Import - error handling

[Danielprobst22]

An error message will be printed and non-zero exit code returned in case of an error.

Notes and questions:

• During one import, multiple deployments may be present on the device
  • Should the import process be aborted completely in case of an error?
  • Or should the import for the current deployment be skipped and the import be continued with the next deployment?
• What should happen with the output directory (i.e. the directory to which the data is being copied, not the directory on the Raspberry Pi/ SSD)
  • Should the entire output directory be deleted?
  • Should the output directory for the failed deployment be deleted?
  • Should the content of the output directory for the failed deployment be deleted and a log file with the error cause placed in that directory?
• How should the user be able to retry a failed import
  • Should the CLI retry the import automatically a given number of times without any user interaction?
  • Should the user be able to retry the import into the same output directory?
  • Should the user be forced to always pick a new output directory (i.e. let the CLI fail when the output directory already exists)

[Danielprobst22]

Suggestion:

• If importing a deployment fails, delete the files produced so far from that deployment's directory and create a log with the failure cause, then continue with the remaining deployments
• Let the user retry the import into the same output directory, list all the deployments that were found on the Raspberry Pi and infer their state based on the output directory:
  • Non-empty deployment directory -> already imported succesfully
  • Deployment directory contains error log file -> failed import
  • Deployment directory doesn't exist -> new deployment
• Before starting the import, prompt the user how each of the deployments should be handled
  • If a successful import was detected, suggest to the user to skip that deployment by default (i.e. just pressing enter without having to type anything and in that case keeping the data in the deployment's output directory that was produced in the previous run) but allow the user to override the behavior and make the cli reimport that deployment (i.e. deleting the deployment's output directory and starting from scratch)
  • If a previous error was detected, suggest to retry that deployment by default, but make it possible to skip that deployment (i.e. leave the previous error log in the deployment's output directory)
  • If a new deployment was detected, suggest to import that deployment by default, but make it possible to skip that deployment (i.e. not create a directory for that deployment at all)
• After the user chose how to handle each deployment, start processing them one by one and produce a summary at the end with the result for each deployment

[Danielprobst22]

See also the following crates:

• https://github.com/dtolnay/anyhow
• https://github.com/dtolnay/thiserror