Provider Upgrade

The upgrade instructions provided in this guide pertain to Akash providers installed via Helm Charts.

If your provider was installed using Praetor, please visit the Praetor Discord server for upgrade guidance and inquiries.

All providers MUST upgrade their Akash Provider and related components (RPC Node, CRDs, etc) on the date of the Mainnet4 upgrade (November 17th 2022) to avoid disruption. Providers which are not upgrade post Mainnet4 will be unable to bid on new orders. Pre-existing deployments on non-upgraded providers will not incur service disruption. But the ability to bid on new deployments, allow container shell access, and other features will be impacted until the provider’s RPC nodes are upgraded to 0.18.0 and Providers are upgraded to use Provider Services version 0.1.0.

NOTE - please do not attempt to upgrade your provider until the Mainnet4 upgrade has been completed. The Helm Charts will be updated to the use the correct/current binaries on the day of the Mainnet4 upgrade.

Provider Upgrade Procedure

Update Local Helm Repo

helm repo update akash

RPC Node

helm -n akash-services get values akash-node | grep -v '^USER-SUPPLIED VALUES' > akash-node-values.yml

Prior to executing the Helm Upgrade command - inspect akash-node-values.yml for image tag and remove it if present

helm upgrade akash-node akash/akash-node -n akash-services -f akash-node-values.yml

Akash Provider

Annotate CRDs

Annotate your current CRDs if you haven’t yet, they are normally delivered by the helm-charts, but they were missing in some releases. They prevent CRD from being removed upon helm uninstall.

kubectl annotate crd manifests.akash.network helm.sh/resource-policy=keep
kubectl annotate crd providerhosts.akash.network helm.sh/resource-policy=keep
kubectl annotate crd providerleasedips.akash.network helm.sh/resource-policy=keep

If you get some errors, ignore them. Most likely you already have these annotations set in place.

It is normal if you do not have providerleasedips yet, it comes with mainnet4 upgrade only.

Manifest Backup

Backup all provider manifest prior to initiating the upgrade process
You can restore them later if you lose them (e.g. by using kubectl apply -f manifests-backup.yaml). To verify the CRD contents use kubectl -n lease get manifests command. (Change manifests to providerhosts or providerleasedips depending on CRD you are verifying).

kubectl -n lease get manifests -o yaml > manifests-backup.yaml

kubectl -n lease get providerhosts -o yaml > providerhosts-backup.yaml

# this is only valid for akash-ip-operator

kubectl -n lease get providerleasedips -o yaml > providerleasedips-backup.yaml

helm -n akash-services get values akash-provider | grep -v '^USER-SUPPLIED VALUES' > akash-provider-values.yml

With the release of the akash-provider helm-chart version 4.0.5, the provider is now deployed as a StatefulSet.
Use the following step to uninstall any pre-existing Provider instances which used Deployment kubernetes resource type before deploying a new one. But first, backup the values using helm get values .... command as covered in prior step.

helm -n akash-services uninstall akash-provider

Prior to executing the Helm Upgrade command - inspect akash-provider-values.yml for image tag and remove it if present

helm upgrade --install akash-provider akash/provider -n akash-services -f akash-provider-values.yml

Hostname Operator

Ensure Any Prior Hostname Operator Instances are Uninstalled

helm uninstall hostname-operator -n akash-services

Install/Upgrade the Hostname Operator

helm upgrade --install akash-hostname-operator akash/akash-hostname-operator -n akash-services

Inventory Operator

Only necessary if providing persistent storage

helm upgrade inventory-operator akash/akash-inventory-operator -n akash-services

Ingress Controller (Optional)

helm -n ingress-nginx get values akash-ingress | grep -v '^USER-SUPPLIED VALUES' > akash-ingress-values.yml

Prior to executing the Helm Upgrade command - inspect akash-ingress-values.yml for image tag and remove it if present

helm upgrade akash-ingress akash/akash-ingress -n ingress-nginx -f akash-ingress-values.yml

Issue following to unblock the akash-provider 8443/tcp port in case you have been using a very old akash-ingress helm chart which used the deployment k8s kind of resource

kubectl -n ingress-nginx delete deployment ingress-nginx-controller