General Process And Challenges While Installing AI Center 23.4.2 With External Orchestrator

General Process and Challenges while installing AI Center 23.4.2 with External Orchestrator.

Scenario: This scenario is for an AI Center v23.4.2 offline installation via Automation Suite with an external Orchestrator, and seeks to act as a guide to the general process and challenges/bugs that may be encountered during such a scenario.

Prerequisites: The User performing installation should take care to ensure all prerequisites like disk configurations and DNS records stipulated in UiPath documentation have been met.

General Process:

Assuming an offline upgrade, the detailed steps are listed in UiPath documentation (Installation Guide:Offline Single Node Installation , Installation Guide: Offline Multi Node Installation)

The important steps are:

  1. Configure Orchestrator (AI Center - Automation Suite 2023.4 - User Guide For Configuring Orchestrator)
  2. Copying necessary files or bundles onto the node(s)
  3. Specifying an external orchestrator, Orchestrator and Identity URLs, and certificate files during the interactive installer process
  4. Install infrastructure components one first server node, and join other nodes to the cluster if applicable
  5. Upload offline bundles using internal Docker registry
  6. Install Fabric and Services on server nodes
  7. Load Document Understanding bundles, and complete the AI Center installation:
  • ./configureUiPathAS.sh aicenter configure --installation-token

Bugs/ Challenges Expected:

  • Preflight Checks Using Wrong Image: Installation progress can be monitored via the ArgoCD ALM tool once that component has been installed.
  • While monitoring the applications from ArgoCD, it can be seen that the aievents application spins up pods with the “aievents-preflight-check-xxxx” and “aievents-register-dmclient-xxxx” naming convention.

  • These pods attempt to pull the image aicenter-jobs-23.4.1-rc1 and will fail with error “Back-off pulling image "localhost:30071/aicenter/aicenter-jobs:23.4.1-rc1” as the needed image is aicenter-jobs-23.4.2-rc1.”.

To fix this problem,
  1. Click on the failed pod and edit the pod configuration
  2. Search for the aicenter-jobs:23.4.1-rc1 and
  3. Amend it to aicenter-jobs:23.4.2-rc1 and save. This step needs to be performed on both pods, and AI Center installation will only continue upon successful completion.

Document Understanding Issue due to CORS Validation

For installs where AI Center and Documentation Understanding are installed, it is likely that Document Understanding fails to complete deployment due to some pods failing with errors indicating attempting CORS validation.

  • As of AS 23.4.2, the quickfix for this issue is to amend the Kubernetes network policy to be more accomodating. Copy and paste the following block onto the Linux prompt, and the issue should resolve itself.

kubectl apply -f - <<EOF

apiVersion: networking.k8s.io/v1

kind: NetworkPolicy

metadata:

name: allow-all-ingress

namespace: uipath

spec:

podSelector: {}

ingress:

- {}

policyTypes:

- Ingress

---

apiVersion: networking.k8s.io/v1

kind: NetworkPolicy

metadata:

name: allow-all-egress

namespace: uipath

spec:

podSelector: {}

egress:

- {}

policyTypes:

- Egress

EOF