MemVerge.ai Install Guide

Prerequisites

• Access to a Kubernetes v1.28+ cluster with cluster-admin role. Supported distributions are:
  • Vanilla Kubernetes
  • K3s
• CRI runtime support (https://kubernetes.io/docs/setup/production-environment/container-runtimes/):
  • Containerd: v1.7+.
  • CRI-O: v1.28+.
  • Others are not supported.
• Ingress Controller is set up in the cluster. There are many choices. See https://kubernetes.io/docs/concepts/services-networking/ingress-controllers/.
• A default storage class is set up in the cluster to dynamically create persistent volume claims.
  • To support checkpoint feature, the storage class must be able to move a persistent volume from one node to another.
• The following helm charts are subcomponents of MemVerge.ai; they are deployed automatically, and should not be present on the cluster prior to MemVerge.ai installation:
  • NVIDIA GPU Operator and its dependencies (DCGM Exporter, NVIDIA Device Plugin, Node Feature Discovery, etc.)
  • HAMi
  • Prometheus Operator and its derivatives, such as kube-state-metrics
  • Rancher
• kubectl version v1.28+.
• Helm version v3.14+.
• Download the mvai package from the releases page that includes:
  • Binary mvaictl to manage MemVerge.ai product from cmdline.
  • Script mvai-cleanup.sh to fully cleanup MemVerge.ai installation.

Acquire GitHub Token

Contact MemVerge Customer Support ([email protected]) to acquire a personal access token of GitHub account mv-customer-support for downloading MemVerge.ai helm chart and container images.

Login to GitHub Registry

helm registry login ghcr.io/memverge
# Username: mv-customer-support
# Password: <personal-access-token>

Create Image Pull Secret

kubectl create namespace cattle-system

kubectl create secret generic memverge-dockerconfig --namespace cattle-system \
  --from-file=.dockerconfigjson=$HOME/.config/helm/registry/config.json \
  --type=kubernetes.io/dockerconfigjson

Install cert-manager

Skip this step if cert-manager is already installed.

helm repo add jetstack https://charts.jetstack.io --force-update

helm install cert-manager jetstack/cert-manager --namespace cert-manager \
  --create-namespace --set crds.enabled=true

Or check https://cert-manager.io/docs/installation for other options.

Install MemVerge.ai

The MemVerge.ai management server is designed to be secure by default and requires SSL/TLS configuration. There are three recommended options for the source of the certificate used for TLS termination at the MemVerge.ai server:

1. MemVerge.ai-generated Certificate

The default is for MemVerge.ai to generate a CA and uses cert-manager to issue the certificate for access to the MemVerge.ai server interface.

helm install --namespace cattle-system mvai oci://ghcr.io/memverge/charts/mvai \
  --wait --timeout 20m --version <version> \
  --set hostname=<load-balancer-hostname> --set bootstrapPassword=admin

• Set the hostname to the DNS name of the load balancer.
  • If there is only one control-plane node and there is no load balancer, the hostname can be the DNS name of the control-plane node.
  • If this single control-plane node has no DNS name, a fake domain name <IP_OF_NODE>.sslip.io can be used.
• Set the bootstrapPassword to something unique for the admin user.
• Note that the version should not include prefix v. To install the latest development version, replace the --version <version> option with --devel in the install command.

2. Let's Encrypt

This option uses `cert-manager` to automatically request and renew `Let's Encrypt` certificates. This is a free service that provides you with a valid certificate as `Let's Encrypt` is a trusted CA.

helm install --namespace cattle-system mvai oci://ghcr.io/memverge/charts/mvai \
  --wait --timeout 20m --version <version> \
  --set hostname=<load-balancer-hostname> --set bootstrapPassword=admin \
  --set ingress.tls.source=letsEncrypt --set letsEncrypt.email=<[email protected]> \
  --set letsEncrypt.ingress.class=<ingress-controller-name>

3. Bring Your Own Certificate

In this option, Kubernetes secrets are created from your own certificates for MemVerge.ai to use.

When you run this command, the hostname option must match the Common Name or a Subject Alternative Names entry in the server certificate or the Ingress controller will fail to configure correctly.

helm install --namespace cattle-system mvai oci://ghcr.io/memverge/charts/mvai \
  --wait --timeout 20m --version <version> \
  --set hostname=<load-balancer-hostname> --set bootstrapPassword=admin \
  --set ingress.tls.source=secret

If you are using a Private CA signed certificate , add --set privateCA=true to the command:

helm install --namespace cattle-system mvai oci://ghcr.io/memverge/charts/mvai \
  --wait --timeout 20m --version <version> \
  --set hostname=<load-balancer-hostname> --set bootstrapPassword=admin \
  --set ingress.tls.source=secret --set privateCA=true

Now that MemVerge.ai is deployed, see Adding TLS Secrets to publish your certificate files so MemVerge.ai and the Ingress controller can use them.

Billing Database Installation

Billing features require persistent storage. By default, the cluster's default StorageClass is used. If the cluster doesn't have a default storage class capable of provisioning a volume for billing, there are a few alternative ways to configure persistent storage.

Note: NFS should be used with caution. root-squash is incompatible, and other configurations may cause issues, such as mount options or NFS version constraints. See the MySQL documentation for details.

Configuring Billing to Use a Non-Default StorageClass

If the cluster has an alternative StorageClass suitable for the billing database, you can override the default StorageClass by adding the following flag to the helm install mvai command:

--set billing.database.volume.pvc.storageClass=[StorageClass name]

Configuring the Installation to Use a Predefined PVC

If the cluster administrator wants to manually define a PVC for the billing database, disable automatic PVC creation by setting the StorageClass to an empty string:

--set billing.database.volume.pvc.storageClass=""

Configuring the Installation to Use a HostPath Mount

The billing database can be configured to use hostPath storage, which directly creates and mounts a directory on the host node to the billing database pod. To enable hostPath storage, add the following flag to the helm install mvai command:

--set billing.database.volume.type="hostPath"

By default, the directory /data/memverge/mvai/mysql-billing is used. This path can be overridden with the following flag:

--set billing.database.volume.hostPath.path="/desired/path/here"

Uninstall MemVerge.ai

This command deletes the MemVerge.ai deployment, but leaves MemVerge.ai CRDs and user-created MemVerge.ai CRs in the cluster.

helm uninstall --namespace cattle-system mvai

To completely cleanup MemVerge.ai resources, run the mvai-cleanup.sh script from the released mvai package.