Bug 1916843: collect logs from openshift-sdn-controller pod

docs/gathered-data.md

@@ -317,6 +317,26 @@ Response see https://docs.openshift.com/container-platform/4.6/rest_api/workload
 Location in archive: config/pod/openshift-apiserver-operator/logs/{pod-name}/errors.log
 
 
+## OpenshiftSDNControllerLogs
+
+collects logs from sdn-controller pod in openshift-sdn namespace with following substrings:
+  - "Node %s is not Ready": A node has been set offline for egress IPs because it is reported not ready at API
+  - "Node %s may be offline... retrying": An egress node has failed the egress IP health check once,
+      so it has big chances to be marked as offline soon or, at the very least, there has been a connectivity glitch.
+  - "Node %s is offline": An egress node has failed enough probes to have been marked offline for egress IPs.
+      If it has egress CIDRs assigned, its egress IPs have been moved to other nodes.
+      Indicates issues at either the node or the network between the master and the node.
+  - "Node %s is back online": This indicates that a node has recovered from the condition described
+      at the previous message, by starting succeeding the egress IP health checks.
+      Useful just in case that previous “Node %s is offline” messages are lost,
+      so that we have a clue that there was failure previously.
+
+The Kubernetes API https://github.com/kubernetes/client-go/blob/master/kubernetes/typed/core/v1/pod_expansion.go#L48
+Response see https://docs.openshift.com/container-platform/4.6/rest_api/workloads_apis/pod-core-v1.html#apiv1namespacesnamespacepodsnamelog
+
+Location in archive: config/pod/openshift-sdn/logs/{pod-name}/errors.log
+
+
 ## OpenshiftSDNLogs
 
 collects logs from pods in openshift-sdn namespace with following substrings:

docs/insights-archive-sample/config/pod/openshift-sdn/logs/sdn-controller-l8gq9/errors.log

@@ -0,0 +1,7 @@
+W0120 20:07:54.645321       1 egressip.go:199] Node ci-ln-0d402jb-f76d1-hts4b-worker-c-2nwdl is not Ready
+W0120 20:08:29.646400       1 egressip.go:199] Node ci-ln-0d402jb-f76d1-hts4b-worker-c-2nwdl is not Ready
+I0120 20:08:59.649184       1 egressip.go:208] Node 10.0.32.3 is back online
+I0120 20:13:49.656538       1 egressip.go:219] Node 10.0.32.3 may be offline... retrying
+I0120 20:13:57.710028       1 egressip.go:219] Node 10.0.32.3 may be offline... retrying
+W0120 20:14:12.749977       1 egressip.go:214] Node 10.0.32.3 is offline
+I0120 20:15:29.663808       1 egressip.go:208] Node 10.0.32.3 is back online

pkg/gather/clusterconfig/0_gatherer.go

@@ -71,6 +71,7 @@ var gatherFunctions = map[string]gatherFunction{
 	"netnamespaces":                     GatherNetNamespace,
 	"openshift_apiserver_operator_logs": GatherOpenShiftAPIServerOperatorLogs,
 	"openshift_sdn_logs":                GatherOpenshiftSDNLogs,
+	"openshift_sdn_controller_logs":     GatherOpenshiftSDNControllerLogs,
 }
 
 // New creates new Gatherer

pkg/gather/clusterconfig/gather_logs.go

@@ -4,6 +4,7 @@ import (
 	"bufio"
 	"context"
 	"fmt"
+	"regexp"
 	"strings"
 
 	corev1 "k8s.io/api/core/v1"
@@ -21,7 +22,9 @@ import (
 //   - limitBytes sets the maximum amount of logs that can be fetched
 //   - logFileName sets the name of the file to save logs to.
 //   - labelSelector allows you to filter pods by their labels
-// Actual location is `config/pod/{namespace}/logs/{podName}/{fileName}.log`
+//   - regexSearch makes messagesToSearch regex patterns, so you can accomplish more complicated search
+//
+// Location of the logs is `config/pod/{namespace}/logs/{podName}/{fileName}.log`
 func gatherLogsFromPodsInNamespace(
 	ctx context.Context,
 	coreClient v1.CoreV1Interface,
@@ -31,6 +34,7 @@ func gatherLogsFromPodsInNamespace(
 	limitBytes int64,
 	logFileName string,
 	labelSelector string,
+	regexSearch bool,
 ) ([]record.Record, error) {
 	pods, err := coreClient.Pods(namespace).List(ctx, metav1.ListOptions{
 		LabelSelector: labelSelector,
@@ -49,7 +53,7 @@ func gatherLogsFromPodsInNamespace(
 				LimitBytes:   &limitBytes,
 			})
 
-			logs, err := filterLogs(ctx, request, messagesToSearch)
+			logs, err := filterLogs(ctx, request, messagesToSearch, regexSearch)
 			if err != nil {
 				return nil, err
 			}
@@ -70,7 +74,9 @@ func gatherLogsFromPodsInNamespace(
 	return records, nil
 }
 
-func filterLogs(ctx context.Context, request *restclient.Request, messagesToSearch []string) (string, error) {
+func filterLogs(
+	ctx context.Context, request *restclient.Request, messagesToSearch []string, regexSearch bool,
+) (string, error) {
 	stream, err := request.Stream(ctx)
 	if err != nil {
 		return "", err
@@ -90,8 +96,18 @@ func filterLogs(ctx context.Context, request *restclient.Request, messagesToSear
 	for scanner.Scan() {
 		line := scanner.Text()
 		for _, messageToSearch := range messagesToSearch {
-			if strings.Contains(strings.ToLower(line), strings.ToLower(messageToSearch)) {
-				result += line + "\n"
+			if regexSearch {
+				matches, err := regexp.MatchString(messageToSearch, line)
+				if err != nil {
+					return "", err
+				}
+				if matches {
+					result += line + "\n"
+				}
+			} else {
+				if strings.Contains(strings.ToLower(line), strings.ToLower(messageToSearch)) {
+					result += line + "\n"
+				}
 			}
 		}
 	}

pkg/gather/clusterconfig/gather_logs_test.go

@@ -11,31 +11,29 @@ import (
 	kubefake "k8s.io/client-go/kubernetes/fake"
 )
 
-func TestGatherLogs(t *testing.T) {
-	const testPodNamespace = "pod-namespace"
+func testGatherLogs(t *testing.T, regexSearch bool, stringToSearch string, shouldExist bool) {
+	const testPodName = "test"
 	const testLogFileName = "errors"
-	// there's no way to specify logs fake pod will have, so we can only search for a hardcoded string "fake logs"
-	const stringToSearch = "fake logs"
 
 	coreClient := kubefake.NewSimpleClientset().CoreV1()
 	ctx := context.Background()
 
-	_, err := coreClient.Pods(testPodNamespace).Create(
+	_, err := coreClient.Pods(testPodName).Create(
 		ctx,
 		&corev1.Pod{
 			ObjectMeta: metav1.ObjectMeta{
-				Name:      testPodNamespace,
-				Namespace: testPodNamespace,
+				Name:      testPodName,
+				Namespace: testPodName,
 			},
 			Status: corev1.PodStatus{
 				Phase: corev1.PodRunning,
 				ContainerStatuses: []corev1.ContainerStatus{
-					{Name: testPodNamespace},
+					{Name: testPodName},
 				},
 			},
 			Spec: corev1.PodSpec{
 				Containers: []corev1.Container{
-					{Name: testPodNamespace},
+					{Name: testPodName},
 				},
 			},
 		},
@@ -48,24 +46,53 @@ func TestGatherLogs(t *testing.T) {
 	records, err := gatherLogsFromPodsInNamespace(
 		ctx,
 		coreClient,
-		testPodNamespace,
+		testPodName,
 		[]string{
 			stringToSearch,
 		},
 		86400,   // last day
 		1024*64, // maximum 64 kb of logs
 		testLogFileName,
 		"",
+		regexSearch,
 	)
 	if err != nil {
 		t.Fatal(err)
 	}
 
+	if !shouldExist {
+		assert.Len(t, records, 0)
+		return
+	}
+
 	assert.Len(t, records, 1)
 	assert.Equal(
 		t,
-		fmt.Sprintf("config/pod/%s/logs/%s/%s.log", testPodNamespace, testPodNamespace, testLogFileName),
+		fmt.Sprintf("config/pod/%s/logs/%s/%s.log", testPodName, testPodName, testLogFileName),
 		records[0].Name,
 	)
-	assert.Equal(t, Raw{stringToSearch + "\n"}, records[0].Item)
+	if regexSearch {
+		assert.Regexp(t, stringToSearch, records[0].Item)
+	} else {
+		assert.Equal(t, Raw{stringToSearch + "\n"}, records[0].Item)
+	}
+}
+
+func TestGatherLogs(t *testing.T) {
+	t.Run("SubstringSearch_ShouldExist", func(t *testing.T) {
+		testGatherLogs(t, false, "fake logs", true)
+	})
+	t.Run("SubstringSearch_ShouldNotExist", func(t *testing.T) {
+		testGatherLogs(t, false, "The quick brown fox jumps over the lazy dog", false)
+	})
+	t.Run("SubstringSearch_ShouldNotExist", func(t *testing.T) {
+		testGatherLogs(t, false, "f.*l", false)
+	})
+
+	t.Run("RegexSearch_ShouldExist", func(t *testing.T) {
+		testGatherLogs(t, true, "f.*l", true)
+	})
+	t.Run("RegexSearch_ShouldNotExist", func(t *testing.T) {
+		testGatherLogs(t, true, "[0-9]99", false)
+	})
 }

pkg/gather/clusterconfig/openshift_apiserver_operator_logs.go

@@ -36,6 +36,7 @@ func GatherOpenShiftAPIServerOperatorLogs(g *Gatherer, c chan<- gatherResult) {
 		1024*64, // maximum 64 kb of logs
 		"errors",
 		"app=openshift-apiserver-operator",
+		false,
 	)
 	if err != nil {
 		c <- gatherResult{nil, []error{err}}

pkg/gather/clusterconfig/openshift_sdn_controller_logs.go

@@ -0,0 +1,57 @@
+package clusterconfig
+
+import (
+	"k8s.io/client-go/kubernetes"
+)
+
+// GatherOpenshiftSDNControllerLogs collects logs from sdn-controller pod in openshift-sdn namespace with following substrings:
+//   - "Node %s is not Ready": A node has been set offline for egress IPs because it is reported not ready at API
+//   - "Node %s may be offline... retrying": An egress node has failed the egress IP health check once,
+//       so it has big chances to be marked as offline soon or, at the very least, there has been a connectivity glitch.
+//   - "Node %s is offline": An egress node has failed enough probes to have been marked offline for egress IPs.
+//       If it has egress CIDRs assigned, its egress IPs have been moved to other nodes.
+//       Indicates issues at either the node or the network between the master and the node.
+//   - "Node %s is back online": This indicates that a node has recovered from the condition described
+//       at the previous message, by starting succeeding the egress IP health checks.
+//       Useful just in case that previous “Node %s is offline” messages are lost,
+//       so that we have a clue that there was failure previously.
+//
+// The Kubernetes API https://github.com/kubernetes/client-go/blob/master/kubernetes/typed/core/v1/pod_expansion.go#L48
+// Response see https://docs.openshift.com/container-platform/4.6/rest_api/workloads_apis/pod-core-v1.html#apiv1namespacesnamespacepodsnamelog
+//
+// Location in archive: config/pod/openshift-sdn/logs/{pod-name}/errors.log
+func GatherOpenshiftSDNControllerLogs(g *Gatherer, c chan<- gatherResult) {
+	messagesToSearch := []string{
+		"Node.+is not Ready",
+		"Node.+may be offline\\.\\.\\. retrying",
+		"Node.+is offline",
+		"Node.+is back online",
+	}
+
+	gatherKubeClient, err := kubernetes.NewForConfig(g.gatherProtoKubeConfig)
+	if err != nil {
+		c <- gatherResult{nil, []error{err}}
+		return
+	}
+
+	coreClient := gatherKubeClient.CoreV1()
+
+	records, err := gatherLogsFromPodsInNamespace(
+		g.ctx,
+		coreClient,
+		"openshift-sdn",
+		messagesToSearch,
+		86400,   // last day
+		1024*64, // maximum 64 kb of logs
+		"errors",
+		"app=sdn-controller",
+		true,
+	)
+	if err != nil {
+		c <- gatherResult{nil, []error{err}}
+		return
+	}
+
+	c <- gatherResult{records, nil}
+	return
+}

pkg/gather/clusterconfig/openshift_sdn_logs.go

@@ -40,6 +40,7 @@ func GatherOpenshiftSDNLogs(g *Gatherer, c chan<- gatherResult) {
 		1024*64, // maximum 64 kb of logs
 		"errors",
 		"app=sdn",
+		false,
 	)
 	if err != nil {
 		c <- gatherResult{nil, []error{err}}