Troubleshooting¶

Common pitfalls¶

Symptom	Cause and fix
`ValueError: No existing parameters found for table ...`	`update_parameters` needs existing parameters. Call `advise_parameters` or `set_parameters` first.
My recommended parameters disappeared	`set_parameters` replaces all the parameters. Use `update_parameters` to change only some of them. See Avatarization parameters.
No avatars, only metrics	`set_parameters` was called without `k` (or a DP epsilon).
`ValueError: Expected either k or fast_dp_epsilon to be set, not both`	After `advise_parameters`, pass `k=None` with the epsilon. See Augmentation, differential privacy and pseudonymization.
`imputation_return_data_imputed=True` has no effect	Pass `imputation_method` with it. See Avatarization parameters.
`ValueError: No jobs found for display name ...`	No run has this exact name (case-sensitive). For a web job, run it in the web application first.
I get the results of an older run	Several runs have the same name: the most recent one is used. Give unique names, or use `create_runner_from_id`. See Reuse past runs and configurations.
`UserWarning` when I call `run()` again	`run()` starts a new run. The warning gives the identifier of the previous run. Use `ignore_warnings=True` to hide it.
The server does not find the data of a replayed YAML	The file name in the YAML (`file: <table>.csv` or `.parquet`) must match the extension of the uploaded data. See Reuse past runs and configurations.
`ImportError` on `ReportLanguage` or `JobKind`	Import `ReportLanguage`, `ReportType` and the other parameter enums from `avatar_yaml.models.parameters`, and `JobKind` from `avatars.models`.
The plot does not show in my notebook	Use `runner.render_plot(..., open_in_browser=True)`.

Understanding errors¶

Most of your actions succeed. This section explains the errors that can happen, and how to correct them.

Timeout("The call timed out. Consider increasing the timeout with the 'timeout' parameter.")

The call takes too long on the server, usually during a job or the upload or download of a dataset. See Handling timeouts.
Validation errors

Validation errors come from a bad input. The error messages use HTTP status codes: codes from 400 to 499 are user errors, codes from 500 to 599 are server errors.

You can correct a user error by changing your parameters and trying again. The error message has this form:
```
Got error in HTTP request: POST https://company.octopize.app/reports.
Error status 400 - privacy_metrics job status is not success
```
It gives:
- the HTTP method (POST, GET, …);
- the endpoint (/reports);
- the status (400);
- a message that explains the error (privacy_metrics job status is not success).
In this example, the report needs a successful privacy metrics job, but the privacy metrics job failed. Read the error of the privacy metrics job (see below), correct the input, and run again.
Failed jobs (JobStatus.error)

run() does not wait for the jobs, so it does not raise when a job fails on the server. The result methods (shuffled, privacy_metrics, …) raise ValueError: Job <name> failed with exception: ... when the job failed. To diagnose the failure, read the status and the exception property of the job:
```
from avatars.models import JobKind

job = runner.get_job(JobKind.standard)
print(job.status)  # JobStatus.error
print(job.exception)  # explanation of the error
```
Correct your input and run the job again.
Internal errors

An internal error is an unexpected error on the server. It comes with a 500 HTTP status code and the message internal error. Try again, if possible with different parameters.

Our error monitoring notifies us of these errors at once. Contact your Octopize contact or support@octopize.io for help while we investigate.

Handling timeouts¶

Asynchronous calls¶

Many endpoints of the Avatar API are asynchronous: you ask for something that runs in the background, and you get the result later with another method, for example runner.shuffled or runner.get_all_results after runner.run.

You do not need to wait for the job to finish. You can close your script, come back later, reload the run and get the results:

from avatars.models import JobKind

runner.run(jobs_to_run=[JobKind.standard, JobKind.privacy_metrics, JobKind.signal_metrics])
runner.statuses()  # the jobs are probably still queued or running

# Take a coffee break, close the script, come back in 3 minutes
runner = manager.create_runner_from_name("my_run")  # in a new session
runner.statuses()  # {'standard': 'finished', ...}
avatars_df = runner.shuffled("my_table")