Performing AI analysis on failed deployments

You can perform AI analysis on failed deployments to identify the root cause of the failures and troubleshoot them easily.

You must have completed the following tasks:
When you run a deployment process, it might fail at a step or multiple steps for some reason. You can troubleshoot the failed steps by using the output log files. By using the Deployment Genie feature, you can get insights such as an AI interpreted root cause analysis and action items intended to ensure a failure doesn’t happen again.
Attention: You must consider the following points before you perform the AI analysis -
  • The recommendations from the configured AI service may be inaccurate or contain errors. If suggested solutions for an error do not seem logical or effective, open a support case to investigate the issue with Deploy.
  • IBM Support does not resolve problems related to your configured AI service as it is a third-party utility beyond our control.
  • Do not apply suggestions that involve deleting files and folders, and executing SQL queries that modify the database.
  1. Go to the Log tab of a deployment request.
    The Log tab displays all steps of the deployment process.
  2. Click Deployment Genie.
    The Log tab displays only the steps that failed in the deployment process.
    Note: You can also use the Deployment Genie option from the menu options of a failed step.
    Troubleshooting: When the Deploy server runs as a container in a Kubernetes cluster and you invoke the Deployment Genie, an alert dialog may display stating that a timeout was encountered. To resolve this timeout issue, modify the Emissary Ingress mapping resource for Deploy to change the timeout_ms value as follows:
    1. Enter the following command:
      kubectl edit mapping <releaseName>-ibm-ucd-prod
    2. Set the timeout_ms value to 0.
    3. Save your changes.
  3. Start the AI analysis of a failed step by clicking the Deployment Genie menu option.
    The logs are sent to the AI service for analysis. The AI Analysis window displays the root cause and its probable solution in a human-readable format.
    Note: The sensitive data such as hostname and IP addresses are not sent to AI service for analysis.
  4. Perform the AI analysis for the remaining failed steps as needed.
You have performed the AI analysis of a failed deployment process.