OpenEBS cstore volume is very slow on delete and overwrite

11/11/2019

I have an OpenEBS setup with 3 data nodes and a default cstore storage class. Creation of a file works pretty good:

time dd if=/dev/urandom of=a.log bs=1M count=100
real    0m 0.53s

I can delete the file and create it again with the quite the same times. But when I rewrite onto the file it takes ages:

time dd if=/dev/urandom of=a.log bs=1M count=100
104857600 bytes (100.0MB) copied, 0.596577 seconds, 167.6MB/s
real    0m 0.59s

time dd if=/dev/urandom of=a.log bs=1M count=100
104857600 bytes (100.0MB) copied, 16.621222 seconds, 6.0MB/s
real    0m 16.62s

time dd if=/dev/urandom of=a.log bs=1M count=100
104857600 bytes (100.0MB) copied, 19.621924 seconds, 5.1MB/s
real    0m 19.62s

When I delete the a.log and write it again, then the ~167MS/s are back. Only writing onto an existing file takes so much time.

My problem is that i think this is why some of my applications (e.g. Databases) are too slow. Creation of a table in mysql took over 7sek.

Here is the spec of my testcluster:

# Create the OpenEBS namespace
apiVersion: v1
kind: Namespace
metadata:
  name: openebs
---
# Create Maya Service Account
apiVersion: v1
kind: ServiceAccount
metadata:
  name: openebs-maya-operator
  namespace: openebs
---
# Define Role that allows operations on K8s pods/deployments
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: openebs-maya-operator
rules:
- apiGroups: ["*"]
  resources: ["nodes", "nodes/proxy"]
  verbs: ["*"]
- apiGroups: ["*"]
  resources: ["namespaces", "services", "pods", "pods/exec", "deployments", "replicationcontrollers", "replicasets", "events", "endpoints", "configmaps", "secrets", "jobs", "cronjobs"]
  verbs: ["*"]
- apiGroups: ["*"]
  resources: ["statefulsets", "daemonsets"]
  verbs: ["*"]
- apiGroups: ["*"]
  resources: ["resourcequotas", "limitranges"]
  verbs: ["list", "watch"]
- apiGroups: ["*"]
  resources: ["ingresses", "horizontalpodautoscalers", "verticalpodautoscalers", "poddisruptionbudgets", "certificatesigningrequests"]
  verbs: ["list", "watch"]
- apiGroups: ["*"]
  resources: ["storageclasses", "persistentvolumeclaims", "persistentvolumes"]
  verbs: ["*"]
- apiGroups: ["volumesnapshot.external-storage.k8s.io"]
  resources: ["volumesnapshots", "volumesnapshotdatas"]
  verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
- apiGroups: ["apiextensions.k8s.io"]
  resources: ["customresourcedefinitions"]
  verbs: [ "get", "list", "create", "update", "delete", "patch"]
- apiGroups: ["*"]
  resources: [ "disks", "blockdevices", "blockdeviceclaims"]
  verbs: ["*" ]
- apiGroups: ["*"]
  resources: [ "cstorpoolclusters", "storagepoolclaims", "storagepoolclaims/finalizers", "cstorpoolclusters/finalizers", "storagepools"]
  verbs: ["*" ]
- apiGroups: ["*"]
  resources: [ "castemplates", "runtasks"]
  verbs: ["*" ]
- apiGroups: ["*"]
  resources: [ "cstorpools", "cstorpools/finalizers", "cstorvolumereplicas", "cstorvolumes", "cstorvolumeclaims"]
  verbs: ["*" ]
- apiGroups: ["*"]
  resources: [ "cstorpoolinstances", "cstorpoolinstances/finalizers"]
  verbs: ["*" ]
- apiGroups: ["*"]
  resources: [ "cstorbackups", "cstorrestores", "cstorcompletedbackups"]
  verbs: ["*" ]
- apiGroups: ["coordination.k8s.io"]
  resources: ["leases"]
  verbs: ["get", "watch", "list", "delete", "update", "create"]
- nonResourceURLs: ["/metrics"]
  verbs: ["get"]
- apiGroups: ["*"]
  resources: [ "upgradetasks"]
  verbs: ["*" ]
---
# Bind the Service Account with the Role Privileges.
kind: ClusterRoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: openebs-maya-operator
subjects:
- kind: ServiceAccount
  name: openebs-maya-operator
  namespace: openebs
- kind: User
  name: system:serviceaccount:default:default
  apiGroup: rbac.authorization.k8s.io
roleRef:
  kind: ClusterRole
  name: openebs-maya-operator
  apiGroup: rbac.authorization.k8s.io
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: maya-apiserver
  namespace: openebs
  labels:
    name: maya-apiserver
    openebs.io/component-name: maya-apiserver
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: maya-apiserver
      openebs.io/component-name: maya-apiserver
  replicas: 1
  strategy:
    type: Recreate
    rollingUpdate: null
  template:
    metadata:
      labels:
        name: maya-apiserver
        openebs.io/component-name: maya-apiserver
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
      - name: maya-apiserver
        imagePullPolicy: IfNotPresent
        image: quay.io/openebs/m-apiserver:1.3.0
        ports:
        - containerPort: 5656
        env:
        - name: OPENEBS_NAMESPACE
          valueFrom:
            fieldRef:
              fieldPath: metadata.namespace
        - name: OPENEBS_SERVICE_ACCOUNT
          valueFrom:
            fieldRef:
              fieldPath: spec.serviceAccountName
        - name: OPENEBS_MAYA_POD_NAME
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: OPENEBS_IO_CREATE_DEFAULT_STORAGE_CONFIG
          value: "true"
        - name: OPENEBS_IO_INSTALL_DEFAULT_CSTOR_SPARSE_POOL
          value: "true"
        - name: OPENEBS_IO_JIVA_CONTROLLER_IMAGE
          value: "quay.io/openebs/jiva:1.3.0"
        - name: OPENEBS_IO_JIVA_REPLICA_IMAGE
          value: "quay.io/openebs/jiva:1.3.0"
        - name: OPENEBS_IO_JIVA_REPLICA_COUNT
          value: "1"
        - name: OPENEBS_IO_CSTOR_TARGET_IMAGE
          value: "quay.io/openebs/cstor-istgt:1.3.0"
        - name: OPENEBS_IO_CSTOR_POOL_IMAGE
          value: "quay.io/openebs/cstor-pool:1.3.0"
        - name: OPENEBS_IO_CSTOR_POOL_MGMT_IMAGE
          value: "quay.io/openebs/cstor-pool-mgmt:1.3.0"
        - name: OPENEBS_IO_CSTOR_VOLUME_MGMT_IMAGE
          value: "quay.io/openebs/cstor-volume-mgmt:1.3.0"
        - name: OPENEBS_IO_VOLUME_MONITOR_IMAGE
          value: "quay.io/openebs/m-exporter:1.3.0"
        - name: OPENEBS_IO_CSTOR_POOL_EXPORTER_IMAGE
          value: "quay.io/openebs/m-exporter:1.3.0"
        - name: OPENEBS_IO_ENABLE_ANALYTICS
          value: "false"
        - name: OPENEBS_IO_INSTALLER_TYPE
          value: "openebs-operator"
        livenessProbe:
          exec:
            command:
            - /usr/local/bin/mayactl
            - version
          initialDelaySeconds: 30
          periodSeconds: 60
        readinessProbe:
          exec:
            command:
            - /usr/local/bin/mayactl
            - version
          initialDelaySeconds: 30
          periodSeconds: 60
---
apiVersion: v1
kind: Service
metadata:
  name: maya-apiserver-service
  namespace: openebs
  labels:
    openebs.io/component-name: maya-apiserver-svc
spec:
  ports:
  - name: api
    port: 5656
    protocol: TCP
    targetPort: 5656
  selector:
    name: maya-apiserver
  sessionAffinity: None
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: openebs-provisioner
  namespace: openebs
  labels:
    name: openebs-provisioner
    openebs.io/component-name: openebs-provisioner
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: openebs-provisioner
      openebs.io/component-name: openebs-provisioner
  replicas: 1
  strategy:
    type: Recreate
    rollingUpdate: null
  template:
    metadata:
      labels:
        name: openebs-provisioner
        openebs.io/component-name: openebs-provisioner
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
      - name: openebs-provisioner
        imagePullPolicy: IfNotPresent
        image: quay.io/openebs/openebs-k8s-provisioner:1.3.0
        env:
        - name: NODE_NAME
          valueFrom:
            fieldRef:
              fieldPath: spec.nodeName
        - name: OPENEBS_NAMESPACE
          valueFrom:
            fieldRef:
              fieldPath: metadata.namespace
        livenessProbe:
          exec:
            command:
            - pgrep
            - ".*openebs"
          initialDelaySeconds: 30
          periodSeconds: 60
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: openebs-snapshot-operator
  namespace: openebs
  labels:
    name: openebs-snapshot-operator
    openebs.io/component-name: openebs-snapshot-operator
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: openebs-snapshot-operator
      openebs.io/component-name: openebs-snapshot-operator
  replicas: 1
  strategy:
    type: Recreate
  template:
    metadata:
      labels:
        name: openebs-snapshot-operator
        openebs.io/component-name: openebs-snapshot-operator
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
        - name: snapshot-controller
          image: quay.io/openebs/snapshot-controller:1.3.0
          imagePullPolicy: IfNotPresent
          env:
          - name: OPENEBS_NAMESPACE
            valueFrom:
              fieldRef:
                fieldPath: metadata.namespace
          livenessProbe:
            exec:
              command:
              - pgrep
              - ".*controller"
            initialDelaySeconds: 30
            periodSeconds: 60
        - name: snapshot-provisioner
          image: quay.io/openebs/snapshot-provisioner:1.3.0
          imagePullPolicy: IfNotPresent
          env:
          - name: OPENEBS_NAMESPACE
            valueFrom:
              fieldRef:
                fieldPath: metadata.namespace
          livenessProbe:
            exec:
              command:
              - pgrep
              - ".*provisioner"
            initialDelaySeconds: 30
            periodSeconds: 60
---
# This is the node-disk-manager related config.
# It can be used to customize the disks probes and filters
apiVersion: v1
kind: ConfigMap
metadata:
  name: openebs-ndm-config
  namespace: openebs
  labels:
    openebs.io/component-name: ndm-config
data:
  node-disk-manager.config: |
    probeconfigs:
      - key: udev-probe
        name: udev probe
        state: true
      - key: seachest-probe
        name: seachest probe
        state: false
      - key: smart-probe
        name: smart probe
        state: true
    filterconfigs:
      - key: os-disk-exclude-filter
        name: os disk exclude filter
        state: true
        exclude: "/,/etc/hosts,/boot"
      - key: vendor-filter
        name: vendor filter
        state: true
        include: ""
        exclude: "CLOUDBYT,OpenEBS"
      - key: path-filter
        name: path filter
        state: true
        include: ""
        exclude: "loop,/dev/fd0,/dev/sr0,/dev/ram,/dev/dm-,/dev/md"
---
apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: openebs-ndm
  namespace: openebs
  labels:
    name: openebs-ndm
    openebs.io/component-name: ndm
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: openebs-ndm
      openebs.io/component-name: ndm
  updateStrategy:
    type: RollingUpdate
  template:
    metadata:
      labels:
        name: openebs-ndm
        openebs.io/component-name: ndm
        openebs.io/version: 1.3.0
    spec:
      nodeSelector:
        "openebs.io/nodegroup": "storage-node"
      serviceAccountName: openebs-maya-operator
      hostNetwork: true
      containers:
      - name: node-disk-manager
        image: quay.io/openebs/node-disk-manager-amd64:v0.4.3
        imagePullPolicy: Always
        securityContext:
          privileged: true
        volumeMounts:
        - name: config
          mountPath: /host/node-disk-manager.config
          subPath: node-disk-manager.config
          readOnly: true
        - name: udev
          mountPath: /run/udev
        - name: procmount
          mountPath: /host/proc
          readOnly: true
        - name: sparsepath
          mountPath: /var/openebs/sparse
        env:
        - name: NAMESPACE
          valueFrom:
            fieldRef:
              fieldPath: metadata.namespace
        - name: NODE_NAME
          valueFrom:
            fieldRef:
              fieldPath: spec.nodeName
        - name: SPARSE_FILE_DIR
          value: "/var/openebs/sparse"
        - name: SPARSE_FILE_SIZE
          value: "10737418240"
        - name: SPARSE_FILE_COUNT
          value: "3"
        livenessProbe:
          exec:
            command:
            - pgrep
            - ".*ndm"
          initialDelaySeconds: 30
          periodSeconds: 60
      volumes:
      - name: config
        configMap:
          name: openebs-ndm-config
      - name: udev
        hostPath:
          path: /run/udev
          type: Directory
      - name: procmount
        hostPath:
          path: /proc
          type: Directory
      - name: sparsepath
        hostPath:
          path: /var/openebs/sparse
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: openebs-ndm-operator
  namespace: openebs
  labels:
    name: openebs-ndm-operator
    openebs.io/component-name: ndm-operator
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: openebs-ndm-operator
      openebs.io/component-name: ndm-operator
  replicas: 1
  strategy:
    type: Recreate
  template:
    metadata:
      labels:
        name: openebs-ndm-operator
        openebs.io/component-name: ndm-operator
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
        - name: node-disk-operator
          image: quay.io/openebs/node-disk-operator-amd64:v0.4.3
          imagePullPolicy: Always
          readinessProbe:
            exec:
              command:
                - stat
                - /tmp/operator-sdk-ready
            initialDelaySeconds: 4
            periodSeconds: 10
            failureThreshold: 1
          env:
            - name: WATCH_NAMESPACE
              valueFrom:
                fieldRef:
                  fieldPath: metadata.namespace
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  fieldPath: metadata.name
            # the service account of the ndm-operator pod
            - name: SERVICE_ACCOUNT
              valueFrom:
                fieldRef:
                  fieldPath: spec.serviceAccountName
            - name: OPERATOR_NAME
              value: "node-disk-operator"
            - name: CLEANUP_JOB_IMAGE
              value: "quay.io/openebs/linux-utils:3.9"
---
apiVersion: v1
kind: Secret
metadata:
  name: admission-server-certs
  namespace: openebs
  labels:
    app: admission-webhook
    openebs.io/component-name: admission-webhook
type: Opaque
data:
  cert.pem: <...pem...>
  key.pem: <...pem...>
---
apiVersion: v1
kind: Service
metadata:
  name: admission-server-svc
  namespace: openebs
  labels:
    app: admission-webhook
    openebs.io/component-name: admission-webhook-svc
spec:
  ports:
  - port: 443
    targetPort: 443
  selector:
    app: admission-webhook
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: openebs-admission-server
  namespace: openebs
  labels:
    app: admission-webhook
    openebs.io/component-name: admission-webhook
    openebs.io/version: 1.3.0
spec:
  replicas: 1
  strategy:
    type: Recreate
    rollingUpdate: null
  selector:
    matchLabels:
      app: admission-webhook
  template:
    metadata:
      labels:
        app: admission-webhook
        openebs.io/component-name: admission-webhook
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
        - name: admission-webhook
          image: quay.io/openebs/admission-server:1.3.0
          imagePullPolicy: IfNotPresent
          args:
            - -tlsCertFile=/etc/webhook/certs/cert.pem
            - -tlsKeyFile=/etc/webhook/certs/key.pem
            - -alsologtostderr
            - -v=2
            - 2>&1
          volumeMounts:
            - name: webhook-certs
              mountPath: /etc/webhook/certs
              readOnly: true
      volumes:
        - name: webhook-certs
          secret:
            secretName: admission-server-certs
---
apiVersion: admissionregistration.k8s.io/v1beta1
kind: ValidatingWebhookConfiguration
metadata:
  name: validation-webhook-cfg
  labels:
    app: admission-webhook
    openebs.io/component-name: admission-webhook
webhooks:
  # failurePolicy Fail means that an error calling the webhook causes the admission to fail.
  - name: admission-webhook.openebs.io
    failurePolicy: Ignore
    clientConfig:
      service:
        name: admission-server-svc
        namespace: openebs
        path: "/validate"
      caBundle: <...ca..>
    rules:
      - operations: [ "CREATE", "DELETE" ]
        apiGroups: ["*"]
        apiVersions: ["*"]
        resources: ["persistentvolumeclaims"]
      - operations: [ "CREATE", "UPDATE" ]
        apiGroups: ["*"]
        apiVersions: ["*"]
        resources: ["cstorpoolclusters"]
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: openebs-localpv-provisioner
  namespace: openebs
  labels:
    name: openebs-localpv-provisioner
    openebs.io/component-name: openebs-localpv-provisioner
    openebs.io/version: 1.3.0
spec:
  selector:
    matchLabels:
      name: openebs-localpv-provisioner
      openebs.io/component-name: openebs-localpv-provisioner
  replicas: 1
  strategy:
    type: Recreate
  template:
    metadata:
      labels:
        name: openebs-localpv-provisioner
        openebs.io/component-name: openebs-localpv-provisioner
        openebs.io/version: 1.3.0
    spec:
      serviceAccountName: openebs-maya-operator
      containers:
      - name: openebs-provisioner-hostpath
        imagePullPolicy: Always
        image: quay.io/openebs/provisioner-localpv:1.3.0
        env:
        - name: NODE_NAME
          valueFrom:
            fieldRef:
              fieldPath: spec.nodeName
        - name: OPENEBS_NAMESPACE
          valueFrom:
            fieldRef:
              fieldPath: metadata.namespace
        - name: OPENEBS_IO_ENABLE_ANALYTICS
          value: "true"
        - name: OPENEBS_IO_INSTALLER_TYPE
          value: "openebs-operator"
        - name: OPENEBS_IO_HELPER_IMAGE
          value: "quay.io/openebs/openebs-tools:3.8"
        livenessProbe:
          exec:
            command:
            - pgrep
            - ".*localpv"
          initialDelaySeconds: 30
          periodSeconds: 60

I am on Kubernetes:

Client Version: version.Info{Major:"1", Minor:"16", GitVersion:"v1.16.2", GitCommit:"c97fe5036ef3df2967d086711e6c0c405941e14b", GitTreeState:"clean", BuildDate:"2019-10-15T19:18:23Z", GoVersion:"go1.12.10", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"16", GitVersion:"v1.16.2", GitCommit:"c97fe5036ef3df2967d086711e6c0c405941e14b", GitTreeState:"clean", BuildDate:"2019-10-15T19:09:08Z", GoVersion:"go1.12.10", Compiler:"gc", Platform:"linux/amd64"}

How can i investigate why it takes so long? What can me bring some more insides?

-- Peter Shaw
kubernetes
openebs

1 Answer

11/28/2019

Spoke to Peter over Slack. The following changes helped to improve the performance:

  • Increasing the queue size of the iSCSI Initiator
  • Increase the cpu from 2 to 6 for each of the three nodes
  • Create my own sc to use a ext3 instead of the default ext4.

With the above changes, the subsequent writes numbers were around 140-150 MB/s.

-- Kiran Mova
Source: StackOverflow