0.0.10

Memory Machine Operator

Memory Machine Architecture

Before continuing, we recommend that you read Memory Machine Container Architecture to understand how Memory Machine™ Container is organized.

Prerequisites

The following instructions assume that you have installed the OpenShift Container Platform, customized to your requirements.

You must have a Kubernetes cluster deployed, and the kubectl command-line tool must be configured to communicate with your cluster.

The Operator SDK is a framework used to develop and run Operators. You will use the operator-sdk command to run the Memory Machine Operator, so you must have the Operator SDK installed.

Please be sure to have a StorageClass that supports dynamic provisioning. You can change the StorageClass in etcdConfig. Please see MemoryMachine Configuration List

Configuring the Memory Machine Platform

Before installing Memory Machine Operator, complete the steps described below.

1. Obtain and install a Memory Machine License.
2. Configure PMem.

Obtaining and Installing a Memory Machine License

Obtain a Memory Machine license from MemVerge. To request a license, send an email to [email protected]. Include the MAC address of the node that will contain the cluster Memory Machine Management Center pod. The license is valid for the entire cluster but is installed only on the node containing the Memory Machine Management Center pod.

Unless otherwise noted, type all commands on a client node that communicates with your cluster.

After you obtain the license:

1. Label the Memory Machine Management Center node as memory-machine.memverge.com=license-server by typing the following command:

   $ kubectl label node <memory-machine-license-node> memory-machine.memverge.com=license-server
   

   where <memory-machine-license-node> is the name of the Memory Machine node.

2. Create a namespace called memverge by typing the following:

   $ kubectl create ns memverge
   

3. Create a Secret for the Memory Machine license with the name memory-machine-license in the memverge namespace as follows:

   $ kubectl create secret generic memory-machine-license \
   -n memverge \
   --from-file=<license=<path-to-your-license-file>
   

   where <path-to-your-license-file> is the Memory Machine license file.

Configuring PMem

Configure the PMem as described in the following steps:

1. For each worker node that you want to use with Memory Machine, label the worker node with storage=pmem as follows:

   $ kubectl label node <node-name> storage=pmem
   

   where <node-name> is the name of the worker node.

2. Run the config-pmem.sh script to reserve PMem capacity (in GiB) for Memory Machine:

   $ config-pmem.sh --mm-capacity <pmem-size>
   

   where pmem-size is the number of GiB of PMem. The value of pmem-size cannot exceed the PMem capacity available on the node.

   If there is no physical PMem device, DRAM can be used to emulate PMem. In that case, add the --pmem-emulation flag as follows:

   $ config-pmem.sh --pmem-emulation --mm-capacity <pmem-size>
   

Installing the Memory Machine Operator

Create an image-pull secret named memverge-dockerconfig and retrieve the Operator image by following these steps:

1. Log into Docker as follows:

   $ cat <file-containing-github-token> | docker login ghcr.io
   

2. Create the secret as follows:

   $ kubectl create secret generic memverge-dockerconfig \
       -n memverge \
       --from-file=.dockerconfigjson=$HOME/.docker/config.json \
       --type=kubernetes.io/dockerconfigjson
   

3. Update the pod credentials as follows:

   $ kubectl patch serviceaccount default \
       -n memverge \
       -p '{"imagePullSecrets": [{"name": "memverge-dockerconfig"}]}'
   

4. Install the Memory Machine Operator node by typing the operator-sdk command as follows:

   $ operator-sdk run bundle ghcr.io/memverge/memory-machine-operator-bundle:0.0.10 \
       -n memverge \
       --pull-secret-name memverge-dockerconfig
   

Deployment Examples

The following examples are available:

• To deploy Memory Machine using a Memory Machine Operator that runs with a test pod, see example.md.

• To deploy a Hazelcast cluster on Memory Machine and checkpoint-restore the Hazelcast cluster, see tutorials/hazelcast/README.md.

Maintaining the Memory Machine Cluster

Configuration Options

For a complete list of configuration options, refer to the Memory Machine Configuration List.

Changing Configurations on Existing Memory Machine Pods

To change the configuration of running Memory Machine pods, redeploy the Memory Machine pods with the new configuration as follows:

1. Stop all applications running on your cluster.

2. Delete the Memory Machine DaemonSet as follows:

   $ kubectl delete daemonset memory-machine -n memverge
   

3. Edit the memorymachine.yml file to change the Memory Machine configuration.

4. Apply the new Memory Machine YAML file to the application space as follows:

   $ kubectl apply -f <path_to_memory_machine_yaml_file>
   

   where <path_to_memory_machine_yaml_file> is the new Memory Machine YAML file.

   The Operator deploys Memory Machine with the new configuration.

Removing Memory Machine Resources

To remove resources created by Memory Machine Operator, do the following:

1. Remove your applications from the cluster.

2. Delete the Memory Machine YAML file as follows:

   $ kubectl delete -f <path-to-memory-machine-yaml-file>
   

   where <path-to-memory-machine-yaml-file> is the Memory Machine YAML file.

The operator detects the deletion of Memory Machine and removes the resources automatically.

Removing the Operator

Remove the Memory Machine Operator and Mutating Webhook by using the following command:

$ operator-sdk cleanup memory-machine-operator