2.2.5. Investigating and Resolving Authentication Issues
💡 First Principle: Every sign-in leaves a record of every decision made about it — which method was used, which Conditional Access policies matched, and which control failed. Troubleshooting isn't guessing; it's reading the record, and simulating the sign-in when the record doesn't exist yet.
Sign-in logs (Microsoft Entra admin center, under Monitoring & health > Sign-in logs) are the first stop. They come in four categories: interactive user, non-interactive user (token refreshes and apps acting for a user), service principal and managed identity. Each entry shows the result, an error code and failure reason, the authentication details (which methods were attempted and satisfied), and a Conditional Access tab listing each policy as applied, not applied, or report-only. A result of "Report-only: Failure" means the policy's conditions matched and its control wasn't met — but the sign-in wasn't blocked, because the policy is only reporting.
Retention matters: sign-in logs are kept 7 days on Entra ID Free and 30 days on P1/P2. To keep them longer — or to query them alongside other data — route them to Azure Monitor (a storage account or a Log Analytics workspace). The Conditional Access insights and reporting workbook needs a Log Analytics workspace receiving sign-in logs.
Tools for "what would happen":
- The Conditional Access What If tool simulates a sign-in for a chosen user, app and set of conditions and shows which policies would apply — without the user having to sign in. Use it before enabling a policy, or when a user reports a block you can't reproduce.
- Report-only mode gives the same answer from real traffic over days.
Common causes and fixes:
| Symptom | Likely cause | Where to look |
|---|---|---|
| User blocked unexpectedly | A Conditional Access policy matched (location, device, risk) | Sign-in log > Conditional Access tab; What If |
| Prompted for MFA repeatedly | Sign-in frequency session control, or MFA claims from a partner tenant not trusted | Policy session controls; cross-tenant trust settings |
| Can't complete MFA | No usable method registered, or a method shown as non-usable (expired TAP, key failed attestation) | User's authentication methods |
| Account locked | Smart lockout after repeated failures | Sign-in log error code; lockout settings |
| Legacy app can't sign in | Legacy authentication blocked by security defaults or Conditional Access | Client app shown in the sign-in log |
⚠️ Exam Trap: If a user "can't sign in, but the sign-in log shows the policy as report-only failure," that policy didn't cause the block — look at the policies that were actually applied. And if the incident is older than your log retention (7 days on Free), the sign-in log can't answer the question at all.
Reflection Question: A user reports being blocked from Exchange Online last Tuesday, but you can't reproduce it today. Which two tools would you use, and why might one of them return nothing?