Upgrading Layar

Introduction

This guide will go over how to upgrade Layar, the method you use will depend on if the server has access to the internet.

Internet Available

If you are using the Layar Amazon Machine Image (AMI) simply run:

/data/layar/deployLayar.sh upgrade

If you installed Layar manually, run:

helm repo update
kubectl delete deployment --all
kubectl delete statefulset --all
sleep 30

Now run the same helm command that was used during installation but replace the word install with upgrade. For example:

helm upgrade layar vyasa/layar --set APPURL=layar.acme.com --set TRITON_GPU_COUNT=3

Air Gapped

Pre-requisites

Before continuing you will need to have the .tar containing the updated version. If you do not have this, please reach out to support at [email protected].

Ensure the file is placed in a directory on the server that runs Layar.

Install

Login to the server as an Admin.
Navigate to the location of the .tar file.
Run tar xvf <name of file>
Once the files have extracted, cdto the newly created directory.
Run for i in 'ls vyasa*'; do echo "loading $i": lz4cat $i | docker load : done

🚧 If you are not changing the LLM during the upgrade skip to step 13

Run kubectl get pvc | grep tgi
Copy the name of the TGI PVC.
Run cd /data/vyasa-volumes/<TGI PVC Name>
Run Tar xvf <path of the models .tar file>
Run vi /data/layar/layar.config
Edit ALL_MODEL_INFOwith the name of the new models.
1. Verify that model arguements do not need to be added to the config file. If you are unsure, review the Model Arguements
2. If you are unsure how to use ALL_MODEL_INFO please reference Multi-Model Configuration Guide.
Save the changes.
Run kubectl delete deployments --all
Run kubectl delete statefulsets --all
cd back to the /data/layar
Run the helm command that was used to install Layar.