- Release Notes
- Requirements
- Installation
- Getting Started
- Projects
- Datasets
- ML Packages
- Pipelines
- ML Skills
- ML Logs
- Document Understanding in AI Fabric
- Basic Troubleshooting Guide
AI Center
5. Verify Installation
443, 8800, 31443, 31390
Recall from step 1. Provision a Machine, that you also need to allow port 6443 to be accessible from the ip of the cluster itself. Said differently, you can set network configuration rules to constraint IP source address on the rule for port 6443.
- Navigate to <linux-machine-ip or
domain-name\>:31390/ai-app, click on the three-dot menu on the top right, and
click on
Log in
- Log-in with the host tenant admin
credentials. Make sure you log in to the host tenant, the default will be
default
.
- Once logged in, you will see the list
of all Orchestrator tenants associated with your account. The first column is
whether that tenant is also provisioned in AI Fabric. Since this is the first time
you are logging in, all tenants will show
Not Provisioned
under AI Fabric. Click on the 3-dot menu for the tenant you would like to provision, and clickProvision
. After a few seconds, refreshing will show that tenant as provisioned
The application certificate needs to be trusted by the Machine where you will be using AI Fabric and the machine that will be using UiPath Robot/Studio. This step can be skipped if you used a trusted domain certificate in Step 5.
Navigate to \<machine-ip or domain-name>:31390/ai-app. This step involves adding the application certificate to your trusted certificate store.
- In Google Chome, click the padlock icon in the address bar (this will be an exclamation icon when you repeat the process for port 31443). Some steps may not appear depending on your Chrome version.
- Click the
Show connection details
arrow - Click the
More Information
button - Click the
View Certificate
orCertificate
button - Click the
Details
tab in the top navigation - Click the
Copy to File
button - Follow the Certificate Export
Wizard and export the certificate as save the certificate as base64-encoded
to some path you can easily reach later.
- Double-click that certificate and
click
Install Certificate
- Set the store location of the
certificate to your Local Machine, and place the certificate into the
Trusted Root Certificate Authorities folder
Important:- Repeat the process outline above for the following certificates:
- Certificate for Orchestrator
- Certificate for <linux-machine-ip or domain-name\))\))>:31443
Note: In this case, there will not be a padlock icon but rather an exclamation icon. The page will say "This XML file does not appear to have any style information associated with it..."
- Navigate to <linux-machine-ip
or domain-name\>:31390/ai-app, if you are still logged-in on the
host
tenant, use the 3-dot menu on the top-right corner to log out and log in on the tenant you just provisioned in Provisioning an AI Fabric Tenant. - If you see something like
Origin authentication failed
, this is due to a rarely occurring problem with permissions. To work-around this, go to Orchestrator and create a new role. Give that role AI Fabric permissions (see About AI Fabric ) and your user to that role. Navigate back to AI Fabric (<linux-machine-ip or domain-name\>:31390/ai-app), the issue should be resolved. - Follow the instructions in About Projects to quickly create a project, you can always delete this later.
- Follow the instructions in About Datasets to quickly create a Dataset. Uploading some files to the Dataset allows your to verify that the chain of trust has been established and that there are no issues the the underlying object store. If you have problems this most likely is due to a misconfigured certificate trust chain. Make sure that the three certificates in Trust the Application Certificate are in your trusted store.
- Follow the instructions in Out of the Box Packages to quickly deploy a machine learning model in your tenant. Due to the fact that some of the out-of-the-box packages are quite large and adding them to your account can fail due to connectivity issues, AI Fabric runs a periodic job that will synchronize the out-of-the-box packages of your account. If you do not have all the ones listed here, don't worry. The recurring synchronization job ensures that eventually your account will have all the models. If after a few days you still do not have all the models listed in Out of the Box Packages, contact UiPath Support.
- On the machine where you are going to use UiPath Studio to build automations, open UiPath Studio and install the UiPath.MLServices.Activities Package if it is not there already.
- Create an empty RPA workflow and add the ML Skill Activity to the workflow.
- Click on the drop-down arrow in the
ML Skill Activity, if you get an error it means that Studio cannot
securely connect to the AI Fabric back-end. The most common error is
Could not establish trust relationship for the SSL/TLS secure channel
. This is due to a misconfigured certificate trust chain. Make sure that the three certificates in Trust the Application Certificate are in your trusted store.
- Go to Orchestrator.
- Click on the `ML Skills` tab in the left navigation bar. If there is some misconfiguration, a red-error banner will appear at the top. - If there is an error, it was likely due to a misconfigured trust chain or the fact that Orchestrator cannot communicate with AI Fabric. Verify your network inbound and outbound rules. - If you open the "ML Skills tab" and you see a red banner with the message "An Error has occurred", please open `Event Viewer` to see a more detailed message.
Event Viewer
application, this can be found through
Windows search.
- Expand
Windows logs
- Click on
Application
underWindow Logs
.
Event Viewer
has a message stating
"UiPath.Orchestrator.AiFabric.AiFabricInternalException: Failed to connect to AI
Fabric. Error: IDX10638: Cannot create the SignatureProvider, 'key.HasPrivateKey' is
false, cannot create signatures. Key: [PII is hidden." proceed with the following
troubleshooting steps:
- Go to IIS Manager and open the Orchestrator Certificate used during installation.
- Click on
Details
and verify that the certificate thumbprint value exactly matches the thumprint values written in Orchestrator web.config (in step 3. Configure Orchestrator ).
Example Web.config:
...
<add key="IDP.CurrentTokenThumbprint" value="fb4e45797efd3d21021828617835d05920945091" />
<add key="IDP.PreviousTokenThumbprint" value="fb4e45797efd3d21021828617835d05920945091" />
...
...
<add key="IDP.CurrentTokenThumbprint" value="fb4e45797efd3d21021828617835d05920945091" />
<add key="IDP.PreviousTokenThumbprint" value="fb4e45797efd3d21021828617835d05920945091" />
...
Event Viewer
has a message stating "Keyset does not Exist’"
proceed with the following troubleshooting steps:
- In the Orchestrator Windows
server, go to
Run
and entermmc.exe
- Click on File ->
Add/Remove Snap-in
- Add
Certificates
in the wizard that opens up and clickOk
- In the wizard that opens up
select the radio button
Computer Account
and hitNext
- In the final wizard screen select
the ratio button
Local computer
and click Finish.
Now open the certificate manager:
- In Windows program search, type
Manager Computer Certificates
- Right click on the Orchestrator
certificate in the
Personal
Store and clickManage Private Keys
under theAll Tasks
menu.
- Click on add. This will enable
you to add a new user to the permissions list.
- In the wizard that opens, Enter
"IIS AppPool\<AppPoolName>" replacing
<AppPoolName>
with your Application Pool. For example:
- Click Ok, and run iisreset from PowerShell
Download and installation of out-of-the-box ML Packages (including Document Understanding) will take between 45 minutes to 3 hours depending on your internet bandwidth. The process that uploads these ML packages to your cluster runs independently of the application installer.
- Provision an AI Fabric tenant
- Trust the application certificate
- Verify the application installation
- Verify if ML Skills activity can communicate with AI Fabric
- Verify if Orchestrator can communicate with AI Fabric
- Error: Mismatched thumbprint values
- Error: Keyset does not exist
- Verify out of the box ML packages