Revert "feat(monitor): optimize monitor and alert (#2132)" (#2146)

This reverts commit 1aad7f4.
tkestack · Nov 3, 2022 · fcf0558 · fcf0558
1 parent c7f9641
commit fcf0558
Show file tree

Hide file tree

Showing 12 changed files with 215 additions and 328 deletions.
diff --git a/api/notify/types.go b/api/notify/types.go
@@ -405,12 +405,10 @@ type MessageList struct {
 
 // MessageSpec is a description of a message.
 type MessageSpec struct {
-	TenantID     string
-	ReceiverName string
-	// +optional
+	TenantID        string
+	ReceiverName    string
 	ReceiverChannel ReceiverChannel
-	// +optional
-	Identity string
+	Identity        string
 	// +optional
 	Username string
 	// +optional
@@ -436,10 +434,6 @@ type MessageStatus struct {
 	// The last time the condition transitioned from one status to another.
 	// +optional
 	LastTransitionTime metav1.Time
-
-	// A human readable message indicating details about the transition.
-	// +optional
-	FailedReason string
 }
 
 // MessagePhase indicates the status of message.

diff --git a/api/notify/v1/generated.pb.go b/api/notify/v1/generated.pb.go
diff --git a/api/notify/v1/generated.proto b/api/notify/v1/generated.proto
diff --git a/api/notify/v1/types.go b/api/notify/v1/types.go
@@ -405,12 +405,10 @@ type MessageList struct {
 
 // MessageSpec is a description of a message.
 type MessageSpec struct {
-	TenantID     string `json:"tenantID" protobuf:"bytes,1,opt,name=tenantID"`
-	ReceiverName string `json:"receiverName" protobuf:"bytes,2,opt,name=receiverName"`
-	// +optional
-	ReceiverChannel ReceiverChannel `json:"receiverChannel,omitempty" protobuf:"bytes,3,opt,name=receiverChannel,casttype=ReceiverChannel"`
-	// +optional
-	Identity string `json:"identity,omitempty" protobuf:"bytes,4,opt,name=identity"`
+	TenantID        string          `json:"tenantID" protobuf:"bytes,1,opt,name=tenantID"`
+	ReceiverName    string          `json:"receiverName" protobuf:"bytes,2,opt,name=receiverName"`
+	ReceiverChannel ReceiverChannel `json:"receiverChannel" protobuf:"bytes,3,opt,name=receiverChannel,casttype=ReceiverChannel"`
+	Identity        string          `json:"identity" protobuf:"bytes,4,opt,name=identity"`
 	// +optional
 	Username string `json:"username,omitempty" protobuf:"bytes,5,opt,name=username"`
 	// +optional
@@ -436,10 +434,6 @@ type MessageStatus struct {
 	// The last time the condition transitioned from one status to another.
 	// +optional
 	LastTransitionTime metav1.Time `json:"lastTransitionTime,omitempty" protobuf:"bytes,2,opt,name=lastTransitionTime"`
-
-	// A human readable message indicating details about the transition.
-	// +optional
-	FailedReason string `json:"failedReason,omitempty" protobuf:"bytes,3,opts,name=failedReason"`
 }
 
 // MessagePhase indicates the status of message.

diff --git a/api/notify/v1/types_swagger_doc_generated.go b/api/notify/v1/types_swagger_doc_generated.go
diff --git a/api/notify/v1/zz_generated.conversion.go b/api/notify/v1/zz_generated.conversion.go
diff --git a/api/openapi/zz_generated.openapi.go b/api/openapi/zz_generated.openapi.go
diff --git a/cmd/tke-installer/app/installer/manifests/tke-notify-api/tke-notify-api.yaml b/cmd/tke-installer/app/installer/manifests/tke-notify-api/tke-notify-api.yaml
@@ -70,8 +70,8 @@ spec:
             periodSeconds: 20
           resources:
             limits:
-              cpu: 500m
-              memory: 2Gi
+              cpu: 100m
+              memory: 256Mi
             requests:
               cpu: 50m
               memory: 128Mi
@@ -141,10 +141,6 @@ data:
       api_server = "https://tke-platform-api"
       api_server_client_config = "/app/conf/tke-platform-config.yaml"
 
-    [features]
-       # Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h"
-       message_ttl: "360h"
-
   tke-auth-webhook.yaml: |
     apiVersion: v1
     kind: Config

diff --git a/cmd/tke-notify-api/app/options/feature.go b/cmd/tke-notify-api/app/options/feature.go
@@ -19,10 +19,9 @@
 package options
 
 import (
-	"time"
-
 	"github.com/spf13/pflag"
 	"github.com/spf13/viper"
+	"time"
 )
 
 const (
@@ -48,7 +47,6 @@ func (o *FeatureOptions) AddFlags(fs *pflag.FlagSet) {
 func (o *FeatureOptions) ApplyFlags() []error {
 	var errs []error
 
-	// Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h".
 	o.MessageTTL = viper.GetDuration(configMessageTTL)
 
 	return errs

diff --git a/pkg/monitor/controller/prometheus/yamls.go b/pkg/monitor/controller/prometheus/yamls.go
@@ -668,7 +668,7 @@ groups:
     expr: sum(k8s_container_fs_write_times) without (container_name,container_id)
 
   - record: k8s_pod_status_ready
-    expr: sum(kube_pod_status_phase{phase=~"Running|Succeeded"}) by (namespace,pod_name) *  on(namespace, pod_name) group_left(workload_kind,workload_name,node, node_role)  __pod_info2
+    expr: sum(kube_pod_status_ready{condition="true"}) by (namespace,pod_name) *  on(namespace, pod_name) group_left(workload_kind,workload_name,node, node_role)  __pod_info2
 
   - record: k8s_pod_restart_total
     expr: sum(idelta(kube_pod_container_status_restarts_total [2m])) by (namespace,pod_name) *  on(namespace, pod_name) group_left(workload_kind,workload_name,node, node_role)  __pod_info2
@@ -1240,8 +1240,8 @@ func configForAlertManager(webhookAddr string, repeatInterval string) string {
 
     route:
       group_by: ['alertname','alarmPolicyName','version']
-      group_wait: 30s
-      group_interval: 5m
+      group_wait: 1s
+      group_interval: 1s
       repeat_interval: %s
       receiver: 'web.hook'
       routes: