Deploy

Alluxio can be run on Kubernetes. This guide demonstrates how to run Alluxio on Kubernetes using the specification included in the Alluxio Docker image or helm.

Prerequisites

A Kubernetes cluster (version 1.11+) with Beta feature gate APIs enabled
- The Alluxio Helm chart which the Kubernetes resource specifications are built from supports Kubernetes version 1.11+.
- Beta feature gates are enabled by default for Kubernetes cluster installations
Cluster access to an Alluxio Docker image alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}. If using a private Docker registry, refer to the Kubernetes private image registry documentation.
Ensure the cluster's Kubernetes Network Policy allows for connectivity between applications (Alluxio clients) and the Alluxio Pods on the defined ports.

Basic Setup

This tutorial walks through a basic Alluxio setup on Kubernetes. Alluxio supports two methods of installation on Kubernetes: either using helm charts or using kubectl. When available, helm is the preferred way to install Alluxio. If helm is not available or if additional deployment customization is desired, kubectl can be used directly using native Kubernetes resource specifications.

Note: From Alluxio 2.3 on, Alluxio only supports helm 3. See how to migrate from helm 2 to 3 here.

(Optional) Copy the Alluxio Helm chart to a private Helm repository

The Alluxio Helm chart source code is located here. Alternatively, you can extract the Helm chart directory from the Alluxio Docker image:

$ id=$(docker create alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}:{{site.ALLUXIO_VERSION_STRING}})
$ docker cp $id:/opt/alluxio/integration/kubernetes/ - > kubernetes.tar
$ docker rm -v $id 1>/dev/null
$ tar -xvf kubernetes.tar
$ cd kubernetes/helm-chart/alluxio

(Optional) Provision a Persistent Volume

Depending on the configuration used to deploy Alluxio, you will likely need to provision Persistent Volumes.

Embedded Journal requires a Persistent Volume for each master Pod to be provisioned and is the preferred HA mechanism for Alluxio on Kubernetes. The volume, once claimed, is persisted across restarts of the master process.
When using the UFS Journal an Alluxio master can also be configured to use a persistent volume for storing the journal. If Alluxio is configured to use a UFS journal and with an external journal location like HDFS, the rest of this section can be skipped.
When Alluxio workers have short-circuit access, you may need to use Volumes to mount the domain socket to the workers.

There are multiple ways to create a PersistentVolume. This is an example which defines one with hostPath for the Alluxio Master journal:

# Name the file alluxio-master-journal-pv.yaml
kind: PersistentVolume
apiVersion: v1
metadata:
  name: alluxio-journal-0
  labels:
    type: local
spec:
  storageClassName: standard
  capacity:
    storage: 1Gi
  accessModes:
    - ReadWriteOnce
  hostPath:
    path: /tmp/alluxio-journal-0

Note:

By default each journal volume should be at least 1Gi, because each Alluxio master Pod will have one PersistentVolumeClaim that requests for 1Gi storage. You will see how to configure the journal size in later sections.
If this hostPath is not already present on the host, Kubernetes can be configured to create it. However the assigned user:group permissions may prevent the Alluxio masters & workers from accessing it. Please ensure the permissions are set to allow the pods to access the directory.
- See the Kubernetes volume docs for more details
- From Alluxio v2.1 on, Alluxio Docker containers will run as non-root user alluxio with UID 1000 and GID 1000 by default.

Then create the persistent volume with kubectl:

$ kubectl create -f alluxio-master-journal-pv.yaml

Deploy

helm

Prerequisites

A. Install Helm

You should have helm 3.X installed. You can install helm following instructions here.

B. A helm repo with the Alluxio helm chart must be available.

$ helm repo add alluxio-charts https://alluxio-charts.storage.googleapis.com/openSource/{{site.ALLUXIO_VERSION_STRING}}

Configuration

Once the helm repository is available, prepare the Alluxio configuration. The minimal configuration must contain the under storage address:

properties:
  alluxio.master.mount.table.root.ufs: "<under_storage_address>"

Note: The Alluxio under filesystem address MUST be modified. Any credentials MUST be modified.

To view the complete list of supported properties run the helm inspect command:

$ helm inspect values alluxio-charts/alluxio

The remainder of this section describes various configuration options with examples.

Example: Amazon S3 as the under store

To mount S3 at the root of Alluxio namespace specify all required properties as a key-value pair under properties.

properties:
  alluxio.master.mount.table.root.ufs: "s3a://<bucket>"
  alluxio.master.mount.table.root.option.s3a.accessKeyId: "<accessKey>"
  alluxio.master.mount.table.root.option.s3a.secretKey: "<secretKey>"

Example: Single Master and Journal in a Persistent Volume

The following configures UFS Journal with a persistent volume claim mounted locally to the master Pod at location /journal.

master:
  count: 1 # For multiMaster mode increase this to >1

journal:
  # [ Required values ]
  type: "UFS" # One of "UFS" or "EMBEDDED"
  folder: "/journal" # Master journal directory or equivalent storage path
  #
  # [ Conditionally required values ]
  #
  ## [ UFS-backed journal options ]
  ## - required when using a UFS-type journal (journal.type="UFS")
  ##
  ## ufsType is one of "local" or "HDFS"
  ## - "local" results in a PV being allocated to each Master Pod as the journal
  ## - "HDFS" results in no PV allocation, it is up to you to ensure you have
  ##   properly configured the required Alluxio properties for Alluxio to access
  ##   the HDFS URI designated as the journal folder
  ufsType: "local"
  #
  ## [ K8s volume options ]
  ## - required when using an EMBEDDED journal (journal.type="EMBEDDED")
  ## - required when using a local UFS journal (journal.type="UFS" and journal.ufsType="local")
  ##
  ## volumeType controls the type of journal volume.
  volumeType: persistentVolumeClaim # One of "persistentVolumeClaim" or "emptyDir"
  ## size sets the requested storage capacity for a persistentVolumeClaim,
  ## or the sizeLimit on an emptyDir PV.
  size: 1Gi
  ### Unique attributes to use when the journal is persistentVolumeClaim
  storageClass: "standard"
  accessModes:
    - ReadWriteOnce

Example: Single Master and Journal in an emptyDir Volume

The following configures UFS Journal with an emptyDir volume mounted locally to the master Pod at location /journal.

master:
  count: 1 # For multiMaster mode increase this to >1

journal:
  # [ Required values ]
  type: "UFS" # One of "UFS" or "EMBEDDED"
  folder: "/journal" # Master journal directory or equivalent storage path
  #
  # [ Conditionally required values ]
  #
  ## [ UFS-backed journal options ]
  ## - required when using a UFS-type journal (journal.type="UFS")
  ##
  ## ufsType is one of "local" or "HDFS"
  ## - "local" results in a PV being allocated to each Master Pod as the journal
  ## - "HDFS" results in no PV allocation, it is up to you to ensure you have
  ##   properly configured the required Alluxio properties for Alluxio to access
  ##   the HDFS URI designated as the journal folder
  ufsType: "local"
  #
  ## [ K8s volume options ]
  ## - required when using an EMBEDDED journal (journal.type="EMBEDDED")
  ## - required when using a local UFS journal (journal.type="UFS" and journal.ufsType="local")
  ##
  ## volumeType controls the type of journal volume.
  volumeType: emptyDir # One of "persistentVolumeClaim" or "emptyDir"
  ## size sets the requested storage capacity for a persistentVolumeClaim,
  ## or the sizeLimit on an emptyDir PV.
  size: 1Gi
  ### Unique attributes to use when the journal is emptyDir
  medium: ""

Note: An emptyDir volume has the same lifetime as the Pod. It is NOT a persistent storage. The Alluxio journal will be LOST when the Pod is restarted or rescheduled. Please only use this for experimental use cases. Check emptyDir for more details.

Example: HDFS as Journal

First create secrets for any configuration required by an HDFS client. These are mounted under /secrets.

$ kubectl create secret generic alluxio-hdfs-config --from-file=${HADOOP_CONF_DIR}/core-site.xml --from-file=${HADOOP_CONF_DIR}/hdfs-site.xml

journal:
  # [ Required values ]
  type: "UFS" # One of "UFS" or "EMBEDDED"
  folder: "hdfs://{$hostname}:{$hostport}/journal" # Master journal directory or equivalent storage path
  #
  # [ Conditionally required values ]
  #
  ## [ UFS-backed journal options ]
  ## - required when using a UFS-type journal (journal.type="UFS")
  ##
  ## ufsType is one of "local" or "HDFS"
  ## - "local" results in a PV being allocated to each Master Pod as the journal
  ## - "HDFS" results in no PV allocation, it is up to you to ensure you have
  ##   properly configured the required Alluxio properties for Alluxio to access
  ##   the HDFS URI designated as the journal folder
  ufsType: "HDFS"

properties:
  alluxio.master.mount.table.root.ufs: "hdfs://{$hostname}:{$hostport}/alluxio"
  alluxio.master.journal.ufs.option.alluxio.underfs.hdfs.configuration: "/secrets/hdfsConfig/core-site.xml:/secrets/hdfsConfig/hdfs-site.xml"

secrets:
  master:
    alluxio-hdfs-config: hdfsConfig
  worker:
    alluxio-hdfs-config: hdfsConfig

Example: Multi-master with Embedded Journal in Persistent Volumes

master:
  count: 3

  # [ Required values ]
  type: "EMBEDDED" # One of "UFS" or "EMBEDDED"
  folder: "/journal" # Master journal directory or equivalent storage path
  #
  # [ Conditionally required values ]
  #
  ## [ K8s volume options ]
  ## - required when using an EMBEDDED journal (journal.type="EMBEDDED")
  ## - required when using a local UFS journal (journal.type="UFS" and journal.ufsType="local")
  ##
  ## volumeType controls the type of journal volume.
  volumeType: persistentVolumeClaim # One of "persistentVolumeClaim" or "emptyDir"
  ## size sets the requested storage capacity for a persistentVolumeClaim,
  ## or the sizeLimit on an emptyDir PV.
  size: 1Gi
  ### Unique attributes to use when the journal is persistentVolumeClaim
  storageClass: "standard"
  accessModes:
    - ReadWriteOnce

Example: Multi-master with Embedded Journal in emptyDir Volumes

master:
  count: 3

journal:
  # [ Required values ]
  type: "EMBEDDED" # One of "UFS" or "EMBEDDED"
  folder: "/journal" # Master journal directory or equivalent storage path
  #
  # [ Conditionally required values ]
  #
  ## [ K8s volume options ]
  ## - required when using an EMBEDDED journal (journal.type="EMBEDDED")
  ## - required when using a local UFS journal (journal.type="UFS" and journal.ufsType="local")
  ##
  ## volumeType controls the type of journal volume.
  volumeType: emptyDir # One of "persistentVolumeClaim" or "emptyDir"
  ## size sets the requested storage capacity for a persistentVolumeClaim,
  ## or the sizeLimit on an emptyDir PV.
  size: 1Gi
  ### Unique attributes to use when the journal is emptyDir
  medium: ""

Example: HDFS as the under store

First create secrets for any configuration required by an HDFS client. These are mounted under /secrets.

$ kubectl create secret generic alluxio-hdfs-config --from-file=${HADOOP_CONF_DIR}/core-site.xml --from-file=${HADOOP_CONF_DIR}/hdfs-site.xml

properties:
  alluxio.master.mount.table.root.ufs: "hdfs://<ns>"
  alluxio.master.mount.table.root.option.alluxio.underfs.hdfs.configuration: "/secrets/hdfsConfig/core-site.xml:/secrets/hdfsConfig/hdfs-site.xml"
secrets:
  master:
    alluxio-hdfs-config: hdfsConfig
  worker:
    alluxio-hdfs-config: hdfsConfig

Example: Off-heap Metastore Management in Persistent Volumes

The following configuration creates a PersistentVolumeClaim for each Alluxio master Pod with the specified configuration and configures the Pod to use the volume for an on-disk RocksDB-based metastore.

properties:
  alluxio.master.metastore: ROCKS
  alluxio.master.metastore.dir: /metastore

metastore:
  volumeType: persistentVolumeClaim # Options: "persistentVolumeClaim" or "emptyDir"
  size: 1Gi
  mountPath: /metastore
  # Attributes to use when the metastore is persistentVolumeClaim
  storageClass: "standard"
  accessModes:
   - ReadWriteOnce

Example: Off-heap Metastore Management in emptyDir Volumes

The following configuration creates an emptyDir Volume for each Alluxio master Pod with the specified configuration and configures the Pod to use the volume for an on-disk RocksDB-based metastore.

properties:
  alluxio.master.metastore: ROCKS
  alluxio.master.metastore.dir: /metastore

metastore:
  volumeType: emptyDir # Options: "persistentVolumeClaim" or "emptyDir"
  size: 1Gi
  mountPath: /metastore
  # Attributes to use when the metastore is emptyDir
  medium: ""

Note: An emptyDir volume has the same lifetime as the Pod. It is NOT a persistent storage. The Alluxio metadata will be LOST when the Pod is restarted or rescheduled. Please only use this for experimental use cases. Check emptyDir for more details.

Example: Multiple Secrets

Multiple secrets can be mounted to both master and worker Pods. The format for the section for each Pod is <secretName>: <mountPath>

secrets:
  master:
    alluxio-hdfs-config: hdfsConfig
    alluxio-ceph-config: cephConfig
  worker:
    alluxio-hdfs-config: hdfsConfig
    alluxio-ceph-config: cephConfig

Examples: Alluxio Storage Management

Alluxio manages local storage, including memory, on the worker Pods. Multiple-Tier Storage can be configured using the following reference configurations.

There 3 supported volume type: hostPath, emptyDir and persistentVolumeClaim.

Memory Tier Only

tieredstore:
  levels:
  - level: 0
    mediumtype: MEM
    path: /dev/shm
    type: emptyDir
    high: 0.95
    low: 0.7

Memory and SSD Storage in Multiple-Tiers

tieredstore:
  levels:
  - level: 0
    mediumtype: MEM
    path: /dev/shm
    type: hostPath
    high: 0.95
    low: 0.7
  - level: 1
    mediumtype: SSD
    path: /ssd-disk
    type: hostPath
    high: 0.95
    low: 0.7

Note: If a hostPath file or directory is created at runtime, it can only be used by the root user. hostPath volumes do not have resource limits. You can either run Alluxio containers with root or make sure the local paths exist and are accessible to the user alluxio with UID and GID 1000. You can find more details here.

Memory and SSD Storage in Multiple-Tiers, using PVC

You can also use PVCs for each tier and provision PersistentVolume. For worker tiered storage please use either hostPath or local volume so that the worker will read and write locally to achieve the best performance.

tieredstore:
  levels:
  - level: 0
    mediumtype: MEM
    path: /dev/shm
    type: persistentVolumeClaim
    name: alluxio-mem
    quota: 1G
    high: 0.95
    low: 0.7
  - level: 1
    mediumtype: SSD
    path: /ssd-disk
    type: persistentVolumeClaim
    name: alluxio-ssd
    quota: 10G
    high: 0.95
    low: 0.7

Note: There is one PVC per tier. When the PVC is bound to a PV of type hostPath or local, each worker Pod will resolve to the local path on the Node. Please also note that a local volumes requires nodeAffinity and Pods using this volume can only run on the Nodes specified in the nodeAffinity rule of this volume. You can find more details here.

Memory and SSD Storage in a Single-Tier

You can also have multiple volumes on the same tier. This configuration will create one persistentVolumeClaim for each volume.

tieredstore:
  levels:
  - level: 0
    mediumtype: MEM,SSD
    path: /dev/shm,/alluxio-ssd
    type: persistentVolumeClaim
    name: alluxio-mem,alluxio-ssd
    quota: 1GB,10GB
    high: 0.95
    low: 0.7

Install

Once the configuration is finalized in a file named config.yaml, install as follows:

$ helm install alluxio -f config.yaml alluxio-charts/alluxio

In order to configure the Alluxio Master pod for use, you will need to format the Alluxio journal.

Format Journal

The master Pods in the StatefulSet use an initContainer to format the journal on startup. This initContainer is switched on by journal.format.runFormat=true. By default, the journal is not formatted when the master starts.

You can trigger the journal formatting by upgrading the existing helm deployment with journal.format.runFormat=true.

# Use the same config.yaml and switch on journal formatting
$ helm upgrade alluxio -f config.yaml --set journal.format.runFormat=true alluxio-charts/alluxio

Note: helm upgrade will re-create the master Pods.

Or you can trigger the journal formatting at deployment.

$ helm install alluxio -f config.yaml --set journal.format.runFormat=true alluxio-charts/alluxio

Note: From Alluxio v2.1 on, Alluxio Docker containers will run as non-root user alluxio with UID 1000 and GID 1000 by default. You should make sure the journal is formatted using the same user that the Alluxio master Pod runs as.

Configure Worker Volumes

Additional configuration is required for the Alluxio Worker pod to be ready for use. See the section for enabling worker short-circuit access.

Uninstall

Uninstall Alluxio as follows:

$ helm delete alluxio

kubectl

Choose the Sample YAML Template

First, extract the pre-templated Kubernetes specification YAMLs from the Alluxio docker image:

$ id=$(docker create alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}:{{site.ALLUXIO_VERSION_STRING}})
$ docker cp $id:/opt/alluxio/integration/kubernetes/ - > kubernetes.tar
$ docker rm -v $id 1>/dev/null
$ tar -xvf kubernetes.tar
$ cd kubernetes

The extracted directory contains a set of YAML templates generated from our Helm chart for common deployment scenarios in the sub-directories: singleMaster-localJournal, singleMaster-hdfsJournal, and multiMaster-embeddedJournal.

singleMaster means the templates generate 1 Alluxio master process, while multiMaster means 3. embedded and ufs are the 2 journal modes that Alluxio supports.

singleMaster-localJournal directory gives you the necessary Kubernetes ConfigMap, 1 Alluxio master process and a set of Alluxio workers. The Alluxio master writes journal to the journal volume requested by volumeClaimTemplates.
multiMaster-EmbeddedJournal directory gives you the Kubernetes ConfigMap, 3 Alluxio masters and a set of Alluxio workers. Each Alluxio master writes journal to its own journal volume requested by volumeClaimTemplates.
singleMaster-hdfsJournal directory gives you the Kubernetes ConfigMap, 1 Alluxio master with a set of workers. The journal is in a shared UFS location. In this template we use HDFS as the UFS.

For customized templated YAMLs, see the README for how to use helm-generate.sh. Otherwise you may manually write or modify YAML files as you see fit.

Configuration

Once the deployment option is chosen, copy the template from the desired sub-directory:

$ cp alluxio-configmap.yaml.template alluxio-configmap.yaml

Modify or add any configuration properties as required. The Alluxio under filesystem address MUST be modified. Any credentials MUST be modified. Add to ALLUXIO_JAVA_OPTS:

-Dalluxio.master.mount.table.root.ufs=<under_storage_address>

Note:

Replace <under_storage_address> with the appropriate URI, for example s3://my-bucket. If using an under storage which requires credentials be sure to specify those as well.
When running Alluxio with host networking, the ports assigned to Alluxio services must not be occupied beforehand.

Create a ConfigMap.

$ kubectl create -f alluxio-configmap.yaml

Install

Prepare the Specification. Prepare the Alluxio deployment specs from the templates. Modify any parameters required, such as location of the Docker image, and CPU and memory requirements for Pods.

For the master(s), create the Service and StatefulSet:

$ mv master/alluxio-master-service.yaml.template master/alluxio-master-service.yaml
$ mv master/alluxio-master-statefulset.yaml.template master/alluxio-master-statefulset.yaml

Note: alluxio-master-statefulset.yaml uses volumeClaimTemplates to define the journal volume for each master if it needs one.

For the workers, create the DaemonSet:

$ mv worker/alluxio-worker-daemonset.yaml.template worker/alluxio-worker-daemonset.yaml

Note: Please make sure that the version of the Kubernetes specification matches the version of the Alluxio Docker image being used.

(Optional) Remote Storage Access

Additional steps may be required when Alluxio is connecting to storage hosts outside the Kubernetes cluster it is deployed on. The remainder of this section explains how to configure the connection to a remote HDFS accessible but not managed by Kubernetes.

Step 1: Add hostAliases for your HDFS connection. Kubernetes Pods don't recognize network hostnames that are not managed by Kubernetes (not a Kubernetes Service), unless if specified by hostAliases.

For example if your HDFS service can be reached at hdfs://<namenode>:9000 where <namenode> is a hostname, you will need to add hostAliases in the spec for all Alluxio Pods creating a map from hostnames to IP addresses.

spec:
  hostAliases:
  - ip: "<namenode_ip>"
    hostnames:
    - "<namenode>"

For the case of a StatefulSet or DaemonSet as used in alluxio-master-statefulset.yaml.template and alluxio-worker-daemonset.yaml.template, hostAliases section should be added to each section of spec.template.spec like below.

kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  ...
  serviceName: "alluxio-master"
  replicas: 1
  template:
    metadata:
      labels:
        app: alluxio-master
    spec:
      hostAliases:
      - ip: "ip for hdfs-host"
        hostnames:
        - "hdfs-host"

Step 2: Create Kubernetes Secret for HDFS configuration files. Run the following command to create a Kubernetes Secret for the HDFS client configuration.

kubectl create secret generic alluxio-hdfs-config --from-file=${HADOOP_CONF_DIR}/core-site.xml --from-file=${HADOOP_CONF_DIR}/hdfs-site.xml

These two configuration files are referred in alluxio-master-statefulset.yaml and alluxio-worker-daemonset.yaml. Alluxio processes need the HDFS configuration files to connect, and the location of these files in the container is controlled by property alluxio.underfs.hdfs.configuration.

Step 3: Modify alluxio-configmap.yaml.template. Now that your Pods know how to talk to your HDFS service, update alluxio.master.journal.folder and alluxio.master.mount.table.root.ufs to point to the desired HDFS destination.

Once all the pre-requisites and configuration have been setup, deploy Alluxio.

$ kubectl create -f ./master/
$ kubectl create -f ./worker/

In order to configure the Alluxio Master pod for use, you will need to format the Alluxio journal.

Format Journal

You can manually add an initContainer to format the journal on Pod creation time. This initContainer will run alluxio formatJournal when the Pod is created and formats the journal.

- name: journal-format
  image: alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}:{{site.ALLUXIO_VERSION_STRING}}
  imagePullPolicy: IfNotPresent
  securityContext:
    runAsUser: 1000
  command: ["alluxio","formatJournal"]
  envFrom:
    - configMapRef:
      name: alluxio-config
  volumeMounts:
    - name: alluxio-journal
      mountPath: /journal

Configure Worker Volumes

Additional configuration is required for the Alluxio Worker pod to be ready for use. See the section for enabling worker short-circuit access.

Upgrade

This section will go over how to upgrade Alluxio in your Kubernetes cluster with kubectl.

Upgrading Alluxio

Step 1: Upgrade the docker image version tag

Each released Alluxio version will have the corresponding docker image released on dockerhub.

You should update the image field of all the Alluxio containers to use the target version tag. Tag latest will point to the latest stable version.

For example, if you want to upgrade Alluxio to the latest stable version, update the containers as below:

containers:
- name: alluxio-master
  image: alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}:latest
  imagePullPolicy: IfNotPresent
  ...
- name: alluxio-job-master
  image: alluxio/{{site.ALLUXIO_DOCKER_IMAGE}}:latest
  imagePullPolicy: IfNotPresent
  ...

Step 2: Stop running Alluxio master and worker Pods

Kill all running Alluxio worker Pods by deleting its DaemonSet.

$ kubectl delete daemonset -l app=alluxio

Then kill all running Alluxio master Pods by killing each StatefulSet and each Service with label app=alluxio.

$ kubectl delete service -l app=alluxio
$ kubectl delete statefulset -l app=alluxio

Make sure all the Pods have been terminated before you move on to the next step.

Step 3: Format journal and Alluxio storage if necessary

Check the Alluxio upgrade guide page for whether the Alluxio master journal has to be formatted. If no format is needed, you are ready to skip the rest of this section and move on to restart all Alluxio master and worker Pods.

You can follow formatting journal with kubectl to format the Alluxio journals.

If you are running Alluxio workers with tiered storage, and you have Persistent Volumes configured for Alluxio, the storage should be cleaned up too. You should delete and recreate the Persistent Volumes.

Once all the journals and Alluxio storage have been formatted, you are ready to restart the Alluxio master and worker Pods.

Step 4: Restart Alluxio master and worker Pods

Now that Alluxio masters and worker containers all use your desired version. Restart them to let it take effect.

Now restart the Alluxio master and worker Pods from the YAML files.

$ kubectl create -f ./master/
$ kubectl create -f ./worker/

Step 5: Verify the Alluxio master and worker Pods are back up

You should verify the Alluxio Pods are back up in Running status.

# You should see all Alluxio master and worker Pods
$ kubectl get pods

You can do more comprehensive verification following Verify Alluxio.

Uninstall

Uninstall Alluxio as follows:

$ kubectl delete -f ./worker/
$ kubectl delete -f ./master/
$ kubectl delete configmap alluxio-config

Note: This will delete all resources under ./master/ and ./worker/. Be careful if you have persistent volumes or other important resources you want to keep under those directories.

Verify

If using persistent volumes, the status of the volume(s) should change to CLAIMED and the status of the volume claims should be BOUNDED. You can validate the status of your PersistentVolume and PersistentVolumeClaims using the follow kubectl commands:

$ kubectl get pv
$ kubectl get pvc

If you have unbound PersistentVolumeClaims, please ensure you have provisioned matching PersistentVolumes. See "(Optional) Provision a Persistent Volume" in Basic Setup.

Once ready, access the Alluxio CLI from the master Pod and run basic I/O tests.

$ kubectl exec -ti alluxio-master-0 -- /bin/bash

From the master Pod, execute the following:

$ alluxio runTests

Access the Web UI

The Alluxio UI can be accessed from outside the kubernetes cluster using port forwarding.

$ kubectl port-forward alluxio-master-$i <local-port>:19999

The command above allocates a port on the local node <local-port> and forwards traffic on <local-port> to port 19999 of pod alluxio-master-$i. The pod alluxio-master-$i does NOT have to be on the node you are running this command.

Note: i=0 for the first master Pod. When running multiple masters, forward port for each master. Only the primary master serves the Web UI.

For example, you are on a node with hostname master-node-1 and you would like to serve the Alluxio master web UI for alluxio-master-0 on master-node-1:8080. Here's the command you can run:

[alice@master-node-1 ~]$ kubectl port-forward --address 0.0.0.0 pods/alluxio-master-0 8080:19999

This forwards the local port master-node-1:8080 to the port on the Pod alluxio-master-0:19999. The Pod alluxio-master-0 does NOT need to be running on master-node-1.

You will see messages like below when there are incoming connections.

[alice@master-node-1 ~]$ kubectl port-forward --address 0.0.0.0 alluxio-master-0 8080:19999
Forwarding from 0.0.0.0:8080 -> 19999
Handling connection for 8080
Handling connection for 8080
Handling connection for 8080
Handling connection for 8080

You can terminate the process to stop the port forwarding, with either Ctrl + C or kill.

For more information about K8s port-forward see the K8s doc.

Advanced Setup

Enable Short-circuit Access

Short-circuit access enables clients to perform read and write operations directly against the worker bypassing the networking interface. For performance-critical applications it is recommended to enable short-circuit operations against Alluxio because it can increase a client's read and write throughput when co-located with an Alluxio worker.

This feature is enabled by default (see next section to disable this feature), however requires extra configuration to work properly in Kubernetes environments.

There are two modes for using short-circuit.

Option1: Use local mode

In this mode, the Alluxio client and local Alluxio worker recognize each other if the client hostname matches the worker hostname. This is called Hostname Introspection. In this mode, the Alluxio client and local Alluxio worker share the tiered storage of Alluxio worker.

helm

You can use local policy by setting the properties as below:

shortCircuit:
  enabled: true
  policy: local

kubectl

In your alluxio-configmap.yaml you should add the following properties to ALLUXIO_WORKER_JAVA_OPTS:

-Dalluxio.user.short.circuit.enabled=true \
-Dalluxio.worker.data.server.domain.socket.as.uuid=false

Also you should remove the property -Dalluxio.worker.data.server.domain.socket.address.

Option2: Use uuid (default)

This is the default policy used for short-circuit in Kubernetes.

If the client or worker container is using virtual networking, their hostnames may not match. In such a scenario, set the following property to use filesystem inspection to enable short-circuit operations and make sure the client container mounts the directory specified as the domain socket path. Short-circuit writes are then enabled if the worker UUID is located on the client filesystem.

Domain Socket Path. The domain socket is a volume which should be mounted on:

All Alluxio workers
All application containers which intend to read/write through Alluxio

This domain socket volume can be either a PersistentVolumeClaim or a hostPath Volume.

Use PersistentVolumeClaim. By default, this domain socket volume is a PersistentVolumeClaim. You need to provision a PersistentVolume to this PersistentVolumeClaim. And this PersistentVolume should be either local or hostPath.

helm

You can use uuid policy by setting the properties as below:

# These are the default configurations
shortCircuit:
  enabled: true
  policy: uuid
  size: 1Mi
  # volumeType controls the type of shortCircuit volume.
  # It can be "persistentVolumeClaim" or "hostPath"
  volumeType: persistentVolumeClaim
  # Attributes to use if the domain socket volume is PVC
  pvcName: alluxio-worker-domain-socket
  accessModes:
    - ReadWriteOnce
  storageClass: standard

The field shortCircuit.pvcName defines the name of the PersistentVolumeClaim for domain socket. This PVC will be created as part of helm install.

kubectl

You should verify the following properties in ALLUXIO_WORKER_JAVA_OPTS. Actually they are set to these values by default:

-Dalluxio.worker.data.server.domain.socket.address=/opt/domain -Dalluxio.worker.data.server.domain.socket.as.uuid=true

Also you should make sure the worker Pods have domain socket defined in the volumes, and all relevant containers have the domain socket volume mounted. The domain socket volume is defined as below by default:

volumes:
  - name: alluxio-domain
    persistentVolumeClaim:
      claimName: "alluxio-worker-domain-socket"

Note: Compute application containers MUST mount the domain socket volume to the same path (/opt/domain) as configured for the Alluxio workers.

The PersistenceVolumeClaim is defined in worker/alluxio-worker-pvc.yaml.template.

Use hostPath Volume. You can also directly define the workers to use a hostPath Volume for domain socket.

helm

You can switch to directly use a hostPath volume for the domain socket. This is done by changing the shortCircuit.volumeType field to hostPath. Note that you also need to define the path to use for the hostPath volume.

shortCircuit:
  enabled: true
  policy: uuid
  size: 1Mi
  # volumeType controls the type of shortCircuit volume.
  # It can be "persistentVolumeClaim" or "hostPath"
  volumeType: hostPath
  # Attributes to use if the domain socket volume is hostPath
  hostPath: "/tmp/alluxio-domain" # The hostPath directory to use

kubectl

You should verify the properties in ALLUXIO_WORKER_JAVA_OPTS in the same way as using PersistentVolumeClaim.

volumes:
  - name: alluxio-domain
    hostPath:
      path: /tmp/alluxio-domain
      type: DirectoryOrCreate

Note: Compute application containers MUST mount the domain socket volume to the same path (/opt/domain) as configured for the Alluxio workers.

Verify Short-circuit Operations

To verify short-circuit reads and writes monitor the metrics displayed under:

the metrics tab of the web UI as Domain Socket Alluxio Read and Domain Socket Alluxio Write
or, the metrics json as cluster.BytesReadDomain and cluster.BytesWrittenDomain
or, the fsadmin metrics CLI as Short-circuit Read (Domain Socket) and Alluxio Write (Domain Socket)

Disable Short-Circuit Operations

To disable short-circuit operations, the operation depends on how you deploy Alluxio.

Note: As mentioned, disabling short-circuit access for Alluxio workers will result in worse I/O throughput

helm

You can disable short circuit by setting the properties as below:

shortCircuit:
  enabled: false

kubectl

You should set the property alluxio.user.short.circuit.enabled to false in your ALLUXIO_WORKER_JAVA_OPTS.

-Dalluxio.user.short.circuit.enabled=false

You should also manually remove the volume alluxio-domain from volumes of the Pod definition and volumeMounts of each container if existing.

Enable remote logging

Alluxio supports a centralized log server that collects logs for all Alluxio processes. You can find the specific section at Remote logging. This can be enabled on K8s too, so that all Alluxio pods will send logs to this log server.

helm

Step 1: Configure the log server

By default, the Alluxio remote log server is not started. You can enable the log server by configuring the following properties:

logserver:
  enabled: true

If you are just testing and it is okay to discard logs, you can use an emptyDir to store the logs in the log server.

logserver:
  enabled: true
  # volumeType controls the type of log volume.
  # It can be "persistentVolumeClaim" or "hostPath" or "emptyDir"
  volumeType: emptyDir
  # Attributes to use when the log volume is emptyDir
  medium: ""
  size: 4Gi

For a production environment, you should always persist the logs with a Persistent Volume. When you specify the logserver.volumeType to be persistentVolumeClaim, the Helm Chart will create a PVC. If you are not using dynamic provisioning for PVs, you will need to manually create the PV. Remember to make sure the selectors for PVC and PV match with each other.

logserver:
  enabled: true
  # volumeType controls the type of log volume.
  # It can be "persistentVolumeClaim" or "hostPath" or "emptyDir"
  volumeType: persistentVolumeClaim
  # Attributes to use if the log volume is PVC
  pvcName: alluxio-logserver-logs
  accessModes:
    - ReadWriteOnce
  storageClass: standard
  # If you are dynamically provisioning PVs, the selector on the PVC should be empty.
  # Ref: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#class-1
  selector: {}
  # If you are manually allocating PV for the logserver,
  # it is recommended to use selectors to make sure the PV and PVC match as expected.
  # You can specify selectors like below:
  # Example:
  # selector:
  #   matchLabels:
  #     role: alluxio-logserver
  #     app: alluxio
  #     chart: alluxio-<chart version>
  #     release: alluxio
  #     heritage: Helm
  #     dc: data-center-1
  #     region: us-east

Step 2: Helm install with the updated configuration

When you enable the remote log server, it will be managed by a K8s Deployment. If you specify the volume type to be persistentVolumeClaim, a PVC will be created and mounted. You will need to provision a PV for the PVC. Then there will be a Service created for the Deployment, which all other Alluxio pods send logs to.

kubectl

Step 1: Configure log server location with environment variables

Add ALLUXIO_LOGSERVER_HOSTNAME and ALLUXIO_LOGSERVER_PORT properties to the configmap.

apiVersion: v1
kind: ConfigMap
metadata:
  ..omitted
data:
  ..omitted
  ALLUXIO_LOGSERVER_HOSTNAME: alluxio-logserver
  ALLUXIO_LOGSERVER_PORT: "45600"

Note: The value for ALLUXIO_LOGSERVER_PORT must be a string or kubectl will fail to read it.

Step 2: Configure and start log server

In the sample YAML directory (e.g. singleMaster-localJournal), the logserver/ directory contains all resources for the log server, including a Deployment, a Service and a PVC if needed.

First you can prepare the YAML file and configure what volume to use for the Deployment.

$ cp logserver/alluxio-logserver-deployment.yaml.template logserver/alluxio-logserver-deployment.yaml

If you are testing and it is okay to discard logs, you can use an emptyDir for the volume like below:

  volumes:      
  - name: alluxio-logs
    emptyDir:
      medium: 
      sizeLimit: "4Gi"

And the volume should be mounted to the log server container at /opt/alluxio/logs.

  volumeMounts:
  - name: alluxio-logs
    mountPath: /opt/alluxio/logs

For a production environment, you should always persist the logs with a Persistent Volume.

  volumes:      
  - name: alluxio-logs
    persistentVolumeClaim:
      claimName: "alluxio-logserver-logs"

There is also a YAML template for PVC alluxio-logserver-logs.

$ cp logserver/alluxio-logserver-pvc.yaml.template logserver/alluxio-logserver-pvc.yaml

You can further configure the resource and selector for the PVC, according to your environment.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: alluxio-logserver-logs
  ..omitted
spec:
  volumeMode: Filesystem
  resources:
    requests:
      storage: 4Gi
  storageClassName: standard
  accessModes:
    - ReadWriteOnce
  # If you are using dynamic provisioning, leave the selector empty.
  selector: {}
  # If you are manually allocating PV for the logserver,
  # it is recommended to use selectors to make sure the PV and PVC match as expected.
  # You can specify selectors like below:
  # Example:
  # selector:
  #   matchLabels:
  #     role: alluxio-logserver
  #     app: alluxio
  #     chart: alluxio-<chart version>
  #     release: alluxio
  #     heritage: Helm
  #     dc: data-center-1
  #     region: us-east

Create the PVC when you are ready.

$ kubectl create -f alluxio-logserver-pvc.yaml

(Optional) If you are not using dynamic provisioning, you need to prepare the PV yourself. Remember to make sure the selectors on the PVC and PV match with each other.

After you configure the volume in the Deployment, you can go ahead to create it.

$ kubectl create -f alluxio-logserver-deployment.yaml

There is also a Service associated to the Deployment.

$ cp logserver/alluxio-logserver-service.yaml.template logserver/alluxio-logserver-service.yaml
$ kubectl create -f logserver/alluxio-logserver-service.yaml

Step 3: Restart other Alluxio pods

You need to restart your other Alluxio pods (masters, workers, FUSE etc) so they capture the updated environment variables and send logs to the remote log server.

Verify log server

You can go into the log server pod and verify the logs exist.

$ kubectl exec -it <logserver-pod-name> -- bash
# In the logserver pod
bash-4.4$ pwd
/opt/alluxio
# You should see logs collected from other Alluxio pods
bash-4.4$ ls -al logs
total 16
drwxrwsr-x    4 1001     bin           4096 Jan 12 03:14 .
drwxr-xr-x    1 alluxio  alluxio         18 Jan 12 02:38 ..
drwxr-sr-x    2 alluxio  bin           4096 Jan 12 03:14 job_master
-rw-r--r--    1 alluxio  bin            600 Jan 12 03:14 logserver.log
drwxr-sr-x    2 alluxio  bin           4096 Jan 12 03:14 master
drwxr-sr-x    2 alluxio  bin           4096 Jan 12 03:14 worker
drwxr-sr-x    2 alluxio  bin           4096 Jan 12 03:14 job_worker

POSIX API

Once Alluxio is deployed on Kubernetes, there are multiple ways in which a client application can connect to it. For applications using the POSIX API, application containers can simply mount the Alluxio FileSystem.

FUSE daemon

One way to use the POSIX API is to deploy the Alluxio FUSE daemon, creating pods running Alluxio Fuse processes at deployment time. The Fuse processes are long-running.

helm

You can deploy the FUSE daemon by configuring the following properties:

fuse:
  enabled: true

To modify the default Fuse mount configuration, one can set

mountPath: The container path to be mounted. Default to /mnt/alluxio-fuse
alluxioPath: The alluxio path to be mounted to container mountPath. Default to /
mountOptions: The Fuse mount options. Default to allow_other. See Fuse mount options for more details.

fuse:
  enabled: true
  clientEnabled: true
  mountPath: /mnt/alluxio-fuse
  alluxioPath: /
  mountOptions: allow_other

Then follow the steps to install Alluxio with helm here.

If Alluxio has already been deployed with helm and now you want to enable FUSE, you use helm upgrade to add the FUSE daemons.

$ helm upgrade alluxio -f config.yaml \
  --set fuse.enabled=true \
  alluxio-charts/alluxio

Advanced POSIX API Configuration

Alluxio fuse/client configuration:

properties:
  alluxio.user.metadata.cache.enabled: true
  alluxio.user.metadata.cache.expiration.time: 2day
  alluxio.user.metadata.cache.max.size: "1000000"
  alluxio.user.direct.memory.io.enabled: true
  alluxio.fuse.logging.threshold: 1000ms

Alluxio fuse java opts

fuse:
  jvmOptions: " -Xmx16G -Xms16G -XX:MaxDirectMemorySize=32g"

Alluxio fuse mount options

fuse:
  mountOptions: direct_io,ro,max_read=131072,attr_timeout=7200,entry_timeout=7200

Alluxio fuse environment variables

fuse:
  env:
    MAX_IDLE_THREADS: "64"

POSIX API docs provides more details about how to configure Alluxio POSIX API.

kubectl

$ cp alluxio-fuse.yaml.template alluxio-fuse.yaml
$ kubectl create -f alluxio-fuse.yaml

Note:

The container running the Alluxio FUSE daemon must have the securityContext.privileged=true with SYS_ADMIN capabilities. Application containers that require Alluxio access do not need this privilege.
Application containers can run on any Docker image.

Then data can then be accessed inside the application container under /mnt/alluxio-fuse.

Advanced POSIX API Configuration

Alluxio fuse/client java opts can be set in alluxio-configmap.yaml:

  ALLUXIO_FUSE_JAVA_OPTS: |-
    -Dalluxio.fuse.mount.point=/mnt/alluxio-fuse 
    -Dalluxio.fuse.mount.alluxio.path=/ 
    -Dalluxio.fuse.mount.options=direct_io,max_read=131072,entry_timeout=7200,attr_timeout=7200 
    -Dalluxio.user.hostname=${ALLUXIO_CLIENT_HOSTNAME} 
    -Dalluxio.user.metadata.cache.enabled=true 
    -Dalluxio.user.metadata.cache.expiration.time=40min 
    -Dalluxio.user.metadata.cache.max.size=10000000 
    -Dalluxio.user.logging.threshold=1000ms 
    -Dalluxio.fuse.logging.threshold=1000ms

Note that if Alluxio Worker and Alluxio Fuse is co-located in the same node, Alluxio fuse can read from the worker storage directly to improve read performance. In this case, Alluxio Fuse need to know about the worker storage information. This is why worker storage configuration is set in ALLUXIO_JAVA_OPTS shared by all Alluxio containers.

Alluxio fuse environment variables can be set in fuse/alluxio-fuse.yaml:

containers:
  - name: alluxio-fuse
    env:
      - name: "MAX_IDLE_THREADS"
        value: "64"

POSIX API docs provides more details about how to configure Alluxio POSIX API.

To access data in Alluxio inside application containers, simply mount Alluxio with a hostPath mount of location /mnt/alluxio-fuse.

kubectl

Below is a sample nginx pod that is able to access data from Alluxio under /mnt/alluxio-fuse inside the pod.

apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  containers:
  - image: nginx
    name: nginx
    ports:
    - containerPort: 80
      protocol: TCP
    volumeMounts:
    - name: alluxio-fuse-mount
      mountPath: /mnt/alluxio-fuse 
  volumes:
  - name: alluxio-fuse-mount
    hostPath:
      path: /mnt/alluxio-fuse
      type: Directory

CSI

Other than using Alluxio FUSE daemon, you could also use CSI to mount the Alluxio FileSystem into application containers. Unlike Fuse daemon which is a long-running process, the Fuse pod launched by CSI has the same life cycle as the application pods who mount Alluxio as a volume. Fuse pod is automatically launched when an application pod mounts Alluxio inside itself, and automatically terminated when such application pods are terminated.

In order to use CSI, you need a Kubernetes cluster with version at least 1.17, with RBAC enabled in API Server.

Step 1: Customize configurations

You can either use the default CSI configurations provided in here under the csi section, or you can customize them to make it suitable for your workload. Here are some common properties that you can customize:

property name

Description

alluxioPath

The path in Alluxio which will be mounted

mountInPod

Set to true to launch Fuse process in an alluxio-fuse pod. Otherwise in the same container as nodeserver

mountPath

The path that Alluxio will be mounted to in the application container

mountOptions

Alluxio Fuse mount options

Step 2: Deploy CSI services You can use Helm to start the Alluxio CSI components with Alluxio cluster, or kubectl to create the resources manually, or parts from Helm and parts manually.

helm

To start the CSI components via helm chart, set the following values in your helm configuration file:

csi:
  enabled: true
  clientEnabled: true

Related Alluxio CSI-related services will be started along with Alluxio cluster.

kubectl

Modify or add any configuration properties inside values.yaml, then please use helm-generate.sh (see here for usage) to generate related templates. All CSI related templates will be under ${ALLUXIO_HOME}/integration/kubernetes/csi.

$ mv alluxio-csi-controller-rbac.yaml.template alluxio-csi-controller-rbac.yaml
$ mv alluxio-csi-controller.yaml.template alluxio-csi-controller.yaml
$ mv alluxio-csi-driver.yaml.template alluxio-csi-driver.yaml
$ mv alluxio-csi-fuse-configmap.yaml.template alluxio-csi-fuse-configmap.yaml
$ mv alluxio-csi-nodeplugin.yaml.template alluxio-csi-nodeplugin.yaml

Then run

$ kubectl apply -f alluxio-csi-controller-rbac.yaml -f alluxio-csi-controller.yaml -f alluxio-csi-driver.yaml -f alluxio-csi-fuse-configmap.yaml -f alluxio-csi-nodeplugin.yaml

to deploy CSI-related services.

Step 3: Provisioning

We provide both templates for k8s dynamic provisioning and static provisioning. Please choose the suitable provisioning methods according to your use case. You can refer to Persistent Volumes | Kubernetes and Dynamic Volume Provisioning | Kubernetes to get more details.

helm

Both dynamic provisioning and static provisioning resources are created via Helm chart. If you need additional resources, you need to create them manaully through kubectl.

kubectl

Persistent Volumes

For static provisioning, we generate two template files: alluxio-pv.yaml.template and alluxio-pvc-static.yaml.template. You can modify these two files based on your needs, then create the respective yaml files.

$ mv alluxio-pv.yaml.template alluxio-pv.yaml
$ mv alluxio-pvc-static.yaml.template alluxio-pvc-static.yaml

Then run

$ kubectl apply -f alluxio-pv.yaml
$ kubectl apply -f alluxio-pvc-static.yaml

to deploy the resources.

Note: If mountInPod is set to true, in alluxio-pv.yaml, the value of spec.csi.volumeHandle needs to be unique for CSI to identify different volumes. If the values of volumeHundle of two PVs are the same, CSI would regard them as the same volume, and thus may not launch Fuse pod, affecting the business pods.

Dynamic Volume Provisioning

For dynamic provisioning, we generate two template files: alluxio-storage-class.yaml.template and alluxio-pvc.yaml.template. You can modify these two files based on your needs, then create the respective yaml files.

$ mv alluxio-storage-class.yaml.template alluxio-storage-class.yaml
$ mv alluxio-pvc.yaml.template alluxio-pvc.yaml

Then run

$ kubectl apply -f alluxio-storage-class.yaml
$ kubectl apply -f alluxio-pvc.yaml

to deploy the resources.

Step 4: Deploy applications

Now you can put the PVC name in your application pod spec to use the Alluxio FileSystem.

Example

Below is a sample nginx pod that is able to access data from Alluxio under /data inside the pod.

apiVersion: v1
kind: Pod
metadata:
  name: nginx
spec:
  containers:
    - image: nginx
      imagePullPolicy: Always
      name: nginx
      ports:
        - containerPort: 80
          protocol: TCP
      volumeMounts:
        - mountPath: /data
          name: alluxio-pvc
  volumes:
    - name: alluxio-pvc
      persistentVolumeClaim:
        claimName: alluxio-pvc

For more information on how to configure a pod to use a persistent volume for storage in Kubernetes, please refer to here.

Start Alluxio Proxy server

One can use either helm or kubectl to set up Alluxio proxy servers inside a kubernetes cluster.

kubectl

By default, proxy uses daemonset, so every node would spawn a pod running proxy server. To start proxy server when deploying Alluxio, set the following property in the helm chart configuration file:

proxy:
  enabled: true

kubectl

Configuration

In the sample YAML directory (e.g. singleMaster-localJournal), the proxy/ directory contains the daemonset configuration file for the proxy. Users can modify the configurations according to the needs.

Deploy proxy server

Run the following commands to deploy proxy daemonset:

$ cp alluxio-proxy-daemonset.yaml.template alluxio-proxy-daemonset.yaml
$ kubectl create -f alluxio-proxy-daemonset.yaml

Stop proxy server

Run the following command to stop proxy daemonset:

$ kubectl delete daemonset alluxio-proxy

Toggle Master or Worker in Helm chart

In use cases where you wish to install Alluxio masters and workers separately with the Helm chart, use the following respective toggles:

master:
  enabled: false

worker:
  enabled: false

Kubernetes Configuration Options

The following options are provided in our Helm chart as additional parameters for experienced Kubernetes users.

ServiceAccounts

By default Kubernetes will assign the namespace's default ServiceAccount to new pods in a namespace. You may specify for Alluxio pods to use any existing ServiceAccounts you may have in your cluster through the following:

helm

You may specify a top-level Helm value serviceAccount which will apply to the Master, Worker, and FUSE pods in the chart.

serviceAccount: sa-alluxio

You can override the top-level Helm value by specifying a value for the specific component's serviceAccount like below:

master:
  serviceAccount: sa-alluxio-master

worker:
  serviceAccount: sa-alluxio-worker

kubectl

You may add a serviceAccountName field to any of the Alluxio Pod template specs to have the Pod run using the matching ServiceAccount. For example:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  template:
    spec:
      serviceAccountName: sa-alluxio

Node Selectors & Tolerations

Kubernetes provides many options to control the scheduling of pods onto nodes in the cluster. The most direct of which is a node selector.

However, Kubernetes will avoid scheduling pods on any tainted nodes. To allow certain pods to schedule on such nodes, Kubernetes allows you to specify tolerations for those taints. See the Kubernetes documentation on taints and tolerations for more details.

helm

You may specify a node selector in JSON as a top-level Helm value, nodeSelector, which will apply to all pods in the chart. Similarly, you may specify a list of tolerations in JSON as a top-level Helm value, tolerations, which will also apply to all pods in the chart.

nodeSelector: {"app": "alluxio"}

tolerations: [ {"key": "env", "operator": "Equal", "value": "prod", "effect": "NoSchedule"} ]

You can override the top-level nodeSelector by specifying a value for the specific component's nodeSelector.

master:
  nodeSelector: {"app": "alluxio-master"}

worker:
  nodeSelector: {"app": "alluxio-worker"}

You can append to the top-level tolerations by specifying a value for the specific component's tolerations.

logserver:
  tolerations: [ {"key": "app", "operator": "Equal", "value": "logging", "effect": "NoSchedule"} ]

kubectl

You may add nodeSelector and tolerations fields to any of the Alluxio Pod template specs. For example:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  template:
    spec:
      nodeSelector:
        app: alluxio
      tolerations:
        - effect: NoSchedule
          key: env
          operator: Equal
          value: prod

Host Aliases

If you wish to add or override hostname resolution in the pods, Kubernetes exposes the containers' /etc/hosts file via host aliases. This can be particularly useful for providing hostname addresses for services not managed by Kubernetes, like HDFS.

helm

You may specify a top-level Helm value hostAliases which will apply to the Master and Worker pods in the chart.

hostAliases:
- ip: "127.0.0.1"
  hostnames:
    - "foo.local"
    - "bar.local"
- ip: "10.1.2.3"
  hostnames:
    - "foo.remote"
    - "bar.remote"

kubectl

You may add the hostAliases field to any of the Alluxio Pod template specs. For example:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  template:
    spec:
      hostAliases:
      - ip: "127.0.0.1"
        hostnames:
          - "foo.local"
          - "bar.local"
      - ip: "10.1.2.3"
        hostnames:
          - "foo.remote"
          - "bar.local"

Deployment Strategy

By default Kubernetes will use the 'RollingUpdate' deployment strategy to progressively upgrade Pods when changes are detected.

helm

The Helm chart currently only supports strategy for the logging server deployment:

logserver:
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 25%
      maxSurge: 1

kubectl

You may add a strategy field to any of the Alluxio Pod template specs to have the Pod run using the matching ServiceAccount. For example:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  template:
    spec:
      strategy:
        type: Recreate

ImagePullSecrets

Kubernetes supports accessing images from a Private Registry. After creating the registry credentials Secret in Kubernetes, you pass the secret to your Pods via imagePullSecrets.

helm

The following value applies the specified imagePullSecrets to all Pods in the Helm chart.

imagePullSecrets:
  - ecr
  - dev

kubectl

Add imagePullSecrets to your Pod specs. Eg:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: alluxio-master
spec:
  template:
    spec:
      containers:
      - name: alluxio-master
        image: private-registry/{{site.ALLUXIO_DOCKER_IMAGE}}:{{site.ALLUXIO_VERSION_STRING}}
      imagePullSecrets:
      - name: ecr
      - name: dev

Troubleshooting

Worker Host Unreachable

Alluxio workers use host networking with the physical host IP as the hostname. Check the cluster firewall if an error such as the following is encountered:

Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: finishConnect(..) failed: Host is unreachable: <host>/<IP>:29999

Check that <host> matches the physical host address and is not a virtual container hostname. Ping from a remote client to check the address is resolvable.

$ ping <host>

Verify that a client can connect to the workers on the ports specified in the worker deployment specification. The default ports are [29998, 29999, 29996, 30001, 30002, 30003]. Check access to the given port from a remote client using a network utility such as ncat:

$ nc -zv <IP> 29999

Permission Denied

From Alluxio v2.1 on, Alluxio Docker containers will run as non-root user alluxio with UID 1000 and GID 1000 by default. Kubernetes hostPath volumes are only writable by root so you need to update the permission accordingly.

Enable Debug Logging

To change the log level for Alluxio servers (master and workers), use the CLI command logLevel as follows:

Access the Alluxio CLI from the master Pod.

$ kubectl exec -ti alluxio-master-0 -- /bin/bash

From the master Pod, execute the following:

$ alluxio logLevel --level DEBUG --logName alluxio

Accessing Logs

The Alluxio master and job master run as separate containers of the master Pod. Similarly, the Alluxio worker and job worker run as separate containers of a worker Pod. Logs can be accessed for the individual containers as follows.

Master:

$ kubectl logs -f alluxio-master-0 -c alluxio-master

Worker:

$ kubectl logs -f alluxio-worker-<id> -c alluxio-worker

Job Master:

$ kubectl logs -f alluxio-master-0 -c alluxio-job-master

Job Worker:

$ kubectl logs -f alluxio-worker-<id> -c alluxio-job-worker

POSIX API

In order for an application container to mount the hostPath volume, the node running the container must have the Alluxio FUSE daemon running. The default spec alluxio-fuse.yaml runs as a DaemonSet, launching an Alluxio FUSE daemon on each node of the cluster.

If there are issues accessing Alluxio using the POSIX API:

Identify the node the application container ran on using the command kubectl describe pods or the dashboard.
Use the command kubectl describe nodes <node> to identify the alluxio-fuse Pod running on that node.
Tail logs for the identified Pod to view any errors encountered: kubectl logs -f alluxio-fuse-<id>.

Filename too long

Alluxio workers create a domain socket used for short-circuit access by default. On Mac OS X, Alluxio workers may fail to start if the location for this domain socket is a path which is longer than what the filesystem accepts.

2020-07-27 21:39:06,030 ERROR GrpcDataServer - Alluxio worker gRPC server failed to start on /opt/domain/1d6d7c85-dee0-4ac5-bbd1-86eb496a2a50
java.io.IOException: Failed to bind
	at io.grpc.netty.NettyServer.start(NettyServer.java:252)
	at io.grpc.internal.ServerImpl.start(ServerImpl.java:184)
	at io.grpc.internal.ServerImpl.start(ServerImpl.java:90)
	at alluxio.grpc.GrpcServer.lambda$start$0(GrpcServer.java:77)
	at alluxio.retry.RetryUtils.retry(RetryUtils.java:39)
	at alluxio.grpc.GrpcServer.start(GrpcServer.java:77)
	at alluxio.worker.grpc.GrpcDataServer.<init>(GrpcDataServer.java:107)
	at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
	at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
	at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
	at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
	at alluxio.util.CommonUtils.createNewClassInstance(CommonUtils.java:273)
	at alluxio.worker.DataServer$Factory.create(DataServer.java:47)
	at alluxio.worker.AlluxioWorkerProcess.<init>(AlluxioWorkerProcess.java:162)
	at alluxio.worker.WorkerProcess$Factory.create(WorkerProcess.java:46)
	at alluxio.worker.WorkerProcess$Factory.create(WorkerProcess.java:38)
	at alluxio.worker.AlluxioWorker.main(AlluxioWorker.java:72)
Caused by: io.netty.channel.unix.Errors$NativeIoException: bind(..) failed: Filename too long

If this is the case, set the following properties to limit the path length:

alluxio.worker.data.server.domain.socket.as.uuid=false
alluxio.worker.data.server.domain.socket.address=/opt/domain/d

Note: You may see performance degradation due to lack of node locality.

Worker Pods get OOMKilled by the Kubernetes scheduler

This is most likely caused due to the Kubernetes configured Pod resource limits having the limits.memory set too low.

Firstly, double check the configured values for your Alluxio worker Pod limits.memory. Note that the Pod consists of two containers, each with their own resource limits.

Check the configured resource requests and limits using kubectl describe pod, kubectl get pod, or equivalent Kube API requests. eg.,

$ kubectl get po -o json alluxio-worker-xxxxx | jq '.spec.containers[].resources'
{
  "limits": {
    "cpu": "4",
    "memory": "4G"
  },
  "requests": {
    "cpu": "1",
    "memory": "2G"
  }
}
{
  "limits": {
    "cpu": "4",
    "memory": "4G"
  },
  "requests": {
    "cpu": "1",
    "memory": "1G"
  }
}

If you used the Helm chart, the default values are:

worker:
  resources:
    limits:
      cpu: "4"
      memory: "4G"
    requests:
      cpu: "1"
      memory: "2G"

jobWorker:
  resources:
    limits:
      cpu: "4"
      memory: "4G"
    requests:
      cpu: "1"
      memory: "1G"

Even if you did not configure any values with Helm, you may still have resource limits in place due to a LimitRange applied to your namespace

Next, ensure that the nodes that the Alluxio worker pods are running on have sufficient resources matching your configured values. You can check that the nodes you intend to schedule Alluxio worker Pods on have sufficient resources to meet your requests using kubectl describe node, kubectl get node, or equivalent Kube API requests. eg.,

$ kubectl get no -o json k8sworkernode-0 | jq '.status.allocatable'
{
  "cpu": "8",
  "ephemeral-storage": "123684658586",
  "hugepages-1Gi": "0",
  "hugepages-2Mi": "0",
  "memory": "64886128Ki",
  "pods": "110"
}

Isolating Alluxio worker Pods from other Pods in your Kubernetes cluster can be accomplished with the help of node selectors and node taints + tolerations.

Keep in mind that the Alluxio worker Pod definition uses a DaemonSet, so there will be worker Pods assigned to all eligible nodes

Next, verify the Alluxio workers' configured ramdisk sizes (if any). See the list of Alluxio configuration properties for additional details.

If you used the Helm chart, the Alluxio site properties are configured using properties. eg.,

properties:
  alluxio.worker.ramdisk.size: 2G
  alluxio.worker.tieredstore.levels: 1
  alluxio.worker.tieredstore.level0.alias: MEM
  alluxio.worker.tieredstore.level0.dirs.mediumtype: MEM
  alluxio.worker.tieredstore.level0.dirs.path: /dev/shm
  alluxio.worker.tieredstore.level0.dirs.quota: 2G

Otherwise, you can view and modify the site properties in the alluxio-config ConfigMap. eg.,

$ kubectl get cm -o json alluxio-config | jq '.data.ALLUXIO_WORKER_JAVA_OPTS'
"-Dalluxio.worker.ramdisk.size=2G
-Dalluxio.worker.tieredstore.levels=1
-Dalluxio.worker.tieredstore.level0.alias=MEM
-Dalluxio.worker.tieredstore.level0.dirs.mediumtype=MEM
-Dalluxio.worker.tieredstore.level0.dirs.path=/dev/shm
-Dalluxio.worker.tieredstore.level0.dirs.quota=2G "

NOTE: Our DaemonSet uses emptyDir volumes as the Alluxio worker Pod's ramdisk in Kubernetes.

spec:
  template:
    spec:
      volumes:
        - name: mem
          emptyDir:
            medium: "Memory"
            sizeLimit: 1G

This results in the following nuances:

sizeLimit has no effect on the size of the allocated ramdisk unless the SizeMemoryBackedVolumes feature gate is enabled (enabled by default as of Kubernetes 1.22).
As stated in the Kubernetes emptyDir documentation, if no size is specified then memory-backed emptyDir volumes will have capacity allocated equal to half the available memory on the host node. This capacity is reflected inside of your containers (for example when running df -u). However if the combined size of your ramdisk and container memory usage exceeds the pod's limits.memory then the Kubernetes scheduler will trigger an OOMKilled on that pod. This is a very likely overlooked source of memory consumption in Alluxio worker Pods.

Lastly, verify the Alluxio worker JVM heap and off-heap maximum capacities. These are configured with the JVM flags -Xmx/-XX:MaxHeapSize and -XX:MaxDirectMemorySize respectively.

See the Oracle Java documentation for more details.

To adjust those values, you would have to manually update the (...)_JAVA_OPTS environment variables in the alluxio-config ConfigMap. For example:

apiVersion: v1
kind: ConfigMap
metadata:
  name: alluxio-config
data:
  ALLUXIO_JAVA_OPTS: |-
    -Xmx2g -Dalluxio.master.hostname=alluxio-master-0 ...
  ALLUXIO_MASTER_JAVA_OPTS: |-
    -Dalluxio.master.hostname=${ALLUXIO_MASTER_HOSTNAME}
  ALLUXIO_JOB_MASTER_JAVA_OPTS: |-
    -Dalluxio.master.hostname=${ALLUXIO_MASTER_HOSTNAME}
  ALLUXIO_WORKER_JAVA_OPTS: |-
    -XX:MaxDirectMemorySize=2g -Dalluxio.worker.hostname=${ALLUXIO_WORKER_HOSTNAME} ...
  ALLUXIO_JOB_WORKER_JAVA_OPTS: |-
    -XX:MaxDirectMemorySize=1g -Dalluxio.worker.hostname=${ALLUXIO_WORKER_HOSTNAME} ...
  ALLUXIO_FUSE_JAVA_OPTS: |-
    -Dalluxio.user.hostname=${ALLUXIO_CLIENT_HOSTNAME} -XX:MaxDirectMemorySize=2g
  ALLUXIO_WORKER_TIEREDSTORE_LEVEL0_DIRS_PATH: /dev/shm

Thus to avoid worker Pods running into OOMKilled errors,

Verify that the nodes your Alluxio worker Pods are scheduled on have sufficient memory to satisfy all the limits.memory specifications assigned.
Ensure you have configured alluxio.worker.ramdisk.size and alluxio.worker.tieredstore.level0.dirs.quota low enough such that the memory consumed by the ramdisk combined with the JVM memory options (-Xmx, -XX:MaxDirectMemorySize) do not exceed the Pod's limits.memory. It is recommended to allow for some overhead as memory may be consumed by other processes as well.

Aside: There is currently an open issue in Alluxio where Alluxio's interpretation of byte sizes differs from Kubernetes (due to Kubernetes distinguishing between "-bibytes"). This is unlikely to cause OOMKilled errors unless you are operating on very tight memory margins.

JVM not seeing correct memory limit from cgroup

It is a known issue that in some early versions of Java 8, the JVM running in a container will determine its heap size(if not specified with -Xmx and -Xms) based on the memory of the physical host instead of the container. In that case, the JVM may attempt to use more memory than the container resource limit and gets killed. You can find more detailed explanations here.

Since Java 8u131, some JVM flags can be turned on in order to correctly read the memory from cgroup. You can refer to our values.yaml from our Helm chart template, and uncomment the below options. These options will be added to the JVM options of all Alluxio containers, including the masters and workers etc. You can find more detailed explanations here.

# Recommended JVM Heap options for running in Docker
# Ref: https://developers.redhat.com/blog/2017/03/14/java-inside-docker/
# These JVM options are common to all Alluxio services
jvmOptions:
  - "-XX:+UnlockExperimentalVMOptions"
  - "-XX:+UseCGroupMemoryLimitForHeap"
  - "-XX:MaxRAMFraction=2"

From Java git 8u191 on, the container support works out-of-the-box. So you don't need to turn on the flags mentioned above any more.

You should check the Java version in the container you are using to ensure the correct memory limits are respected. Also it is recommended to go to the running container and double check the JVM process is running with the correct memory consumption.

tmpfs is smaller than the configured size

In Kubernetes context, g or GB means 1000^3 and gi or GiB means 1024^3. However, in Alluxio context, g or GB means 1024^3. So when we use g and pass the quota to Alluxio and K8s, K8s grants 1000^3 but Alluxio tries to utilize 1024^3. For example if it is an emptyDir, then the pod using the emptyDir will be killed for overusing resources.

Therefore, we recommend using Gi whenever possible in helm chart or yaml files to avoid such issue.

Last updated 1 month ago