
|
|
}
|
|
discovery.relabel "redis_memcache" {
|
|
targets = concat(
|
|
discovery.gce.redis_memcache_a.targets,
|
|
discovery.gce.redis_memcache_b.targets,
|
|
discovery.gce.redis_memcache_c.targets,
|
|
)
|
|
|
|
rule {
|
|
source_labels = ["__meta_gce_label_monitoring"]
|
|
regex = "no"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
regex = "(DR|perf)"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_pod"]
|
|
regex = "(devops)"
|
|
action = "drop"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_mcache"]
|
|
regex = "(redis-server)"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_name"]
|
|
target_label = "name"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_mcache"]
|
|
target_label = "mcache"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_bu"]
|
|
target_label = "bu"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_team"]
|
|
target_label = "team"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_gce_label_env"]
|
|
target_label = "env"
|
|
}
|
|
}
|
|
prometheus.scrape "redis_memcache" {
|
|
targets = discovery.relabel.redis_memcache.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "redis_memcache"
|
|
scrape_timeout = "5s"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
// model inference service
|
|
discovery.relabel "model_inference_service_kubernetes_pods" {
|
|
targets = discovery.kubernetes.kubernetes_pods.targets
|
|
|
|
rule {
|
|
source_labels = ["__address__"]
|
|
regex = "(.*8002)"
|
|
action = "keep"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_node_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
regex = "__meta_kubernetes_pod_label_(.+)"
|
|
action = "labelmap"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_namespace"]
|
|
target_label = "kubernetes_namespace"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_node_name"]
|
|
target_label = "kubernetes_node"
|
|
}
|
|
rule {
|
|
source_labels = ["__meta_kubernetes_pod_name"]
|
|
target_label = "kubernetes_pod_name"
|
|
}
|
|
rule {
|
|
target_label = "cluster"
|
|
replacement = "k8s-datascience-prd-ase1"
|
|
}
|
|
}
|
|
prometheus.scrape "model_inference_service_kubernetes_pods" {
|
|
targets = discovery.relabel.model_inference_service_kubernetes_pods.output
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
job_name = "model_inference_service_kubernetes_pods"
|
|
clustering {
|
|
enabled = true
|
|
}
|
|
}
|
|
|
|
prometheus.exporter.self "default" {}
|
|
prometheus.scrape "alloy_pods" {
|
|
targets = prometheus.exporter.self.default.targets
|
|
forward_to = [prometheus.remote_write.mimir.receiver]
|
|
}
|
|
|
|
// writers
|
|
prometheus.remote_write "mimir" {
|
|
endpoint {
|
|
url = "http://mimir-nginx.mimir-distributed.svc.clusterset.local/api/v1/push"
|
|
headers = {"X-Scope-OrgID" = "anonymous"}
|
|
tls_config {
|
|
insecure_skip_verify = true
|
|
}
|
|
queue_config {
|
|
min_shards = 25
|
|
}
|
|
}
|
|
} |