Reduce timeout for scrapping IMDS and give instruction when fail to s…

…crape IMDS inside container
aws · Jun 21, 2022 · 9921a14 · 9921a14
2 parents 0910c4e + 5335531
commit 9921a14
Show file tree

Hide file tree

Showing 68 changed files with 4,432 additions and 1,110 deletions.
diff --git a/.github/workflows/build-test-linux.yml b/.github/workflows/build-test-linux.yml
@@ -27,7 +27,7 @@ jobs:
     - name: Set up Go 1.x
       uses: actions/setup-go@v2
       with:
-        go-version: ~1.15.15
+        go-version: ~1.17.8
 
     - name: Check out code
       uses: actions/checkout@v2
@@ -41,15 +41,15 @@ jobs:
 
     - uses: zencargo/github-action-go-mod-tidy@v1
       with:
-        go-version: 1.15
+        go-version: 1.17.8
 
     - name: Cache build output
       uses: actions/cache@v2
       with:
         path: |
               ~/.cache/go-build
               ~/go/pkg/mod
-        key: ${{ runner.os }}-go-${{ hashFiles('**/go.sum') }}
+        key: v1-go-pkg-mod-${{ runner.os }}-${{ hashFiles('**/go.sum') }}
 
     - name: Test
       run: make test

diff --git a/.github/workflows/build-test-macos.yml b/.github/workflows/build-test-macos.yml
@@ -27,7 +27,7 @@ jobs:
     - name: Set up Go 1.x
       uses: actions/setup-go@v2
       with:
-        go-version: ~1.15.15
+        go-version: ~1.17.8
 
     - name: Check out code
       uses: actions/checkout@v2
@@ -41,7 +41,7 @@ jobs:
         path: |
               ~/Library/Caches/go-build
               ~/go/pkg/mod
-        key: ${{ runner.os }}-go-${{ hashFiles('**/go.sum') }}
+        key: v1-go-pkg-mod-${{ runner.os }}-${{ hashFiles('**/go.sum') }}
 
     - name: Test
       run: make test

diff --git a/.github/workflows/build-test-windows.yml b/.github/workflows/build-test-windows.yml
@@ -31,15 +31,15 @@ jobs:
       - name: Setup Go
         uses: actions/setup-go@v2
         with:
-          go-version: ~1.15.15
+          go-version: ~1.17.8
 
       - name: Cache Go
         uses: actions/cache@v2
         with:
           path: |
             %LocalAppData%\go-build
             ~/go/pkg/mod
-          key: ${{ runner.os }}-go-${{ hashFiles('**/go.sum') }}
+          key: v1-go-pkg-mod-${{ runner.os }}-${{ hashFiles('**/go.sum') }}
 
       - name: Install make
         run: choco install make

diff --git a/.github/workflows/integrationTest.yml b/.github/workflows/integrationTest.yml
@@ -89,7 +89,7 @@ jobs:
       - name: Set up Go 1.x
         uses: actions/setup-go@v2
         with:
-          go-version: ~1.15.15
+          go-version: ~1.17.8
 
       - name: Install rpm
         run: sudo apt install rpm
@@ -116,7 +116,7 @@ jobs:
           path: |
             ~/go/pkg/mod
             ~/.cache/go-build
-          key: v1-go-pkg-mod-${{ hashFiles('**/go.sum') }}
+          key: v1-go-pkg-mod-${{ runner.os }}-${{ hashFiles('**/go.sum') }}
 
       - name: Cache binaries
         id: cached_binaries
@@ -181,7 +181,7 @@ jobs:
       - name: Set up Go 1.x
         uses: actions/setup-go@v2
         with:
-          go-version: ~1.15.15
+          go-version: ~1.17.8
 
       - name: Configure AWS Credentials
         uses: aws-actions/configure-aws-credentials@v1
@@ -238,7 +238,7 @@ jobs:
       - name: Set up Go 1.x
         uses: actions/setup-go@v2
         with:
-          go-version: ~1.15.15
+          go-version: ~1.17.8
 
       - name: Configure AWS Credentials
         uses: aws-actions/configure-aws-credentials@v1

diff --git a/.github/workflows/nightly-build.yml b/.github/workflows/nightly-build.yml
@@ -15,7 +15,7 @@ jobs:
       - name: Set up Go 1.x
         uses: actions/setup-go@v2
         with:
-          go-version: ~1.15.15
+          go-version: ~1.17.8
         id: go
 
       #Need to install rpm so ubuntu can make rpm by default ubuntu can make deb

diff --git a/amazon-cloudwatch-container-insights/cloudwatch-agent-dockerfile/source/Dockerfile b/amazon-cloudwatch-container-insights/cloudwatch-agent-dockerfile/source/Dockerfile
@@ -6,11 +6,17 @@ FROM $GO_IMAGE as builder
 # NOTE: This arg will be populated by docker buildx
 # https://docs.docker.com/engine/reference/builder/#automatic-platform-args-in-the-global-scope
 ARG TARGETARCH
+ENV GOPROXY=direct
 RUN mkdir -p /go/src/github.com/aws/amazon-cloudwatch-agent/
 WORKDIR /go/src/github.com/aws/amazon-cloudwatch-agent/
 
 ARG GO111MODULE="on"
 ENV GO111MODULE=${GO111MODULE}
+
+COPY go.mod /go/src/github.com/aws/amazon-cloudwatch-agent/
+COPY go.sum /go/src/github.com/aws/amazon-cloudwatch-agent/
+RUN go mod download -x
+
 COPY . /go/src/github.com/aws/amazon-cloudwatch-agent/
 RUN make build-for-docker-${TARGETARCH}
 

diff --git a/...loyment-manifest-templates/deployment-mode/service/cwagent-prometheus/prometheus-eks.yaml b/...loyment-manifest-templates/deployment-mode/service/cwagent-prometheus/prometheus-eks.yaml
@@ -47,8 +47,8 @@ data:
                   ]
                 },
                 {
-                  "source_labels": ["Service", "backend"],
-                  "label_matcher": ".*haproxy-ingress-.*metrics;(httpback-shared-backend|httpback-default-backend|httpsback-shared-backend|_default_backend|.*backend_http)",
+                  "source_labels": ["Service"],
+                  "label_matcher": ".*haproxy-ingress-.*metrics",
                   "dimensions": [["Service","Namespace","ClusterName","backend","code"]],
                   "metric_selectors": [
                     "^haproxy_backend_http_responses_total$"
@@ -230,7 +230,6 @@ data:
         source_labels:
         - __meta_kubernetes_pod_phase
         target_label: pod_phase
-
     - job_name: 'kubernetes-pod-fluentbit-plugin'
       sample_limit: 10000
       metrics_path: /api/v1/metrics/prometheus
@@ -274,7 +273,6 @@ data:
         source_labels:
         - __meta_kubernetes_pod_node_name
         target_label: NodeName
-
     - job_name: kubernetes-service-endpoints
       sample_limit: 10000
       kubernetes_sd_configs:
@@ -341,7 +339,6 @@ data:
         replacement: "$1"
       - regex: proxy
         action: labeldrop
-
     - job_name: 'kubernetes-pod-jmx'
       sample_limit: 10000
       metrics_path: /metrics
@@ -380,7 +377,6 @@ data:
       - source_labels: [__name__]
         regex: 'jvm_gc_collection_seconds.*'
         action: drop
-
 kind: ConfigMap
 metadata:
   name: prometheus-config
@@ -402,16 +398,16 @@ metadata:
 rules:
   - apiGroups: [""]
     resources:
-    - nodes
-    - nodes/proxy
-    - services
-    - endpoints
-    - pods
+      - nodes
+      - nodes/proxy
+      - services
+      - endpoints
+      - pods
     verbs: ["get", "list", "watch"]
   - apiGroups:
-    - extensions
+      - extensions
     resources:
-    - ingresses
+      - ingresses
     verbs: ["get", "list", "watch"]
   - nonResourceURLs: ["/metrics"]
     verbs: ["get"]
@@ -461,7 +457,7 @@ spec:
           # Please don't change below envs
           env:
             - name: CI_VERSION
-              value: "k8s/1.2.4"
+              value: "k8s/1.3.9"
           # Please don't change the mountPath
           volumeMounts:
             - name: prometheus-cwagentconfig
@@ -477,4 +473,4 @@ spec:
           configMap:
             name: prometheus-config
       terminationGracePeriodSeconds: 60
-      serviceAccountName: cwagent-prometheus
+      serviceAccountName: cwagent-prometheus
diff --git a/...ontainer-insights/k8s-yaml-templates/cwagent-kubernetes-monitoring/cwagent-daemonset.yaml b/...ontainer-insights/k8s-yaml-templates/cwagent-kubernetes-monitoring/cwagent-daemonset.yaml
@@ -16,6 +16,7 @@ spec:
       containers:
         - name: cloudwatch-agent
           image: amazon/cloudwatch-agent:1.230621.0
+          imagePullPolicy: Always
           #ports:
           #  - containerPort: 8125
           #    hostPort: 8125
@@ -56,6 +57,9 @@ spec:
             - name: varlibdocker
               mountPath: /var/lib/docker
               readOnly: true
+            - name: containerdsock
+              mountPath: /run/containerd/containerd.sock
+              readOnly: true
             - name: sys
               mountPath: /sys
               readOnly: true
@@ -75,6 +79,9 @@ spec:
         - name: varlibdocker
           hostPath:
             path: /var/lib/docker
+        - name: containerdsock
+          hostPath:
+            path: /run/containerd/containerd.sock
         - name: sys
           hostPath:
             path: /sys

diff --git a/cmd/amazon-cloudwatch-agent/amazon-cloudwatch-agent.go b/cmd/amazon-cloudwatch-agent/amazon-cloudwatch-agent.go
@@ -31,7 +31,6 @@ import (
 	"github.com/aws/amazon-cloudwatch-agent/logs"
 	"github.com/aws/amazon-cloudwatch-agent/profiler"
 
-	lumberjack "github.com/aws/amazon-cloudwatch-agent/logger"
 	_ "github.com/aws/amazon-cloudwatch-agent/plugins"
 	"github.com/influxdata/telegraf/agent"
 	"github.com/influxdata/telegraf/config"
@@ -47,6 +46,7 @@ import (
 
 const (
 	defaultEnvCfgFileName = "env-config.json"
+	LogTargetEventLog = "eventlog"
 )
 
 var fDebug = flag.Bool("debug", false,
@@ -273,14 +273,14 @@ func runAgent(ctx context.Context,
 		return errors.New("Error: no inputs found, did you provide a valid config file?")
 	}
 
-	if int64(c.Agent.Interval.Duration) <= 0 {
-		return fmt.Errorf("Agent interval must be positive, found %s",
-			c.Agent.Interval.Duration)
+	if int64(c.Agent.Interval) <= 0 {
+		return fmt.Errorf("Agent interval must be positive, found %v",
+			c.Agent.Interval)
 	}
 
-	if int64(c.Agent.FlushInterval.Duration) <= 0 {
-		return fmt.Errorf("Agent flush_interval must be positive; found %s",
-			c.Agent.Interval.Duration)
+	if int64(c.Agent.FlushInterval) <= 0 {
+		return fmt.Errorf("Agent flush_interval must be positive; found %v",
+			c.Agent.FlushInterval)
 	}
 
 	if *fSchemaTest {
@@ -304,6 +304,7 @@ func runAgent(ctx context.Context,
 		RotationInterval:    ag.Config.Agent.LogfileRotationInterval,
 		RotationMaxSize:     ag.Config.Agent.LogfileRotationMaxSize,
 		RotationMaxArchives: ag.Config.Agent.LogfileRotationMaxArchives,
+		LogWithTimezone:     "",
 	}
 
 	logger.SetupLogging(logConfig)
@@ -548,11 +549,10 @@ func main() {
 			}
 			os.Exit(0)
 		} else {
-			winlogger, err := s.Logger(nil)
-			if err == nil {
-				//When in service mode, register eventlog target and setup default logging to eventlog
-				logger.RegisterEventLogger(winlogger)
-				logger.SetupLogging(logger.LogConfig{LogTarget: lumberjack.LogTargetLumberjack})
+			// When in service mode, register eventlog target and setup default logging to eventlog
+			e := RegisterEventLogger()
+			if e != nil {
+				log.Println("E! Cannot register event log " + e.Error())
 			}
 			err = s.Run()
 

diff --git a/cmd/amazon-cloudwatch-agent/register_event_logger_notwindows.go b/cmd/amazon-cloudwatch-agent/register_event_logger_notwindows.go
@@ -0,0 +1,17 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: MIT
+
+//go:build !windows
+// +build !windows
+
+package main
+
+// RegisterEventLogger is for supporting Windows Event, it should only be created when running on Windows
+// To minimize duplicate code for Windows vs non-Windows build amazon-cloudwatch-agent.go main class,
+// create this dummy method so amazon-cloudwatch-agent.go can be build independent to the OS. Because of this method is
+// invoked inside the if statement of "if runtime.GOOS == "windows" && windowsRunAsService() {" effectively it
+// is unreachable, but Go compiler needs to see this exits to build.
+func RegisterEventLogger() error {
+	// Unreachable code, do nothing.
+	return nil
+}
diff --git a/cmd/amazon-cloudwatch-agent/register_event_logger_windows.go b/cmd/amazon-cloudwatch-agent/register_event_logger_windows.go
@@ -0,0 +1,24 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: MIT
+
+//go:build windows
+// +build windows
+
+package main
+
+import (
+	lumberjack "github.com/aws/amazon-cloudwatch-agent/logger"
+	"github.com/influxdata/telegraf/logger"
+)
+
+// RegisterEventLogger is for supporting Windows Event
+func RegisterEventLogger() error {
+	// When in service mode, register eventlog target and setup default logging to eventlog
+
+	e := logger.RegisterEventLogger(LogTargetEventLog)
+	if e != nil {
+		return e
+	}
+	logger.SetupLogging(logger.LogConfig{LogTarget: lumberjack.LogTargetLumberjack})
+	return nil
+}
diff --git a/cmd/start-amazon-cloudwatch-agent/start-amazon-cloudwatch-agent.go b/cmd/start-amazon-cloudwatch-agent/start-amazon-cloudwatch-agent.go
@@ -5,6 +5,7 @@ package main
 
 import (
 	"io"
+	"io/ioutil"
 	"log"
 	"os"
 	"os/exec"
@@ -92,8 +93,29 @@ func main() {
 		log.Fatalf("E! Cannot translate JSON config into TOML, ERROR is %v \n", err)
 	}
 	log.Printf("I! Config has been translated into TOML %s \n", tomlConfigPath)
+	printFileContents(tomlConfigPath)
 
 	if err := startAgent(writer); err != nil {
+		log.Printf("E! Error when starting Agent, Error is %v \n", err)
 		os.Exit(1)
 	}
 }
+
+func printFileContents(path string) {
+	file, err := os.Open(path)
+	if err != nil {
+		log.Printf("E! Error when printing file(%s) contents, Error is %v \n", path, err)
+		os.Exit(1)
+	}
+	defer func() {
+		if err = file.Close(); err != nil {
+			log.Printf("E! Error when closing file,  Error is %v \n", err)
+		}
+	}()
+
+	b, err := ioutil.ReadAll(file)
+	if err != nil {
+		log.Printf("E! Error when reading file(%s), Error is %v \n", path, err)
+	}
+	log.Printf("D! toml config %v", string(b))
+}