Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

NPS event 6273 'Network Policy Server denied access to a user': read the reason code and find the policy

Technical article
Troubleshooting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

Event 6273 is the audit event "Network Policy Server denied access to a user." It records a RADIUS access request that NPS rejected, for example from a VPN, dial-up, or 802.1X wired or wireless connection. The event text alone does not say why. The event's own Reason Code and Reason fields do, and so does its Network Policy Name field. Microsoft's troubleshooting guidance says: "The reason code indicates the cause of the failure."

Find the failed request, read its reason code and policy name from the event, then match them against the table below. Change policy only after you know whether NPS rejected the user's credentials, found no matching policy, or could not reach the directory.

What the error means

Microsoft's audit documentation lists the NPS events in one family: 6272 "Network Policy Server granted access to a user", 6273 "Network Policy Server denied access to a user", 6274 "Network Policy Server discarded the request for a user", and 6275 to 6280 for accounting discards, quarantine, probation, and account lock and unlock. The Audit Network Policy Server subcategory generates events only if the NPS or IAS role is installed on the server, and Microsoft rates the event volume as "Medium to High" on servers running NPS.

NPS makes two decisions. Microsoft's network policy page says that during authentication NPS "verifies the identity of the user or computer", and during authorisation it "determines whether the user or computer is allowed to access the network." A 6273 can come from either.

For authorisation, NPS "checks each connection request against the first rule in the list, then the second, and so on, until a match is found." Policy conditions decide whether a policy matches. Constraints are different: "If a constraint is not matched by the connection request, NPS automatically rejects the request" and it does so "without evaluating additional network policies." If no policy matches, the configure-policies page says the request "is rejected unless the dial-in properties on the user account are set to grant access." If the user's dial-in properties deny access, NPS rejects the request.

Microsoft's current NPS troubleshooting page for supported versions of Windows Server tells you to check events 6273 or 6274 and then "the reason codes of the authentication failure events". It gives one worked example: for reason code 16 it lists a wrong username or password, an account locked in Active Directory, and a request sent to the wrong domain controller or for an account that does not exist there. Microsoft's archived page for event 6273 (written for Windows Server 2008 R2) lists the event's fields, including Reason Code, Reason, Network Policy Name, Proxy Policy Name, Authentication Type and EAP Type.

For the full numeric list, Microsoft documents reason codes in the NPS log-file reference (an archived page for Windows Server 2008 to 2012). It defines the Reason-Code field as "the reason for rejecting a user". That page is about the log file and does not mention event 6273, so treat its table as a guide and compare the log record with the event for the same request when you can.

Likely causes

Cause What you see How to confirm
Authentication failed (reason code 16, IAS_AUTH_FAILURE) Request reaches a policy but the user or computer fails to authenticate Reason-Code 16 in the log; check Authentication-Type and EAP-Friendly-Name for the method used
No network policy matched (48, IAS_NO_POLICY_MATCH) Policy-Name field is blank Microsoft: Policy-Name is blank if "none of the network policies matched"
Account state (34 disabled, 35 expired, 36 locked out, 37 invalid logon hours, 8 no such user, 33 password must change) Single user fails while others succeed Account properties in Active Directory Users and Computers
Directory unreachable (4 global catalog unavailable, 5 domain unavailable, 6 server unavailable, 7 no such domain) Many users fail at once NPS server's domain membership and reachability of a domain controller
Dial-in permission or restriction (64 dial-in locked out, 65 dial-in disabled, 66 invalid auth type, 67 to 71 invalid calling or called station, hours, port type, restriction) Policy matches but a constraint or user permission refuses Network policy constraints, and the user's Network Access Permission
Internal or request problem (1 internal error, 3 malformed request, 80 no record, 96 session timeout, 97 unexpected request) Failures not tied to one user Same code across clients; NPS and RADIUS client settings
Policy disabled or in the wrong order Intended policy never applies Policy State and position in the ordered list

Codes 2 (IAS_ACCESS_DENIED), 17, 18, 32 and 38 also appear in Microsoft's table. The page gives only the symbolic name for each, with no further explanation.

Prove the cause before changing anything

  1. Check that NPS auditing is on. In an administrative Command Prompt run:

    auditpol /get /subcategory:"Network Policy Server"
    

    Microsoft says a result of "Success and Failure" or "Failure" means auditing is enabled. If it is not, Microsoft gives auditpol /set /subcategory:"Network Policy Server" /success:enable /failure:enable.

    Risk: the /set command changes the audit policy on this server, so NPS starts writing events it did not write before. Back out by running auditpol /set again with the success and failure values the /get command showed first.

  2. In Event Viewer, open Custom views, then Server roles, then Network Policy and Access Services. Find the 6273 event and read its Reason Code, Reason and Network Policy Name fields. Note the time, user name and the RADIUS client. Microsoft's archived event page lists these fields. Match the Reason Code against the table above.

  3. Open the NPS accounting log for the same minute. Microsoft's accounting page names three logging types: event logging, local file logging and SQL Server logging. The default log directory is systemroot\System32\LogFiles, and the setup steps select the DTS Compliant format, with ODBC and IAS legacy formats as alternatives.

  4. In that record read Packet-Type. Microsoft defines 2 as Access-Accept and 3 as Access-Reject. Then read Reason-Code, Policy-Name, Proxy-Policy-Name (the connection request policy that matched), Client-Friendly-Name (the RADIUS client), Authentication-Type and EAP-Friendly-Name.

  5. Read what the combination says:

    • Reason-Code 48 and a blank Policy-Name: no network policy matched. Read the order and conditions.
    • A Policy-Name present with Reason-Code 16: Microsoft's current page ties code 16 to a failed authentication (wrong password, locked account, wrong domain controller). A policy name alongside it suggests the request reached that policy first; that reading is reasoning, not a Microsoft statement.
    • Provider-Type 2 (Provider-Name "Radius Proxy"): NPS forwarded the request to a remote RADIUS server, so the rejection may come from there. Microsoft lists Remote-Server-Address for that case.
  6. Open the NPS console, expand Policies and Network Policies. Check each policy's Policy State and order, and the conditions and constraints of the policy you expected.

  7. For a single user, check the account in Active Directory Users and Computers: enabled, not locked out, not expired, and the Network Access Permission value on the Dial-in tab.

Microsoft's archived event page lists the event's fields but shows no filled-in sample, so match your own event against that field list rather than a template.

Fix

Reason code 16, authentication failed. Check that the credentials, certificate or EAP method the client uses matches what the policy allows. Microsoft's log reference lists the Authentication-Type values (PAP, CHAP, MS-CHAP, MS-CHAP v2, EAP, None, Custom); compare the one logged with the policy's authentication methods.

Reason code 48, no policy match. Adjust the policy conditions or order so the intended policy is evaluated first.

Risk: moving a policy up or loosening a condition can grant access to requests that were correctly refused. Back out by restoring the previous order and conditions; note them before the change.

Account state codes (33 to 37, 8). Fix the account in Active Directory (re-enable, unlock, extend expiry, adjust logon hours). Risk: each change gives a person network access; confirm identity and approval first. Back out by reversing the account setting.

Directory unreachable (4 to 7). Confirm the NPS server can reach a domain controller and remains joined to the domain. These are infrastructure faults; do not change policy for them.

Dial-in permission (64, 65). Microsoft says the Access Permission page offers Access granted, Access denied, or "Access is determined by User Dial-in properties." If a user account is set to deny access, NPS rejects the request.

Disruptive step: selecting Ignore user account dial-in properties in a policy makes NPS use only the policy's own settings and skip the account's dial-in property. That changes who is allowed for every user the policy matches. Back out by clearing the check box.

Logging, if the failures are noise. Some proxies and network access servers send periodic "ping" requests with a fictional user name, which fill the logs with rejects. Microsoft documents a ping user-name registry string under HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\IAS\Parameters.

Disruptive step: editing the registry can damage the system if done incorrectly; back up first and export the key. NPS then rejects matching ping requests without processing them and does not record them. Back out by deleting the value you added.

Example (illustrative): nps01 serves staff Wi-Fi for CONTOSO through a controller wlc01 at 10.0.20.5. Three network policies exist in order: 1 "Wi-Fi Staff" (condition: Windows group), 2 "Wi-Fi Guests", 3 "VPN". A new contractor in no condition group connects and 6273 appears. The log shows Packet-Type 3, Reason-Code 48 and a blank Policy-Name, so none of the three policies matched. The administrator adds the contractor to the group used by policy 1; the next request shows Packet-Type 2 and Policy-Name "Wi-Fi Staff". Had Policy-Name read "Wi-Fi Staff" with Reason-Code 16, the administrator would have looked at the credentials instead, using the checks Microsoft lists for code 16.

Confirm the fix

  • A fresh attempt produces event 6272, and the log record shows Packet-Type 2 (Access-Accept).
  • Policy-Name names the policy you intended.
  • A test with a known bad password still produces 6273, so you know the policy still refuses what it should.
  • Check the same user from every RADIUS client that sends their requests.

What to watch afterwards

  • Count 6273 events per day by Reason-Code. A jump in code 5 or 6 across many users is a reason to check the directory path before individual accounts (advice, not a Microsoft statement).
  • Watch the accounting log's disk. Microsoft states that if RADIUS accounting fails because of a full drive or other causes, "NPS stops processing connection requests." Keep logs on a partition separate from the system partition, as Microsoft strongly recommends, and note the "When disk is full delete older log files" option.
  • After any policy change, record the order and conditions, because Microsoft's evaluation stops at the first matching policy.

Risks and limits

These commands and steps have not been run or lab-tested. Microsoft's numeric reason-code table is on an archived page written for Windows Server 2008 to 2012 and the page does not mention event 6273; newer builds may emit codes this table lacks. The event list comes from an archived Windows 10 audit page. The "What you see" column in the causes table describes the pattern to expect and is advice, not a Microsoft statement.

The If logging fails, discard connection requests setting stops NPS processing when logs are full or unavailable. Turn it on only if you accept that sign-ins stop when logging fails.

Sources