improve koordlet log verbosity (#338)

Signed-off-by: saintube <saintube@foxmail.com>

Author: saintube

pkg/koordlet/metricsadvisor/collector.go

@@ -362,7 +362,7 @@ func (c *collector) collectPodThrottledInfo() {
 		if err != nil || currentCPUStat == nil {
 			if pod.Status.Phase == corev1.PodRunning {
 				// print running pod collection error
-				klog.Infof("collect pod %s/%s, uid %v cpu throttled failed, err %v, metric %v",
+				klog.V(4).Infof("collect pod %s/%s, uid %v cpu throttled failed, err %v, metric %v",
 					pod.Namespace, pod.Name, uid, err, currentCPUStat)
 			}
 			continue
@@ -402,7 +402,7 @@ func (c *collector) collectContainerThrottledInfo(podMeta *statesinformer.PodMet
 		collectTime := time.Now()
 		containerStat := &pod.Status.ContainerStatuses[i]
 		if len(containerStat.ContainerID) == 0 {
-			klog.Infof("container %s/%s/%s id is empty, maybe not ready, skip this round",
+			klog.V(4).Infof("container %s/%s/%s id is empty, maybe not ready, skip this round",
 				pod.Namespace, pod.Name, containerStat.Name)
 			continue
 		}
@@ -414,7 +414,7 @@ func (c *collector) collectContainerThrottledInfo(podMeta *statesinformer.PodMet
 		}
 		currentCPUStat, err := system.GetCPUStatRaw(containerCgroupPath)
 		if err != nil {
-			klog.Infof("collect container %s/%s/%s cpu throttled failed, err %v, metric %v",
+			klog.V(4).Infof("collect container %s/%s/%s cpu throttled failed, err %v, metric %v",
 				pod.Namespace, pod.Name, containerStat.Name, err, currentCPUStat)
 			continue
 		}

pkg/koordlet/reporter/reporter.go

@@ -211,7 +211,7 @@ func (r *reporter) sync() {
 	if retErr != nil {
 		klog.Warningf("update node metric status failed, status %v, err %v", util.DumpJSON(newStatus), retErr)
 	} else {
-		klog.Infof("update node metric status success, detail: %v", util.DumpJSON(newStatus))
+		klog.V(4).Infof("update node metric status success, detail: %v", util.DumpJSON(newStatus))
 	}
 }
 

pkg/koordlet/resmanager/cgroup_reconcile.go

@@ -467,6 +467,8 @@ func makeCgroupResourcesForAnolis(owner *OwnerRef, parentDir string, summary *cg
 	var resources []MergeableResourceUpdater
 
 	if !system.HostSystemInfo.IsAnolisOS {
+		klog.V(5).Infof("ignored cgroup resources which required non Anolis OS, owner: %v, parentDir: %v",
+			owner, parentDir)
 		return nil
 	}
 

pkg/koordlet/resmanager/cpu_burst.go

@@ -509,12 +509,16 @@ func (b *CPUBurst) applyCPUBurst(burstCfg *slov1alpha1.CPUBurstConfig, podMeta *
 			containerCFSBurstValStr := strconv.FormatInt(containerCFSBurstVal, 10)
 			updater := NewCommonCgroupResourceUpdater(ownerRef, containerDir, system.CPUBurst, containerCFSBurstValStr)
 			updated, err := b.executor.UpdateByCache(updater)
-			if err != nil {
+			if err == nil {
+				klog.V(5).Infof("apply container %v/%v/%v cpu burst value success, dir %v, value %v",
+					pod.Namespace, pod.Name, containerStat.Name, containerDir, containerCFSBurstVal)
+			} else if system.HostSystemInfo.IsAnolisOS {
+				// cgroup `cpu.burst_us` is expected available on anolis os, and it may not exist in other kernels.
 				klog.Infof("update container %v/%v/%v cpu burst failed, dir %v, updated %v, error %v",
 					pod.Namespace, pod.Name, containerStat.Name, containerDir, updated, err)
 			} else {
-				klog.V(5).Infof("apply container %v/%v/%v cpu burst value success, dir %v, value %v",
-					pod.Namespace, pod.Name, containerStat.Name, containerDir, containerCFSBurstVal)
+				klog.V(4).Infof("update container %v/%v/%v cpu burst ignored on non Anolis OS, dir %v, "+
+					"updated %v, info %v", pod.Namespace, pod.Name, containerStat.Name, containerDir, updated, err)
 			}
 		}
 	} // end for containers
@@ -525,12 +529,16 @@ func (b *CPUBurst) applyCPUBurst(burstCfg *slov1alpha1.CPUBurstConfig, podMeta *
 		podCFSBurstValStr := strconv.FormatInt(podCFSBurstVal, 10)
 		updater := NewCommonCgroupResourceUpdater(ownerRef, podDir, system.CPUBurst, podCFSBurstValStr)
 		updated, err := b.executor.UpdateByCache(updater)
-		if err != nil {
+		if err == nil {
+			klog.V(5).Infof("apply pod %v/%v cpu burst value success, dir %v, value %v",
+				pod.Namespace, pod.Name, podDir, podCFSBurstValStr)
+		} else if system.HostSystemInfo.IsAnolisOS {
+			// cgroup `cpu.burst_us` is expected available on anolis os, and it may not exist in other kernels.
 			klog.Infof("update pod %v/%v cpu burst failed, dir %v, updated %v, error %v",
 				pod.Namespace, pod.Name, podDir, updated, err)
 		} else {
-			klog.V(5).Infof("apply pod %v/%v cpu burst value success, dir %v, value %v",
-				pod.Namespace, pod.Name, podDir, podCFSBurstValStr)
+			klog.V(4).Infof("update pod %v/%v cpu burst ignored on non Anolis OS, dir %v, updated %v, "+
+				"info %v", pod.Namespace, pod.Name, podDir, updated, err)
 		}
 	}
 }

pkg/koordlet/resmanager/memory_evict.go

@@ -51,11 +51,11 @@ func NewMemoryEvictor(mgr *resmanager) *MemoryEvictor {
 }
 
 func (m *MemoryEvictor) memoryEvict() {
-	klog.Infof("starting memory evict process")
-	defer klog.Infof("memory evict process completed")
+	klog.V(5).Infof("starting memory evict process")
+	defer klog.V(5).Infof("memory evict process completed")
 
 	if time.Now().Before(m.lastEvictTime.Add(time.Duration(m.resManager.config.MemoryEvictCoolTimeSeconds) * time.Second)) {
-		klog.Infof("skip memory evict process, still in evict cooling time")
+		klog.V(5).Infof("skip memory evict process, still in evict cooling time")
 		return
 	}
 
@@ -110,7 +110,7 @@ func (m *MemoryEvictor) memoryEvict() {
 
 	nodeMemoryUsage := nodeMetric.MemoryUsed.MemoryWithoutCache.Value() * 100 / memoryCapacity
 	if nodeMemoryUsage < *thresholdPercent {
-		klog.Infof("skip memory evict, node memory usage(%v) is below threshold(%v)", nodeMemoryUsage, thresholdConfig)
+		klog.V(5).Infof("skip memory evict, node memory usage(%v) is below threshold(%v)", nodeMemoryUsage, thresholdConfig)
 		return
 	}