chore: fix database monitor. (#4440)

* chore: add monitor for database and kubernetes. Signed-off-by: yy <lingdie.yy@outlook.com> Signed-off-by: yy <lingdie.yy@outlook.com> * chore: add monitor for database and kubernetes. Signed-off-by: yy <lingdie.yy@outlook.com> Signed-off-by: yy <lingdie.yy@outlook.com> * chore: fix pull image logic. Signed-off-by: yy <lingdie.yy@outlook.com> Signed-off-by: yy <lingdie.yy@outlook.com> * fix offline build * chore: fix database monitor. Signed-off-by: yy <lingdie.yy@outlook.com> Signed-off-by: yy <lingdie.yy@outlook.com> * chore: fix kb prometheus. Signed-off-by: yy <lingdie.yy@outlook.com> Signed-off-by: yy <lingdie.yy@outlook.com> --------- Signed-off-by: yy <lingdie.yy@outlook.com>
2026-09-01 15:38:06 +08:00 · 2023-12-21 15:51:59 +08:00
parent 18aff5d84a
commit c0660dbc59
4 changed files with 69 additions and 34 deletions
@@ -8,7 +8,7 @@ COPY manifests manifests
 ENV cloudDomain="127.0.0.1.nip.io"
 ENV cloudPort=""
 ENV certSecretName="wildcard-cert"
-ENV monitorUrl=""
+ENV monitorUrl="http://database-monitor.sealos.svc.cluster.local:9090"
 ENV minioUrl=""
 ENV minioAccessKey=""
 ENV minioSecretKey=""
@@ -14,7 +14,7 @@ images=(
  docker.io/labring/cilium:v1.12.14
  docker.io/labring/cert-manager:v1.8.0
  docker.io/labring/openebs:v3.4.0
-  docker.io/labring/kubernetes-reflector:v7.0.151
+  docker.io/labring/kube-prometheus-stack:v0.63.0
  docker.io/labring/ingress-nginx:v1.5.1
  docker.io/labring/kubeblocks:v0.6.4
  docker.io/labring/metrics-server:v0.6.4
@@ -32,7 +32,7 @@ done
 # get and save cli
 mkdir -p output/cli

-VERSION="v4.3.5"
+VERSION="v4.3.7"

 wget https://github.com/labring/sealos/releases/download/${VERSION}/sealos_${VERSION#v}_linux_${ARCH}.tar.gz \
   && tar zxvf sealos_${VERSION#v}_linux_${ARCH}.tar.gz sealos && chmod +x sealos && mv sealos output/cli
@@ -385,16 +385,76 @@ data:
    rule_files:
    - /etc/config/recording_rules.yml
    - /etc/config/alerting_rules.yml
+    - /etc/config/kubelet_alert_rules.yml
    - /etc/config/mysql_alert_rules.yml
    - /etc/config/postgresql_alert_rules.yml
    - /etc/config/redis_alert_rules.yml
    - /etc/config/kafka_alert_rules.yml
-    - /etc/config/mongodb_replicaset_alert_rules.yaml
+    - /etc/config/mongodb_alert_rules.yml
    scrape_configs:
    - job_name: prometheus
      static_configs:
      - targets:
        - localhost:9090
+    - honor_labels: true
+      job_name: kubeblocks-service
+      kubernetes_sd_configs:
+      - role: endpoints
+      relabel_configs:
+      - action: keep
+        regex: kubeblocks
+        source_labels:
+        - __meta_kubernetes_service_label_app_kubernetes_io_managed_by
+      - action: drop
+        regex: agamotto
+        source_labels:
+        - __meta_kubernetes_service_label_monitor_kubeblocks_io_managed_by
+      - action: keep
+        regex: true
+        source_labels:
+        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_scrape
+      - action: replace
+        regex: (https?)
+        source_labels:
+        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_scheme
+        target_label: __scheme__
+      - action: replace
+        regex: (.+)
+        source_labels:
+        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_path
+        target_label: __metrics_path__
+      - action: replace
+        regex: (.+?)(?::\d+)?;(\d+)
+        replacement: $1:$2
+        source_labels:
+        - __address__
+        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_port
+        target_label: __address__
+      - action: labelmap
+        regex: __meta_kubernetes_service_annotation_monitor_kubeblocks_io_param_(.+)
+        replacement: __param_$1
+      - action: labelmap
+        regex: __meta_kubernetes_service_label_(.+)
+      - action: replace
+        source_labels:
+        - __meta_kubernetes_namespace
+        target_label: namespace
+      - action: replace
+        source_labels:
+        - __meta_kubernetes_service_name
+        target_label: service
+      - action: replace
+        source_labels:
+        - __meta_kubernetes_pod_node_name
+        target_label: node
+      - action: replace
+        source_labels:
+        - __meta_kubernetes_pod_name
+        target_label: pod
+      - action: drop
+        regex: Pending|Succeeded|Failed|Completed
+        source_labels:
+        - __meta_kubernetes_pod_phase
    - bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
      job_name: kubernetes-apiservers
      kubernetes_sd_configs:
@@ -447,22 +507,18 @@ data:
        ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
        insecure_skip_verify: true
    - honor_labels: true
-      job_name: kubeblocks-service-endpoints
+      job_name: kubeblocks-agamotto
      kubernetes_sd_configs:
      - role: endpoints
      relabel_configs:
      - action: keep
-        regex: kubeblocks
+        regex: agamotto
        source_labels:
-        - __meta_kubernetes_service_label_app_kubernetes_io_managed_by
+        - __meta_kubernetes_service_label_monitor_kubeblocks_io_managed_by
      - action: keep
        regex: true
        source_labels:
        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_scrape
-      - action: drop
-        regex: true
-        source_labels:
-        - __meta_kubernetes_service_annotation_monitor_kubeblocks_io_scrape_slow
      - action: replace
        regex: (https?)
        source_labels:
@@ -483,30 +539,10 @@ data:
      - action: labelmap
        regex: __meta_kubernetes_service_annotation_monitor_kubeblocks_io_param_(.+)
        replacement: __param_$1
-      - action: labelmap
-        regex: __meta_kubernetes_service_label_(.+)
-      - action: replace
-        source_labels:
-        - __meta_kubernetes_namespace
-        target_label: namespace
-      - action: replace
-        source_labels:
-        - __meta_kubernetes_service_name
-        target_label: service
-      - action: replace
-        source_labels:
-        - __meta_kubernetes_pod_node_name
-        target_label: node
-      - action: replace
-        source_labels:
-        - __meta_kubernetes_pod_name
-        target_label: pod
      - action: drop
        regex: Pending|Succeeded|Failed|Completed
        source_labels:
        - __meta_kubernetes_pod_phase
-      scrape_interval: 5m
-      scrape_timeout: 30s
    alerting:
      alertmanagers:
      - kubernetes_sd_configs:
@@ -531,7 +567,7 @@ data:
          regex: "9093"
          action: keep
 '
-    echo $kb_addon_prometheus_server_patch > $CLOUD_DIR/kb-addon-prometheus-server-patch.yaml
+    echo "$kb_addon_prometheus_server_patch" > $CLOUD_DIR/kb-addon-prometheus-server-patch.yaml


    sealos_gen_cmd="sealos gen ${image_registry}/${image_repository}/kubernetes:v${kubernetes_version#v:-1.25.6}\
@@ -36,9 +36,8 @@ spec:
        - /manager
        env:
        - name: PROMETHEUS_SERVICE_HOST
-          value: http://kb-addon-prometheus-server.
+          value: http://kb-addon-prometheus-server.kb-system.svc.cluster.local
        - name: OBJECT_STORAGE_INSTANCE
-          value:
        image: ghcr.io/labring/sealos-database-service:latest
        imagePullPolicy: Always
        name: database-monitor