hoststatus: add lasted warning log (#35)

* hoststatus: add lasted warning log Signed-off-by: weizhoublue <[email protected]> * d Signed-off-by: weizhoublue <[email protected]> --------- Signed-off-by: weizhoublue <[email protected]>
spidernet-io · Jan 20, 2025 · 1cee628 · 1cee628
1 parent be83a70
commit 1cee628
Show file tree

Hide file tree

Showing 10 changed files with 50 additions and 9 deletions.
diff --git a/VERSION b/VERSION
@@ -1 +1 @@
-v0.4.0
+v0.5.0
diff --git a/chart/Chart.yaml b/chart/Chart.yaml
@@ -3,10 +3,10 @@ name: bmc-operator
 description: A Helm chart for BMC Operator
 
 # This is the chart version, which will be taken from VERSION file
-version: 0.4.0
+version: 0.5.0
 
 # This is the version number of the application being deployed, which will be taken from VERSION file
-appVersion: "0.4.0"
+appVersion: "0.5.0"
 
 type: application
 

diff --git a/chart/crds/bmc.spidernet.io_hoststatuses.yaml b/chart/crds/bmc.spidernet.io_hoststatuses.yaml
@@ -104,6 +104,16 @@ spec:
                     - message
                     - time
                     type: object
+                  lastestWarningLog:
+                    properties:
+                      message:
+                        type: string
+                      time:
+                        type: string
+                    required:
+                    - message
+                    - time
+                    type: object
                   totalLogAccount:
                     format: int32
                     type: integer

diff --git a/doc/usage/dhcp.md b/doc/usage/dhcp.md
@@ -4,6 +4,10 @@
 
 Agent 中的 DHCP server，支持把 DHCP client 的 IP 固定到 DHCP server 的配置中， 从而实现 DHCP client 的 IP 固定。
 
+### DHCP server 的配置
+
+DHCP server 的配置文件，位于 confimap bmc-dhcp-config 中，若有修改需求，设置后，再重启 agent pod
+
 ### DHCP client 的 IP 固定
 
 当`EnableDhcpDiscovery`功能开启时：

diff --git a/doc/usage/quickstart.md b/doc/usage/quickstart.md
@@ -314,6 +314,10 @@ test-hostendpoint                bmc-clusteragent   true      192.168.0.50    ho
 ```bash
 # 获取所有 BMC 主机的日志
 kubectl get events -n bmc --field-selector reason=BMCLogEntry
+    LAST SEEN   TYPE      REASON        OBJECT                                      MESSAGE
+    30s         Warning   BMCLogEntry   hoststatus/bmc-clusteragent-192-168-0-100   [2012-03-07T14:45:00Z][Critical]:  Temperature threshold exceeded
+    2m13s       Warning   BMCLogEntry   hoststatus/bmc-clusteragent-192-168-0-101   [2012-03-07T14:45:00Z][Critical]:  Temperature threshold exceeded
+    105s        Normal    BMCLogEntry   hoststatus/device-safe                      [2018-08-31T13:33:54+00:00][]:  [ PS1 Status ] Power Supply Failure
 
 # 获取指定 BMC 主机的日志
 kubectl get events -n bmc --field-selector reason=BMCLogEntry,involvedObject.name=${HoststatusName}
@@ -322,11 +326,15 @@ kubectl get events -n bmc --field-selector reason=BMCLogEntry,involvedObject.nam
 kubectl get hoststatus ${HoststatusName} -n bmc -o jsonpath='{.status.log}' | jq .
   {
     "lastestLog": {
-      "message": "[][2018-08-31T13:33:54+00:00][]: [ PS1 Status ] Power Supply Failure",
-      "time": "2018-08-31T13:33:54+00:00"
+      "message": "[2024-10-16T22:47:28Z][Critical]:  [GS-0002] GPU Temp, 6 is not present",
+      "time": "2024-10-16T22:47:28Z"
     },
-    "totalLogAccount": 67,
-    "warningLogAccount": 67
+    "lastestWarningLog": {
+      "message": "[2024-10-16T22:47:28Z][Critical]:  [GS-0002] GPU Temp, 6 is not present",
+      "time": "2024-10-16T22:47:28Z"
+    },
+    "totalLogAccount": 52,
+    "warningLogAccount": 35
   }
 
 ```
diff --git a/pkg/agent/hostendpoint/controller.go b/pkg/agent/hostendpoint/controller.go
@@ -180,6 +180,7 @@ func (r *HostEndpointReconciler) handleHostEndpoint(ctx context.Context, hostEnd
 			TotalLogAccount:   0,
 			WarningLogAccount: 0,
 			LastestLog:        nil,
+			LastestWarningLog: nil,
 		},
 	}
 

diff --git a/pkg/agent/hoststatus/HostStatusReconcile.go b/pkg/agent/hoststatus/HostStatusReconcile.go
@@ -29,12 +29,14 @@ var hostStatusLock = &sync.Mutex{}
 // ------------------------------  update the spec.info of the hoststatus
 
 // GenerateEvents creates Kubernetes events from Redfish log entries and returns the latest message and count
-func (c *hostStatusController) GenerateEvents(logEntrys []*gofishredfish.LogEntry, hostStatusName string, lastLogTime string) (newLastestTime, newLastestMsg string, totalMsgCount, warningMsgCount, newLogAccount int) {
+func (c *hostStatusController) GenerateEvents(logEntrys []*gofishredfish.LogEntry, hostStatusName string, lastLogTime string) (newLastestTime, newLastestMsg, newLastestWarningTime, newLastestWarningMsg string, totalMsgCount, warningMsgCount, newLogAccount int) {
 	totalMsgCount = 0
 	warningMsgCount = 0
 	newLogAccount = 0
 	newLastestTime = ""
 	newLastestMsg = ""
+	newLastestWarningTime = ""
+	newLastestWarningMsg = ""
 
 	if len(logEntrys) == 0 {
 		return
@@ -52,6 +54,10 @@ func (c *hostStatusController) GenerateEvents(logEntrys []*gofishredfish.LogEntr
 		ty := corev1.EventTypeNormal
 		if entry.Severity != gofishredfish.OKEventSeverity && entry.Severity != "" {
 			ty = corev1.EventTypeWarning
+			if newLastestWarningMsg == "" {
+				newLastestWarningTime = entry.Created
+				newLastestWarningMsg = msg
+			}
 			warningMsgCount++
 		}
 
@@ -145,14 +151,18 @@ func (c *hostStatusController) UpdateHostStatusInfo(name string, d *hoststatusda
 			if updated.Status.Log.LastestLog != nil {
 				lastLogTime = updated.Status.Log.LastestLog.Time
 			}
-			newLastestTime, newLastestMsg, totalMsgCount, warningMsgCount, newLogAccount := c.GenerateEvents(logEntrys, name, lastLogTime)
+			newLastestTime, newLastestMsg, newLastestWarningTime, newLastestWarningMsg, totalMsgCount, warningMsgCount, newLogAccount := c.GenerateEvents(logEntrys, name, lastLogTime)
 			if newLastestTime != "" {
 				updated.Status.Log.TotalLogAccount = int32(totalMsgCount)
 				updated.Status.Log.WarningLogAccount = int32(warningMsgCount)
 				updated.Status.Log.LastestLog = &bmcv1beta1.LogEntry{
 					Time:    newLastestTime,
 					Message: newLastestMsg,
 				}
+				updated.Status.Log.LastestWarningLog = &bmcv1beta1.LogEntry{
+					Time:    newLastestWarningTime,
+					Message: newLastestWarningMsg,
+				}
 				log.Logger.Infof("find %d new logs for hostStatus %s", newLogAccount, name)
 			}
 		}

diff --git a/pkg/agent/hoststatus/dhcp.go b/pkg/agent/hoststatus/dhcp.go
@@ -152,6 +152,7 @@ func (c *hostStatusController) handleDHCPAdd(client dhcptypes.ClientInfo) error
 			TotalLogAccount:   0,
 			WarningLogAccount: 0,
 			LastestLog:        nil,
+			LastestWarningLog: nil,
 		},
 	}
 	if c.config.AgentObjSpec.Endpoint.SecretName != "" {

diff --git a/pkg/k8s/apis/bmc.spidernet.io/v1beta1/hoststatus_types.go b/pkg/k8s/apis/bmc.spidernet.io/v1beta1/hoststatus_types.go
@@ -43,6 +43,8 @@ type LogStruct struct {
 	WarningLogAccount int32 `json:"warningLogAccount"`
 	// +optional
 	LastestLog *LogEntry `json:"lastestLog,omitempty"`
+	// +optional
+	LastestWarningLog *LogEntry `json:"lastestWarningLog,omitempty"`
 }
 
 type LogEntry struct {

diff --git a/pkg/k8s/apis/bmc.spidernet.io/v1beta1/zz_generated.deepcopy.go b/pkg/k8s/apis/bmc.spidernet.io/v1beta1/zz_generated.deepcopy.go