The Flex Volume Driver has been deprecated in favour of the CSI Driver
Flex should only be used where the CSI driver is not supported due to a lower level of Kubernetes version.
This helm chart installs the FlexVolume plugin on a Kubernetes cluster.
Platform and Software Dependencies
- #### Operating Systems Supported*:
- CentOS 7
- RHEL 7
- CoreOS (Ladybug 1298.6.0 and above)
- Ubuntu 16.04
- Ubuntu 18.04
- #### Environments Supported*:
- Kubernetes 1.6+
- Helm 2.9.1+ (NOTE: Helm3 is not supported for FlexDriver)
- OpenShift 3.11
- AWS EKS 1.14
- #### Other software dependencies:
- Latest linux multipath software package for your operating system (Required)
- Latest Filesystem utilities/drivers (XFS by default, Required)
- Latest iSCSI initiator software for your operating system (Optional, required for iSCSI connectivity)
- Latest NFS software package for your operating system (Optional, required for NFS connectivity)
- Latest FC initiator software for your operating system (Optional, required for FC connectivity, FC Supported on Bare-metal K8s installations only)
- #### FlashArray and FlashBlade:
- The FlashArray and/or FlashBlade should be connected to the compute nodes using Pure’s best practices
* Please see release notes for details
How to install
Add the Pure Storage helm repo
helm repo add pure https://purestorage.github.io/helm-charts helm repo update helm search repo pure-k8s-plugin
Optional (offline installation): Download the helm chart
git clone https://github.com/purestorage/helm-charts.git
Create your own
values.yaml and install the helm chart with it, and keep it. The easiest way is to copy
the default ./values.yaml
The following table lists the configurable parameters and their default values.
||The image name to pull from||
||The image tag to pull||
||Image pull policy||
||Enable/disable debug mode for app||
||Control to create the built-in StorageClasses ‘pure’, ‘pure-file’ and ‘pure-block’||
||Name of K8s/openshift service account for installing the plugin||
||Block volume default filesystem type. Not recommended to change!||
||Block volume default mkfs options. Not recommended to change!||
||Block volume default filesystem mount options. Not recommended to change!||””|
||iSCSI login timeout in seconds. Not recommended to change!||
||List of CIDR blocks allowed as iSCSI targets, e.g. 10.0.0.0/24,10.1.0.0/16. Use comma (,) as the separator, and empty string means allowing all addresses.||””|
||Enable/Disable attachment preemption!||
||Block volume access protocol, either ISCSI or FC||
||Enable/Disable FlashBlade snapshots||
||Namespace for the backend storage||
||Orchestrator type, such as openshift, k8s||
||Full path of directory to install flex plugin, works with image.tag >= 2.0.1||
||Array list of all the backend FlashArrays and FlashBlades||must be set by user, see an example below|
||NodeSelectors Select node-labels to schedule flex-plugin. See this for setting node selectors on Openshift.||
||NodeSelectors Select node-labels to schedule provisioner. See this for setting node selectors on Openshift.||
arrays: FlashArrays: - MgmtEndPoint: "126.96.36.199" APIToken: "a526a4c6-18b0-a8c9-1afa-3499293574bb" Labels: rack: "22" env: "prod" - MgmtEndPoint: "188.8.131.52" APIToken: "b526a4c6-18b0-a8c9-1afa-3499293574bb" FlashBlades: - MgmtEndPoint: "184.108.40.206" APIToken: "T-c4925090-c9bf-4033-8537-d24ee5669135" NfsEndPoint: "220.127.116.11" Labels: rack: "7b" env: "dev" - MgmtEndPoint: "18.104.22.168" APIToken: "T-d4925090-c9bf-4033-8537-d24ee5669135" NfsEndPoint: "22.214.171.124" Labels: rack: "6a"
Assigning Pods to Nodes
It is possible to make flex-daemon and provisioner run on specific nodes
affinity. You can set these config
separately for flex-daemon and provisioner (e.g.
The unprefixed parameters have been deprecated but will still be used as a
fallback if the prefixed parameters are not set. (e.g. if
provisioner.nodeSelector is not, provisioner will use the value of
nodeSelector as a fallback) This makes sure the behavior is backward
Install the plugin in a separate namespace (i.e. project)
For security reason, it’s strongly recommended to install the plugin in a separate namespace/project. Make sure the namespace is existing, otherwise create it before installing the plugin.
Customize your values.yaml including arrays info (replacement for pure.json), and then install with your values.yaml.
Dry run the installation, and make sure your values.yaml is working correctly:
helm install pure-storage-driver pure/pure-k8s-plugin --namespace <namespace> -f <your_own_dir>/yourvalues.yaml --dry-run --debug
Run the Install:
# For Openshift 3.11 only: # you need to add the privileged securityContextConstraints (scc) to the service account which is created for plugin installation. # You can find the serviceaccount info from your values.yaml (if not in it, find in the default values.yaml). # The service account should be "system:serviceaccount:<project>:<clusterrolebinding.serviceAccount.name>" oc adm policy add-scc-to-user privileged system:serviceaccount:<project>:<clusterrolebinding.serviceAccount.name> # Install the plugin (works for both openshift and kubernetes) helm install pure-storage-driver pure/pure-k8s-plugin --namespace <namespace> -f <your_own_dir>/yourvalues.yaml
The values in your
values.yaml overwrite the ones in
pure-k8s-plugin/values.yaml, but any specified with the
--set option will take precedence.
helm install pure-storage-driver pure/pure-k8s-plugin --namespace <namespace> -f <your_own_dir>/yourvalues.yaml \ --set flasharray.sanType=fc \ --set namespace.pure=k8s_xxx \ --set orchestrator.name=openshift
How to update
Update your values.yaml with the correct arrays info, and then upgrade the helm as below.
Note: Ensure that the values for
--set options match when run with the original install step. It is highly recommended
to use the values.yaml and not specify options with
--set to make this easier.
helm upgrade pure-storage-driver pure/pure-k8s-plugin --namespace <namespace> -f <your_own_dir>/yourvalues.yaml --set ...
Using Snapshots with a FlashArray
More details on using the snapshot functionality can be found here
Using Labels to control volume topology
More details on using configuration labels can be found here
How to upgrade the driver version
It’s not recommended to upgrade by setting the
image.tag in the image section of values.yaml. Use the version of
the helm repository with the tag version required. This ensures the supporting changes are present in the templates.
# list the avaiable version of the plugin helm repo update helm search repo pure-k8s-plugin -l # select a target chart version to upgrade as helm upgrade pure-storage-driver pure/pure-k8s-plugin --namespace <namespace> -f <your_own_dir>/yourvalues.yaml --version <target chart version>
How to upgrade from the legacy installation to helm version
This upgrade will not impact the in-use volumes/filesystems from data path perspective. However, it will affect the in-flight volume/filesystem management operations. So, it is recommended to stop all the volume/filesystem management operations before doing this upgrade. Otherwise, these operations may need to be retried after the upgrade.
Uninstall the legacy installation using the install script provided as
Reinstall via helm
arraysinfo in your
values.yaml(online tool: https://www.json2yaml.com/).
flexPathmatch up exactly with kubelet’s
a. How to find the full path of the directory for pure flex plugin (i.e.
# ssh to a node which has pure flex plugin installed, and check the default value of "volume-plugin-dir" from "kubelet --help" # and then find the full path of the directory as below: # for k8s root@k8s-test-k8s-0:~# find /usr/libexec/kubernetes/kubelet-plugins/ -name "flex" | xargs dirname /usr/libexec/kubernetes/kubelet-plugins/volume/exec/pure~flex # for openshift on RHEL Server root@k8s-test-openshift-0:~# find /usr/libexec/kubernetes/kubelet-plugins/ -name "flex" | xargs dirname /usr/libexec/kubernetes/kubelet-plugins/volume/exec/pure~flex # for openshift 3.10+ on RHEL Atomic root@k8s-test-openshift-0:~# find /etc/origin/kubelet-plugins/ -name "flex" | xargs dirname /etc/origin/kubelet-plugins/volume/exec/pure~flex
If Kubernetes is deployed using containerized kubelet services then there may be steps required to ensure it can use the FlexVolume plugin. In general there are a few requirements that must be met for the plugin to work.
The container running the kubelet service must have: * Access to the host systems PID namespace * Access to host devices and sysfs (
/sys) * Access to the kubelet volume plugin directory For the volume plugin directory this defaults to
/usr/libexec/kubernetes/kubelet-plugins/volume/exec/but can be adjusted with the kubelet
volume-plugin-diroption. Where possible the containerized kubelet should have this directory passed in from the host system. To change the volume plugin directory a few steps are required: * Update the kubelet service to use the
volume-plugin-diroption, and direct it to the new location. * Ensure the kubelet container is configured to mount the new location into the container. * Ensure that the
pure-flex-daemon.yamlis configured to to use the new plugin directory for the
kubelet-pluginshost volume mount. This allows for the
pure-flexplugin to be installed in the new location on the filesystem, and for the kubelet to have access to the plugin.
Platform Specific Considerations
Some Kubernetes environments will require special configuration, especially on restrictive host operating systems where parts of it are mounted read-only.
Atomic is configured to have the
/usrdirectory tree mounted as read-only. This will cause problems installing the
pure-flexplugin as write permission is required. To get things working an alternate plugin directory should be used, a good option is
/etc/kubernetes/volumeplugins/. This is convienient for both because it is writable, and the kubelet container will already be mounting the
/etc/kubernetes/directory in to the kubelet. Once changed the kublet parameters need to be updated to set the
/etc/kubernetes/volumeplugins/, and the
pure-flexDaemonSet needs to be adjusted to install there as well via the
flexPathoption in your
Similar to the Atomic hosts this has a read-only
/usrtree and requires the plugin to be installed to an alternate location. Follow the same recommendations to use
/etc/kubernetes/volumeplugins/and adjust the kubelet service to use the
--volume-plugin-dirCLI argument and mount the
/etc/kubernetesdirectory into the container.
openshiftand configure the other OpenShift specific options. Note: the deployment is done with the default service account, and requires privileged containers. This means you may need to modify the service account used to use a new or existing service account with the right permissions or add the privileged scc to the default service account.
OpenShift Containerized Deployment
When deploying OpenShift with the containerized deployment method it is going to require mounting the plugin directory through to the container running the kubelet service. The kubelet configuration is then set via the
kubeletArgumentssection to set the
volume-plugin-dir. The easiest path to use is something like
/etc/origin/kubelet-pluginsor similar as the node config path is passed through to the container.
Release notes can be found here