Install Aizen remote component - Static deployment

Prerequisites

Below are the software components that should be installed on GPU node(s) before installing Aizen remote components

  • Kubernetes cluster 1.26+

    • Kubernetes is a portable, extensible, open source platform for managing containerized workloads and services, that facilitates both declarative configuration and automation

  • Persistent storage class

    • Aizen uses presistent volumes (PV) for storing ephemeral cache and transient data. It needs storage class that can dynamically provision volumes. This storage should have high performance, typical of the SSD class. Storage class options are avaiable for public/private cloud and on prem deployments

  • Object storage

    • Aizen needs access to a scalable cloud storage for object storage

  • Nvidia drivers and libraries

  • Helm 3.2.0+

    • helm is a tool that automates creation, packaging, configuration, deployment of Kubernetes applications into a single reusable package

  • kubectl

    • kubectl is a command line tool that enables communications between Kubernetes API and the control plane. It allows application deployment, cluster resource management, and resource monitoring

  • Dockerhub credentials to access microservice images

Note

The assumption is this GPU node(s) has Kubernetes installed along with Container Storage Interface(CSI) and gpu operator

  • Create namespace for remote aizen components

kubectl create ns aizen

  • Label the gpu nodes

kubectl label node <nodename or ip> aizen.com/gpu.deploy=true

If you have Docker credential information, first create kubernetes secret for accessing Aizen images

kubectl create secret docker-registry aizenrepo-creds
--docker-username=aizencorp
--docker-password=<YOUR DOCKER CREDENTIALS>
-n aizen

Deploy remote components

#Script for Aizen remote deployment
#----------------------------------
NAMESPACE=aizen
HELMCHART_LOCATION=aizenremote-helmcharts-1.0.0

STORAGE_CLASS=
BUCKET_NAME=
CLUSTER_NAME=

CLOUD_ENDPOINT_URL=
CLOUD_ACCESSKEY_ID=
CLOUD_SECRET_KEY=
CLOUD_PROVIDER_REGION=
CLOUD_PROVIDER_TYPE=

#Needed only for cloudian
CLOUD_ENDPOINT_IP=

#IMAGE
IMAGE_REPO=aizencorp
IMAGE_REPO_SECRET=
IMAGE_TAG=1.0.0

#PVC
METASTORAGE_PERSISTENCE_SIZE=200Gi

#You don't need to change anything below this line
helm -n $NAMESPACE install aizenremote $HELMCHART_LOCATION/aizenremote \
--set global.clustername=$CLUSTER_NAME,\
global.s3.endpoint_url=$CLOUD_ENDPOINT_URL,\
global.s3.secrets.values.s3_access_key=$CLOUD_ACCESSKEY_ID,\
global.s3.secrets.values.s3_secret_key=$CLOUD_SECRET_KEY,\
global.customer_bucket_name=$BUCKET_NAME,\
global.storage_class=$STORAGE_CLASS,\
global.cloud_provider_type=$CLOUD_PROVIDER_TYPE,\
global.cloud_provider_region=$CLOUD_PROVIDER_REGION,\
global.image_registry=$IMAGE_REPO,\
global.storage_class=$STORAGE_CLASS,\
global.image_secret=$IMAGE_REPO_SECRET,\
global.image_tag=$IMAGE_TAG,\
metastorage.volume_size=$METASTORAGE_PERSISTENCE_SIZE

Check deployment status of remote components

  • Check the status of all remote components

kubectl -n aizen get pods