Microsoft Graph API intermittent error "Token not found: token is either invalid or expired" resolves itself after retry

Question

I am making requests to the Microsoft Graph REST api (beta), specifically to the sign-in event endpoint: https://graph.microsoft.com/beta/auditLogs/signIns

I am making batch requests that retrieve sign-ins for a particular user in batches of 1000. Starting around the 25th of september, these requests would fail after roughly 10-50 batches, with the following response, with a 400 (Invalid Request) HTTP error code:

"error": {
  "code": "",
  "message": "Token not found: token is either invalid or expired",
  "innerError": {
    "request-id": "[request-id-redacted]",
    "date": "2019-09-30T22:27:36"
  }
}

However, if I retry the requests after waiting ~1s, with the very same JWT web token, the requests succeed and I'm able to complete all the batch requests for the job I'm running. The access token I receive when initially authenticating expires in 1 hour, but this error crops up ~1-15 minutes after I receive the token (I've confirmed the unix timestamp expiry date I get w/ the token).

I'm wondering what could be the cause of this error, and how I could avoid it, other than hard-coding the specific response message and retrying. I was unable to find any matches on google for the error message, either. Has anyone seen this error before from the Microsoft Graph API?

Answer 1

"I was unable to find any matches on google for the error message, either. Has anyone seen this error before from the Microsoft Graph API?"

Yes. I'm querying the https://graph.microsoft.com/beta/reports/credentialUserRegistrationDetails MS Graph endpoint, and I'm seeing exactly the same thing you are. Intermittent responses with status code 400 and this error message: "Token not found: token is either invalid or expired", only a minute after requesting an access token.

I'm using the @odata.nextLink value to page through the results, however it's proven to be very unreliable due to this issue. Sometimes I will get the entire set of results, other times I will only get a portion of it. Even when I sleep for 5 seconds between each nextLink request, the error still appears.

This started happening for me this week. Before I was getting consistent results.

Your approach to perform a retry might be the best one for now:

if response.status_code == 400:
  retry_request()