Merge pull request #33168 from Crassirostris/fix-es-logging-test

Automatic merge from submit-queue

Fix elasticsearch cluster logging e2e test on GCI

Fix #33126
Fix #32804

Additional timeouts are proven unnecessary, the problem lies with the cpu limits

cluster/addons/fluentd-elasticsearch/es-controller.yaml

@@ -23,9 +23,9 @@ spec:
       - image: gcr.io/google_containers/elasticsearch:1.9
         name: elasticsearch-logging         
         resources:
-          # keep request = limit to keep this container in guaranteed class
+          # need more cpu upon initialization, therefore burstable class
           limits:
-            cpu: 100m
+            cpu: 1000m
           requests:
             cpu: 100m
         ports:

test/e2e/es_cluster_logging.go

@@ -219,11 +219,6 @@ func ClusterLevelLoggingWithElasticsearch(f *framework.Framework) {
 	}
 	framework.Logf("Found %d healthy nodes.", len(nodes.Items))
 
-	// TODO: Figure out why initialization time influences
-	// results of the test and remove this step
-	By("Wait some more time to ensure ES and fluentd are ready")
-	time.Sleep(2 * time.Minute)
-
 	// Wait for the Fluentd pods to enter the running state.
 	By("Checking to make sure the Fluentd pod are running on each healthy node")
 	label = labels.SelectorFromSet(labels.Set(map[string]string{k8sAppKey: fluentdValue}))
@@ -306,9 +301,6 @@ func ClusterLevelLoggingWithElasticsearch(f *framework.Framework) {
 		Expect(err).NotTo(HaveOccurred())
 	}
 
-	// Wait a bit for the log information to make it into Elasticsearch.
-	time.Sleep(30 * time.Second)
-
 	// Make several attempts to observe the logs ingested into Elasticsearch.
 	By("Checking all the log lines were ingested into Elasticsearch")
 	totalMissing := 0