Merge pull request #37236 from bowei/flake-kube-proxy

Automatic merge from submit-queue Update the timeout in CLOSE_WAIT e2e test I see some test flakes due to the timeout being too strict, updating to a larger value. Adding tail -n 1, it looks like there may be leftover state for other runs. We really only care about one of the CLOSE_WAIT entries.

Merge pull request #37236 from bowei/flake-kube-proxy
4dd0320d · Kubernetes Submit Queue · GitHub · 4ac2ef2a · 9aef074e · 4dd0320d
Commit 4dd0320d authored Nov 21, 2016 by Kubernetes Submit Queue Committed by GitHub Nov 21, 2016
Hide whitespace changes
Inline Side-by-side

Showing with 3 additions and 2 deletions

kube_proxy.go test/e2e/kube_proxy.go +3 -2

No files found.
--- a/test/e2e/kube_proxy.go
+++ b/test/e2e/kube_proxy.go
@@ -180,7 +180,8 @@ var _ = framework.KubeDescribe("Network", func() {
 			fmt.Sprintf(
 				"sudo cat /proc/net/ip_conntrack "+
 					"| grep 'CLOSE_WAIT.*dst=%v.*dport=%v' "+
-					"| awk '{print $3}'",
+					"| tail -n 1"+
+					"| awk '{print $3}' ",
 				serverNodeInfo.nodeIp,
 				testDaemonTcpPort),
 			framework.TestContext.Provider,
@@ -193,7 +194,7 @@ var _ = framework.KubeDescribe("Network", func() {
 		// These must be synchronized from the default values set in
 		// pkg/apis/../defaults.go ConntrackTCPCloseWaitTimeout. The
 		// current defaults are hidden in the initialization code.
-		const epsilonSeconds = 10
+		const epsilonSeconds = 60
 		const expectedTimeoutSeconds = 60 * 60

 		framework.Logf("conntrack entry timeout was: %v, expected: %v",