19

The deployment of one of my apps to a Service Fabric Cluster failed and triggered an Unhealthy Evaluation with an error event saying: There was an error during CodePackage activation.The service host terminated with exit code:3762504530

However, on the node where the app is deployed, Health State indicates: The application was activated successfully.

Is there any way to get a more detailed report on the error event?

Horia Toma
  • 1,099
  • 2
  • 17
  • 29

1 Answers1

31

I usually connect via RDP to the affected node and do the following things in such a case:

  • Check Console-Out / Console-Error logs: Service Fabric stores console output (if enabled via <ConsoleRedirection> in your ServiceManifest.xml) and errors in a log folder. On your DEV cluster, this should be C:\SfDevCluster\Data\_App\Node.x\<ApplicationTypeFolder>\log. On a default installation in Azure, it should be D:\SvcFab\_App\<ApplicationTypeFolder>\log

  • EventLog: .NET exceptions sometimes show up in the "Application" log, but Service Fabric also has its own subfolder which might contain helpful events.

  • PerfView: PerfView is a very powerful tool to monitor ETW events (Event Tracing for Windows). Since .NET exceptions are logged as ETW events, PerfView might show you helpful exceptions. Here's a quick tutorial:

    • Download and run PerfView
    • Go to "Collect -> Collect". De-Select "Merge".
    • Click "Start Collection".
    • Now kill your Service Fabric Service through Process Explorer, in case it is running. Moments later, Service Fabric will start it again.
    • If your service is not running, re-deploy your service.
    • After the service failed, press "Stop collection" in PerfView.
    • Now double-click on "Events" in the left tree - this will open all recorded ETW events.
    • Search for "Microsoft-Windows-DotNETRuntime/Exception/Start" and double click on it.
    • You should now see all .NET exceptions that occurred, ordered by time.
Christian Weiss
  • 1,060
  • 10
  • 12