- Release Notes
- Requirements
- Installation
- Getting Started
- Projects
- Datasets
- ML Packages
- Pipelines
- ML Skills
- ML Logs
- Document Understanding in AI Fabric
- Basic Troubleshooting Guide
4. Run the AI Fabric Application Installer
- Cloud shell console logs will have
the URL to access Kots Admin console, so login to the console via the same URL. If
you have opted not to expose Kots via Load Balancer, then set the kube context as
shown in this section “How to set Kubernetes Cluster Context” followed by triggering
the below
command.
kubectl -n aifabric port-forward service/kotsadm 8800:3000
kubectl -n aifabric port-forward service/kotsadm 8800:3000 - Now the KotsAdmin will be accessible
via: http://localhost:8800
- If you kotsAdmin password isn’t
working or if you want to reset
it.
Option 1: if you have a linux machine or WSL enabled on your windows, then Install Kots CLI: curl https://kots.io/install | bash Reset password: kubectl kots reset-password -n aifabric or login to cloud shell cd aks-arm ./kots reset-password -n aifabric
Option 1: if you have a linux machine or WSL enabled on your windows, then Install Kots CLI: curl https://kots.io/install | bash Reset password: kubectl kots reset-password -n aifabric or login to cloud shell cd aks-arm ./kots reset-password -n aifabric
Upload the license provided to you by your UiPath representative via the UI as below.
You must enter the details per the screen below.
To obtain Identity Server access token:
The IS access token can be found by login in with the tenant “host“(opposed to tenant “default“) as admin user at the following address “https://<IdentityServerEndpoint>/identity/configuration”. This address should already be known by the customer because here is the place where he can configure external identity providers, such as Azure AD, Windows, or Google.
Copy the token and paste it into replicated installation console.
Enable HA for Core Service
Enabling this will ensure that 2 replicas will be always running for AI Fabric core services along with horizontal pod scaling enabled i.e based on workload if required, no of pods related to core services will be auto scaled/down scaled accordingly. If we don’t enable HA, then only one replica of core service will be running always but horizontal pod scaling is still enabled which ensures that if need arises pods will be auto scaled for short duration.
Enable HA for CPU Based ML Skills
Enabling this will ensure that we deploy 2 replicas for all CPU based ML Skills and these 2 replicas will be deployed across nodes present under multiple zones. If HA is not enabled only replica will be deployed.
Enable HA for GPU Based ML Skills
Enabling this will ensure that we deploy 2 replicas for all GPU based ML Skills and these 2 replicas will be deployed across nodes present under multiple zones. If HA is not enabled only replica will be deployed. Since GPU machines are quite expensive on azure, we are providing this option so that 2 nodes are not required to deploy an ML skill as GPU’s can be shared across deployments.
Configure Max CPU & Memory a GPU based job can consume
Since Standard NC6 is the smallest config machine available on Azure with GPU availability, that’s why we are defaulting max CPU to 5000 (5 CPU) and max memory to 50 GB. But if the customer is using Standard_NC6s_v2 or some other config machine instead of Standard_NC6 nodes for GPU node pool, in that scenario customer can override the defaults (i.e max CPU and Mem) a GPU based training job can consume.
On Saving the Config, KOTS will start validating the Inputs and if all the pre-flight checks passed, KOTS will trigger the deployment exactly similar to that of one box installation
If all the preflight checks are passed, config screen would look something like this. Click continue to proceed.
To initiate installation, click on Deploy button. Once the status is Deployed, installation has been started and setup admin can go to Application tab to check the current status
Provisioning job status can be tracked vi querying this pod from local. Please ensure you have setup Kubernetes Cluster Context as described :
rajivchodisetti@DESKTOP-LOUPTI1:/mnt/c/Users/rajiv.chodisetti$ kubectl -n aifabric get pods | grep provision
provision-4xls7mzjpnui8j7n-s9tct 0/1 Completed 0 14h
To check the logs of this pod:
kubectl -n aifabric logs -f provision-4xls7mzjpnui8j7n-s9tct
If AIFabric deployment is successful, this is what you would see in the pod logs at the end,
Successfully setup cronjob for oob installation run on daily basis.
< Total steps: Current step: 8 Estimated time: 2s >
AiFabric in Azure AKS has been provisioned successfully
rajivchodisetti@DESKTOP-LOUPTI1:/mnt/c/Users/rajiv.chodisetti$ kubectl -n aifabric get pods | grep provision
provision-4xls7mzjpnui8j7n-s9tct 0/1 Completed 0 14h
To check the logs of this pod:
kubectl -n aifabric logs -f provision-4xls7mzjpnui8j7n-s9tct
If AIFabric deployment is successful, this is what you would see in the pod logs at the end,
Successfully setup cronjob for oob installation run on daily basis.
< Total steps: Current step: 8 Estimated time: 2s >
AiFabric in Azure AKS has been provisioned successfully
Please see here