diff --git a/.github/workflows/dataflow_engine_chaos.yaml b/.github/workflows/dataflow_engine_chaos.yaml
new file mode 100644
index 00000000..b19c03b3
--- /dev/null
+++ b/.github/workflows/dataflow_engine_chaos.yaml
@@ -0,0 +1,227 @@
+name: Dataflow Engine Chaos
+
+on:
+  schedule:
+    - cron: '0 17-23 * * *' # run at minute 0 every hour from 01:00 ~ 07:00 UTC+8
+  pull_request:
+    branches: [ master ]
+
+# See: https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#concurrency.
+concurrency:
+  group: ${{ github.ref }}-${{ github.workflow }}
+  cancel-in-progress: true
+
+# A workflow run is made up of one or more jobs that can run sequentially or in parallel
+jobs:
+  # This workflow contains a single job called "base"
+  base:
+    # The type of runner that the job will run on
+    runs-on: ubuntu-18.04
+    timeout-minutes: 30
+    strategy:
+      fail-fast: false
+      matrix:
+        chaos-obj:
+          [
+            "pod-failure-dfe",
+            "pod-kill-dfe",
+          ]
+
+    # Steps represent a sequence of tasks that will be executed as part of the job
+    steps:
+      - uses: actions/checkout@v2
+
+      - uses: actions/setup-go@v3
+        with:
+          go-version: 1.18
+
+      - name: Cache go modules
+        uses: actions/cache@v2
+        with:
+          path: ~/go/pkg/mod
+          key: ${{ runner.os }}-dataflow-${{ hashFiles('go.sum') }}
+
+      # Set up Kubernetes with K3s
+      - name: Set up K3s cluster
+        run: |
+          curl -sfL https://get.k3s.io | INSTALL_K3S_VERSION=v1.18.9+k3s1 sh -s - \
+            --write-kubeconfig-mode=644 \
+            "${k3s_disable_command:---disable}" metrics-server \
+            "${k3s_disable_command:---disable}" traefik \
+            --flannel-backend=none \
+            --docker
+        shell: bash
+
+      # this may be failed sometimes, and I want to exit the workflow directly if failed,
+      # but GitHub Actions doesnt' support early-exit yet, see https://github.com/actions/runner/issues/662.
+      # so, simply wait for a long time.
+      - name: Wait for coredns
+        run: |
+          kubectl rollout status --watch --timeout 600s deployment/coredns -n kube-system
+        shell: bash
+        env:
+          KUBECONFIG: /etc/rancher/k3s/k3s.yaml
+
+      - name: Export KUBECONFIG environment variable
+        run: |
+          echo 'KUBECONFIG=/etc/rancher/k3s/k3s.yaml' >> $GITHUB_ENV
+        shell: bash
+
+      - name: Print cluster information
+        run: |
+          kubectl config view
+          kubectl cluster-info
+          kubectl get nodes
+          kubectl get pods -n kube-system
+          kubectl get sc
+          kubectl version
+
+      - name: Build dataflow engine binary
+        run: make df-master df-executor df-chaos-case
+
+      - name: Build Dataflow engine docker image
+        run: |
+          cp -r $GITHUB_WORKSPACE/chaos/manifests/conf/ $GITHUB_WORKSPACE/bin/
+          docker build -f $GITHUB_WORKSPACE/chaos/manifests/Dockerfile -t dataflow:chaos $GITHUB_WORKSPACE/bin
+          docker image list
+
+      # Set up metastore and basic services
+      - name: Set up metastore and basic services
+        run: |
+          kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/metastore.yaml
+          kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/metastore.yaml
+          kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/metastore.yaml
+      - name: Wait for metastore ready
+        run: |
+          kubectl wait --for=condition=Ready pod/metastore-framework-mysql-0 --timeout=60s || true
+          kubectl wait --for=condition=Ready pod/metastore-user-etcd-0 --timeout=60s || true
+
+          echo show pvc
+          kubectl get pvc -l app=metastore -o wide
+          echo show pv
+          kubectl get pv -o wide
+          echo show svc
+          kubectl get svc -l app=metastore -o wide
+          echo show sts
+          kubectl get sts -l app=metastore -o wide
+          echo show po
+          kubectl get po -l app=metastore -o wide
+          echo describe po
+          kubectl describe po -l app=metastore
+          echo describe pvc
+          kubectl describe pvc -l app=metastore
+          kubectl wait --for=condition=Ready pod/metastore-framework-mysql-0 --timeout=0s
+          kubectl wait --for=condition=Ready pod/metastore-user-etcd-0 --timeout=0s
+
+      - name: Set up server-master
+        run: |
+          kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/server-master.yaml
+          kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/server-master.yaml
+          kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/server-master.yaml
+
+      - name: Wait for server-master ready
+        run: |
+          kubectl wait --for=condition=Ready pod -l app=server-master --all --timeout=60s|| true
+          echo "<<<<< show pvc >>>>>"
+          kubectl get pvc -l app=server-master -o wide
+          echo "<<<<< show pv >>>>>"
+          kubectl get pv -o wide
+          echo "<<<<< show svc >>>>>"
+          kubectl get svc -l app=server-master -o wide
+          echo "<<<<< show sts >>>>>"
+          kubectl get sts -l app=server-master -o wide
+          echo "<<<<< show po >>>>>"
+          kubectl get po -l app=server-master -o wide
+          echo "<<<<< describe po >>>>>"
+          kubectl describe po -l app=server-master
+          echo "<<<<< describe pvc >>>>>"
+          kubectl describe pvc -l app=server-master
+          echo "<<<<< show current log for server-master-0 >>>>>"
+          kubectl logs server-master-0 || true
+          echo "<<<<< show previous log for server-master-0 >>>>>"
+          kubectl logs server-master-0 -p || true
+          echo "<<<<< show current log for server-master-1 >>>>>"
+          kubectl logs server-master-1 || true
+          echo "<<<<< show previous log for server-master-1 >>>>>"
+          kubectl logs server-master-1 -p || true
+          echo "<<<<< show current log for server-master-2 >>>>>"
+          kubectl logs server-master-2 || true
+          echo "<<<<< show previous log for server-master-2 >>>>>"
+          kubectl logs server-master-2 -p || true
+
+      - name: Set up executor
+        run: |
+          kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/executor.yaml
+          kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/executor.yaml
+          kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/executor.yaml
+
+      - name: Wait for executor ready
+        run: |
+          kubectl wait --for=condition=Ready pod -l app=executor --all --timeout=60s|| true
+          echo "<<<<< show pvc >>>>>"
+          kubectl get pvc -l app=executor -o wide
+          echo "<<<<< show pv >>>>>"
+          kubectl get pv -o wide
+          echo "<<<<< show svc >>>>>"
+          kubectl get svc -l app=executor -o wide
+          echo "<<<<< show sts >>>>>"
+          kubectl get sts -l app=executor -o wide
+          echo "<<<<< show po >>>>>"
+          kubectl get po -l app=executor -o wide
+          echo "<<<<< describe po >>>>>"
+          kubectl describe po -l app=executor
+          echo "<<<<< describe pvc >>>>>"
+          kubectl describe pvc -l app=executor
+          echo "<<<<< show current log for executor-0 >>>>>"
+          kubectl logs executor-0 || true
+          echo "<<<<< show previous log for executor-0 >>>>>"
+          kubectl logs executor-0 -p || true
+          echo "<<<<< show current log for executor-1 >>>>>"
+          kubectl logs executor-1 || true
+          echo "<<<<< show previous log for worker-master-1 >>>>>"
+          kubectl logs executor-1 -p || true
+          echo "<<<<< show current log for executor-2 >>>>>"
+          kubectl logs executor-2 || true
+          echo "<<<<< show previous log for executor-2 >>>>>"
+          kubectl logs executor-2 -p || true
+
+      - name: Set up chaos test cases
+        run: |
+          kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/cases.yaml
+          kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/cases.yaml
+          kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/cases.yaml
+
+      # - name: Encode chaos-mesh action
+      #   run: |
+      #     echo CFG_BASE64=$(base64 -w 0 $GITHUB_WORKSPACE/chaos/manifests/${{ matrix.chaos-obj }}.yaml) >> $GITHUB_ENV
+
+      # - name: Run chaos mesh action
+      #   uses: chaos-mesh/chaos-mesh-action@master
+      #   env:
+      #     CFG_BASE64: ${{ env.CFG_BASE64 }}
+      #     CHAOS_MESH_VERSION: v1.0.0
+
+      # check whether complete with 1m * 20 times.
+      - name: Wait for chaos test case complete
+        run: |
+          $GITHUB_WORKSPACE/chaos/scripts/check-case.sh
+
+      - name: Copy logs to hack permission
+        if: ${{ always() }}
+        run: |
+          mkdir ./logs
+          sudo cp -r -L /var/log/containers/. ./logs
+          sudo find /var/ -type f -regex '.*/(server-master|executor).log$' | sudo xargs -i cp {} ./logs || true
+          sudo chown -R runner ./logs
+
+      # Upload logs as artifact seems not stable, so we set `continue-on-error: true` here.
+      - name: Upload logs
+        continue-on-error: true
+        uses: actions/upload-artifact@v2
+        if: ${{ always() }}
+        with:
+          name: chaos-base-logs.${{ matrix.chaos-obj }}
+          path: |
+            ./logs
+            !./logs/coredns-*
+            !./logs/local-path-provisioner-*
diff --git a/Makefile b/Makefile
index 66c4df40..ae510ba8 100644
--- a/Makefile
+++ b/Makefile
@@ -1,6 +1,7 @@
 TEST_DIR := /tmp/dataflow_engine_test
 PARALLEL=3
 GO       := GO111MODULE=on go
+GOBUILD  := CGO_ENABLED=0 $(GO) build -trimpath
 GOTEST := CGO_ENABLED=1 go test -p $(PARALLEL) --race
 FAIL_ON_STDOUT := awk '{ print  } END { if (NR > 0) { exit 1  }  }'
 
@@ -21,20 +22,23 @@ df-proto:
 	./generate-proto.sh
 
 df-master:
-	go build -o bin/master ./cmd/master
+	$(GOBUILD) -o bin/master ./cmd/master
 	cp ./bin/master ./ansible/roles/common/files/master.bin
 
 df-executor:
-	go build -o bin/executor ./cmd/executor
+	$(GOBUILD) -o bin/executor ./cmd/executor
 	cp ./bin/executor ./ansible/roles/common/files/executor.bin
 
 df-master-client:
-	go build -o bin/master-client ./cmd/master-client
+	$(GOBUILD) -o bin/master-client ./cmd/master-client
 
 df-demo:
-	go build -o bin/demoserver ./cmd/demoserver
+	$(GOBUILD) -o bin/demoserver ./cmd/demoserver
 	cp ./bin/demoserver ./ansible/roles/common/files/demoserver.bin
 
+df-chaos-case:
+	$(GOBUILD) -o bin/df-chaos-case ./chaos/cases
+
 unit_test: check_failpoint_ctl
 	mkdir -p "$(TEST_DIR)"
 	$(FAILPOINT_ENABLE)
diff --git a/chaos/cases/cases.go b/chaos/cases/cases.go
new file mode 100644
index 00000000..b9ec39af
--- /dev/null
+++ b/chaos/cases/cases.go
@@ -0,0 +1,30 @@
+// Copyright 2022 PingCAP, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"context"
+
+	"github.com/pingcap/tiflow/dm/pkg/log"
+	"go.uber.org/zap"
+)
+
+var cases = []string{"fake-job-normal", "fake-job-fast-finish"}
+
+func runCases(ctx context.Context) error {
+	for _, c := range cases {
+		log.L().Info("run case successfully", zap.String("case", c))
+	}
+	return nil
+}
diff --git a/chaos/cases/config.go b/chaos/cases/config.go
new file mode 100644
index 00000000..3e80d99f
--- /dev/null
+++ b/chaos/cases/config.go
@@ -0,0 +1,50 @@
+// Copyright 2020 PingCAP, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"flag"
+	"time"
+)
+
+// config is used to run chaos tests.
+type config struct {
+	*flag.FlagSet `toml:"-" yaml:"-" json:"-"`
+
+	MasterAddr string        `toml:"master-addr" yaml:"master-addr" json:"master-addr"`
+	Duration   time.Duration `toml:"duration" yaml:"duration" json:"duration"`
+
+	MasterCount int `toml:"master-count" yaml:"master-count" json:"master-count"`
+	WorkerCount int `toml:"worker-count" yaml:"worker-count" json:"worker-count"`
+}
+
+// newConfig creates a config for this chaos testing suite.
+func newConfig() *config {
+	cfg := &config{}
+	cfg.FlagSet = flag.NewFlagSet("chaos-case", flag.ContinueOnError)
+	fs := cfg.FlagSet
+
+	fs.StringVar(&cfg.MasterAddr, "master-addr", "server-master:10240", "address of server-master")
+	fs.DurationVar(&cfg.Duration, "duration", 20*time.Minute, "duration of cases running")
+
+	fs.IntVar(&cfg.MasterCount, "master-count", 3, "expect count of server-master")
+	fs.IntVar(&cfg.WorkerCount, "worker-count", 4, "expect count of executor")
+
+	return cfg
+}
+
+// parse parses flag definitions from the argument list.
+func (c *config) parse(args []string) error {
+	return c.FlagSet.Parse(args)
+}
diff --git a/chaos/cases/main.go b/chaos/cases/main.go
new file mode 100644
index 00000000..ab15da6d
--- /dev/null
+++ b/chaos/cases/main.go
@@ -0,0 +1,91 @@
+// Copyright 2022 PingCAP, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"math/rand"
+	"net/http"
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
+
+	"github.com/pingcap/errors"
+	"github.com/pingcap/tiflow/dm/pkg/log"
+	"go.uber.org/zap"
+)
+
+// main starts to run the test case logic after MySQL, TiDB and DM have been set up.
+// NOTE: run this in the same K8s namespace as DM-master.
+func main() {
+	code := 0
+	defer func() {
+		os.Exit(code)
+	}()
+
+	cfg := newConfig()
+	err := cfg.parse(os.Args[1:])
+	switch errors.Cause(err) {
+	case nil:
+	case flag.ErrHelp:
+		return
+	default:
+		fmt.Println("parse cmd flags err:", err.Error())
+		code = 2
+		return
+	}
+
+	err = log.InitLogger(&log.Config{
+		File:  "chaos-case.log",
+		Level: "info",
+	})
+	if err != nil {
+		fmt.Println("init logger error:", err.Error())
+		code = 2
+		return
+	}
+
+	go func() {
+		//nolint:errcheck
+		http.ListenAndServe("0.0.0.0:8899", nil) // for pprof
+	}()
+
+	rand.Seed(time.Now().UnixNano())
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	sc := make(chan os.Signal, 1)
+	signal.Notify(sc,
+		syscall.SIGHUP,
+		syscall.SIGINT,
+		syscall.SIGTERM,
+		syscall.SIGQUIT)
+	go func() {
+		sig := <-sc
+		log.L().Info("got signal to exit", zap.Stringer("signal", sig))
+		cancel()
+	}()
+
+	// run tests cases
+	err = runCases(ctx)
+	if err != nil {
+		log.L().Error("run cases failed", zap.Error(err))
+		code = 2
+		return
+	}
+}
diff --git a/chaos/manifests/Dockerfile b/chaos/manifests/Dockerfile
new file mode 100644
index 00000000..7a78ed7e
--- /dev/null
+++ b/chaos/manifests/Dockerfile
@@ -0,0 +1,12 @@
+FROM alpine:3.14
+
+ADD master /df-server-master
+ADD executor /df-executor
+ADD df-chaos-case /df-chaos-case
+ADD conf /conf
+
+RUN chmod a+x /df-server-master /df-executor /df-chaos-case
+
+WORKDIR /
+
+EXPOSE 10239 10240 10241
diff --git a/chaos/manifests/cases.yaml b/chaos/manifests/cases.yaml
new file mode 100644
index 00000000..21316c5c
--- /dev/null
+++ b/chaos/manifests/cases.yaml
@@ -0,0 +1,16 @@
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: chaos-test-case
+spec:
+  template:
+    spec:
+      containers:
+        - name: chaos-test-case
+          image: dataflow:chaos
+          imagePullPolicy: IfNotPresent
+          command:
+            - "/df-chaos-case"
+            - "--duration=20m"
+      restartPolicy: Never
+  backoffLimit: 0 # fail immediately
diff --git a/chaos/manifests/conf/executor.toml b/chaos/manifests/conf/executor.toml
new file mode 100644
index 00000000..395201d8
--- /dev/null
+++ b/chaos/manifests/conf/executor.toml
@@ -0,0 +1,3 @@
+keepalive-ttl = "20s"
+keepalive-interval = "500ms"
+session-ttl = 20
diff --git a/chaos/manifests/conf/server-master.toml b/chaos/manifests/conf/server-master.toml
new file mode 100644
index 00000000..6981dd49
--- /dev/null
+++ b/chaos/manifests/conf/server-master.toml
@@ -0,0 +1,10 @@
+[etcd]
+data-dir = "/data/etcd"
+
+[frame-metastore-conf]
+endpoints = ["metastore-framework-mysql-0.metastore:3306"]
+auth.user = "root"
+auth.passwd = ""
+
+[user-metastore-conf]
+endpoints = ["metastore-user-etcd-0.metastore:12479"]
diff --git a/chaos/manifests/executor.yaml b/chaos/manifests/executor.yaml
new file mode 100644
index 00000000..c444cdf6
--- /dev/null
+++ b/chaos/manifests/executor.yaml
@@ -0,0 +1,75 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: executor
+  labels:
+    app: executor
+spec:
+  ports:
+    - name: executor
+      port: 10241
+      targetPort: 10241
+  selector:
+    app: executor
+---
+apiVersion: apps/v1
+kind: StatefulSet
+metadata:
+  name: executor
+  labels:
+    app: executor
+spec:
+  selector:
+    matchLabels:
+      app: executor
+  serviceName: executor
+  replicas: 4
+  podManagementPolicy: Parallel
+  template:
+    metadata:
+      labels:
+        app: executor
+    spec:
+      containers:
+        - name: executor
+          image: dataflow:chaos # build this image in GitHub action workflow
+          imagePullPolicy: IfNotPresent
+          volumeMounts:
+            - mountPath: /log
+              name: executor-log
+          env:
+            - name: MY_POD_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.name
+            - name: MY_POD_NAMESPACE
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.namespace
+          ports:
+            - containerPort: 10241
+              name: executor
+          command:
+            - "/df-executor"
+            - "--name=$(MY_POD_NAME)"
+            - "--worker-addr=0.0.0.0:10241"
+            - "--advertise-addr=$(MY_POD_NAME).executor.$(MY_POD_NAMESPACE):10241"
+            - "--join=server-master-0.server-master.$(MY_POD_NAMESPACE):10240,server-master-1.server-master.$(MY_POD_NAMESPACE):10240,server-master-2.server-master.$(MY_POD_NAMESPACE):10240"
+            - "--config=/conf/executor.toml"
+          readinessProbe:
+            httpGet:
+              port: 10241
+              path: /metrics # TODO: use a real probe url
+              scheme: HTTP
+            initialDelaySeconds: 5
+            periodSeconds: 5
+            failureThreshold: 5
+  volumeClaimTemplates:
+    - metadata:
+        name: executor-log
+      spec:
+        accessModes:
+          - ReadWriteOnce
+        resources:
+          requests:
+            storage: 1Gi
diff --git a/chaos/manifests/metastore.yaml b/chaos/manifests/metastore.yaml
new file mode 100644
index 00000000..0306a1c1
--- /dev/null
+++ b/chaos/manifests/metastore.yaml
@@ -0,0 +1,106 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: metastore
+  labels:
+    app: metastore
+spec:
+  ports:
+    - name: port-mysql # note the name is no more than 15 characters
+      port: 3306
+      targetPort: 3306
+    - name: port-etcd # note the name is no more than 15 characters
+      port: 12479
+      targetPort: 2479
+  selector:
+    app: metastore
+
+---
+apiVersion: apps/v1
+kind: StatefulSet
+metadata:
+  name: metastore-framework-mysql
+  labels:
+    app: metastore
+spec:
+  selector:
+    matchLabels:
+      app: metastore
+  serviceName: metastore
+  replicas: 1
+  podManagementPolicy: Parallel
+  template:
+    metadata:
+      labels:
+        app: metastore
+    spec:
+      containers:
+        - name: metastore-framework-mysql
+          image: mysql:5.7
+          imagePullPolicy: IfNotPresent
+          volumeMounts:
+            - mountPath: "/var/lib/mysql"
+              name: metastore-framework-mysql
+          env:
+            - name: MYSQL_ALLOW_EMPTY_PASSWORD
+              value: "true"
+          ports:
+            - containerPort: 3306
+              name: port-mysql
+          args:
+            - "--server-id=1"
+  volumeClaimTemplates:
+    - metadata:
+        name: metastore-framework-mysql
+      spec:
+        accessModes:
+          - ReadWriteOnce
+        resources:
+          requests:
+            storage: 5Gi
+
+---
+apiVersion: apps/v1
+kind: StatefulSet
+metadata:
+  name: metastore-user-etcd
+  labels:
+    app: metastore
+spec:
+  selector:
+    matchLabels:
+      app: metastore
+  serviceName: metastore
+  replicas: 1
+  podManagementPolicy: Parallel
+  template:
+    metadata:
+      labels:
+        app: metastore
+    spec:
+      containers:
+        - name: metastore-user-etcd
+          image: quay.io/coreos/etcd:v3.5.4
+          imagePullPolicy: IfNotPresent
+          ports:
+            - containerPort: 2479
+              name: port-etcd
+          command:
+            - "etcd"
+          args:
+            - "--name=metastore-user-etcd"
+            - "--advertise-client-urls=http://0.0.0.0:2479"
+            - "--listen-client-urls=http://0.0.0.0:2479"
+            - "--listen-peer-urls=http://127.0.0.1:2480"
+            - "--initial-advertise-peer-urls=http://127.0.0.1:2480"
+            - "--initial-cluster=metastore-user-etcd=http://127.0.0.1:2480"
+            - "--initial-cluster-state=new"
+  volumeClaimTemplates:
+    - metadata:
+        name: metastore-user-etcd
+      spec:
+        accessModes:
+          - ReadWriteOnce
+        resources:
+          requests:
+            storage: 5Gi
diff --git a/chaos/manifests/server-master.yaml b/chaos/manifests/server-master.yaml
new file mode 100644
index 00000000..fb7e5d3d
--- /dev/null
+++ b/chaos/manifests/server-master.yaml
@@ -0,0 +1,84 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: server-master
+  labels:
+    app: server-master
+spec:
+  ports:
+    - name: port-master
+      port: 10240
+      targetPort: 10240
+    - name: port-mpeer
+      port: 10239
+      targetPort: 10239
+  selector:
+    app: server-master
+---
+apiVersion: apps/v1
+kind: StatefulSet
+metadata:
+  name: server-master
+  labels:
+    app: server-master
+spec:
+  selector:
+    matchLabels:
+      app: server-master
+  serviceName: server-master
+  replicas: 3
+  podManagementPolicy: Parallel
+  template:
+    metadata:
+      labels:
+        app: server-master
+    spec:
+      containers:
+        - name: server-master
+          image: dataflow:chaos
+          imagePullPolicy: IfNotPresent
+          volumeMounts:
+            - mountPath: /data
+              name: server-master-data
+            - mountPath: /log
+              name: server-master-log
+          env:
+            - name: MY_POD_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.name
+            - name: MY_POD_NAMESPACE
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.namespace
+          ports:
+            - containerPort: 10240
+              name: port-master
+            - containerPort: 10239
+              name: port-mpeer
+          command:
+            - "/df-server-master"
+            - "--name=$(MY_POD_NAME)"
+            - "--master-addr=0.0.0.0:10240"
+            - "--advertise-addr=$(MY_POD_NAME).server-master.$(MY_POD_NAMESPACE):10240"
+            - "--peer-urls=0.0.0.0:10239"
+            - "--advertise-peer-urls=http://$(MY_POD_NAME).server-master.$(MY_POD_NAMESPACE):10239"
+            - "--initial-cluster=server-master-0=http://server-master-0.server-master.$(MY_POD_NAMESPACE):10239,server-master-1=http://server-master-1.server-master.$(MY_POD_NAMESPACE):10239,server-master-2=http://server-master-2.server-master.$(MY_POD_NAMESPACE):10239"
+            - "--config=/conf/server-master.toml"
+  volumeClaimTemplates:
+    - metadata:
+        name: server-master-data
+      spec:
+        accessModes:
+          - ReadWriteOnce
+        resources:
+          requests:
+            storage: 2Gi
+    - metadata:
+        name: server-master-log
+      spec:
+        accessModes:
+          - ReadWriteOnce
+        resources:
+          requests:
+            storage: 1Gi
diff --git a/chaos/scripts/check-case.sh b/chaos/scripts/check-case.sh
new file mode 100755
index 00000000..d7cf7395
--- /dev/null
+++ b/chaos/scripts/check-case.sh
@@ -0,0 +1,29 @@
+#!/bin/bash
+
+completed=false
+for i in {1..20}; do
+	kubectl wait --for=condition=complete job/chaos-test-case --timeout=1m
+	if [ $? -eq 0 ]; then
+		completed=true
+		echo "chaos-test-case has completed"
+		break
+	else
+		echo "chaos-test-case has not completed" ${i}
+		kubectl get job chaos-test-case -o wide
+		if [ $? -ne 0 ]; then
+			echo "chaos-test-case job has been cleared"
+			break
+		fi
+		failed=$(kubectl get job chaos-test-case -o jsonpath={.status.failed})
+		if [[ $failed -gt 0 ]]; then
+			echo "chaos-test-case job has failed"
+			kubectl logs job.batch/chaos-test-case
+			kubectl describe jobs/chaos-test-case
+			break
+		fi
+	fi
+done
+
+if ! $completed; then
+	exit 1
+fi