AI Center User Guide

DELIVERY:

Last updated Nov 11, 2024

General AI Center troubleshooting and FAQs

Message: Failed to upload Item(s), it may be due to a slow or lost Internet connection

When uploading dataset files, the following error can occur:

Failed to upload item(s), it may be due to a slow or lost internet connection

Possible cause

This error message can show up because of some browser configurations.

Solution

Open browser console and get the DNS of the objectstore url. It will be of the form objectstore.xxx.xx Make sure that the objectstore DNS is resolvable either by adding to host file or talking to your network administrator Once the DNS is resolved, if the certificate is not trusted, make sure you trust the certificate inside your browser before uploading the item.

Issue: Error on Pipelines pages even though permissions are in place for running pipelines

When trying to view or run pipelines, an error can be occur, even though permissions to run pipelines are in place.

Solution

In order to run and view pipelines, Read permissions on the ML Packages are mandatory.

Issue: Service deployment can get stuck because of the DATABASECHANGELOGLOCK lock not being released by one service

On rare occasions, if you restart the machine two times consecutively, service deployment can get stuck because of the DATABASECHANGELOGLOCK lock not being released by one service. In this case you will see AI Center pods restarting continuously.

Solution

Run the following SQL command in the AI Center database to release the lock:

UPDATE DATABASECHANGELOGLOCK SET LOCKED = 0, LOCKGRANTED = null, LOCKEDBY = nullUPDATE DATABASECHANGELOGLOCK SET LOCKED = 0, LOCKGRANTED = null, LOCKEDBY = null

Issue: import/export script fails

The import/export script is failing with the following error message:

cookfile_new.txt: Permission denied

Solution

Remove the cookfile.txt and cookfile_new.txt files generated locally by the import/export script.

Message: ./export.sh: Line 2: $'r': Command Not Found

When running the import or export scripts, the following error message can occur:

./export.sh: line 2: $'\r': command not found

This error message is displayed when importing or exporting ML Packages using scripts.

Solution

Run the following command before running the import or export script:

dos2unix <filename>dos2unix <filename>

Issue: signed URL for public datasets is failing

This issue can occur when running a UiPath Studio automation and uploading validation data for training using a public dataset.

Solution

If you encounter this issue, try one of the following solutions:

Upload the package manually.
Use a private dataset so you can connect through a robot that can leverage this.

Issue: the Update-mlskills-cm cronjob is missing

The update-mlskills-cm cronjob is missing in AI Center versions 2021.10.1 and 2021.10.2.

Solution

To avoid failing ML Skills deployments and predictions because of this issue, create the cronjob manually by using the YAML file below.

apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: update-mlskill-cm
  namespace: uipath
spec:
  concurrencyPolicy: Forbid
  failedJobsHistoryLimit: 1
  jobTemplate:
    spec:
      template:
        metadata:
          annotations:
            sidecar.istio.io/inject: "false"
        spec:
          containers:
          - args:
            - -XPOST
            - ai-deployer-svc.uipath.svc.cluster.local/ai-deployer/v1/system/mlskills:update-cm
            image: registry.uipath.com/aicenter/alpine-curl:7.78.0
            imagePullPolicy: IfNotPresent
            name: update-mlskill-cm
            securityContext: 
              allowPrivilegeEscalation: false
              capabilities:
                drop:
                - NET_RAW
              privileged: false
              readOnlyRootFilesystem: true
              runAsNonRoot: true
          dnsPolicy: ClusterFirst
          imagePullSecrets:
          - name: regcred
          restartPolicy: OnFailure
          schedulerName: default-scheduler
          securityContext: {}
          terminationGracePeriodSeconds: 30
      ttlSecondsAfterFinished: 120
  schedule: 0 */2 * * *
  startingDeadlineSeconds: 200
  successfulJobsHistoryLimit: 1
  suspend: falseapiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: update-mlskill-cm
  namespace: uipath
spec:
  concurrencyPolicy: Forbid
  failedJobsHistoryLimit: 1
  jobTemplate:
    spec:
      template:
        metadata:
          annotations:
            sidecar.istio.io/inject: "false"
        spec:
          containers:
          - args:
            - -XPOST
            - ai-deployer-svc.uipath.svc.cluster.local/ai-deployer/v1/system/mlskills:update-cm
            image: registry.uipath.com/aicenter/alpine-curl:7.78.0
            imagePullPolicy: IfNotPresent
            name: update-mlskill-cm
            securityContext: 
              allowPrivilegeEscalation: false
              capabilities:
                drop:
                - NET_RAW
              privileged: false
              readOnlyRootFilesystem: true
              runAsNonRoot: true
          dnsPolicy: ClusterFirst
          imagePullSecrets:
          - name: regcred
          restartPolicy: OnFailure
          schedulerName: default-scheduler
          securityContext: {}
          terminationGracePeriodSeconds: 30
      ttlSecondsAfterFinished: 120
  schedule: 0 */2 * * *
  startingDeadlineSeconds: 200
  successfulJobsHistoryLimit: 1
  suspend: false

Disabling streaming logs

Versions up to 2021.10.4

To disable log streaming on existing skills, edit the skill deployment and change the LOGS_STREAMING_ENABLED environment variable to false. You can also add a logsStreamingEnabled global variable with the value set as false using ArgoCD under the aicenter app details. Make sure to sync ArgoCD after the change is done.