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

NetLeverage
NetLeverage

The revocation function was unable to check revocation because the revocation server was offline (0x80092013)

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.

Error 0x80092013 means a Windows machine tried to confirm that a certificate had not been revoked and could not obtain a usable certificate revocation list (CRL). For a DirectAccess IP-HTTPS connection Microsoft lists four causes, and the same CRL retrieval applies to any other certificate check: the CRL location is unreachable, the CRL location is unpublished, the CRL has expired and a new one was not published, or the CRL is reachable but the client is picking up an old cached copy.

To tell them apart, run certutil -verify -urlfetch against the failing certificate, from the machine that reports the error. The output shows which address was tried and whether a CRL came back.

What the error means

Microsoft's COM error table defines 0x80092013 as CRYPT_E_REVOCATION_OFFLINE: "The revocation function was unable to check revocation because the revocation server was offline." The code 0x80092012 (CRYPT_E_NO_REVOCATION_CHECK) reads "The revocation function was unable to check revocation for the certificate." The two are different. 0x80092013 says a check was attempted and the source of revocation data was out of reach or stale.

When Windows validates a certificate it builds the chain, then checks revocation from the root CA certificate down to the certificate being evaluated. The CRL address comes from the CRL distribution point (CDP) extension inside the certificate. Microsoft states that Windows clients try the URLs in that extension in order until they retrieve a valid CRL.

A CRL has a limited life. Microsoft separates the publishing interval from the validity period and says overlap should give clients time to fetch a successor before the previous CRL becomes unusable. A CRL that was generated on the CA but never reached the web server still looks expired to every client.

Microsoft's DirectAccess page shows the error in netsh interface http show interface output, where the IP-HTTPS connection fails because the revocation check fails. Any other feature that validates a certificate with a CDP the client cannot satisfy depends on the same retrieval.

Likely causes

Cause What you see How to confirm
CDP address unreachable from the client -urlfetch reports a retrieval failure for the address in the certificate. Other networks may work. Fetch the exact URL from the failing network. Check DNS, proxy, firewall, sign-in prompts.
CDP unpublished: the certificate points at a URL nothing serves Every client fails on every network Compare the CDP extension in the certificate with the entries from Get-CACrlDistributionPoint on the CA, and fetch the certificate's own URL
CRL expired: the CA made a new one but it never arrived at the endpoint, or no new one was made The URL answers, but the file on it is past its next-update time File date on the web server, then -urlfetch output
CRL files not allowed to be served The URL answers with an HTTP error or a block File system permissions on the CRL and delta CRL files; double escaping and MIME mapping on the web server
Stale client cache One machine fails; a new machine works Cache clear fixes it, and -urlfetch from the same machine succeeds afterwards

Microsoft's DirectAccess page lists causes of an unreachable CDP: a system-context proxy that is unreachable or needs sign-in, a hotspot sign-in that is pending, a CRL location not available on the internet, a CRL location that wants authentication, and CRL files that are not allowed to be served.

Prove the cause before changing anything

These checks only read.

1. Get the failing certificate into a file. From the certificate's store, Export-Certificate writes the public part only. The Microsoft store page gives this form, with THUMBPRINT replaced by the real value:

$certificate = Get-ChildItem -Path 'Cert:\LocalMachine\My\THUMBPRINT'
Export-Certificate -Cert $certificate -FilePath '.\server.cer' -NoClobber

2. Fetch the revocation data from the failing machine.

Copy the exported server.cer to the machine that reports the error, because the next step must run there.

certutil -verify -urlfetch .\server.cer

Microsoft documents -urlfetch as "Retrieve and verify AIA Certs and CDP CRLs", and the AD CS page says to confirm the output shows a successful retrieval of the base CRL and, when delta CRLs are used, of the delta CRL. The command contacts the addresses the certificate carries, so run it from an approved troubleshooting machine. The Microsoft pages do not print sample output for this command. Read the result in words: a good run retrieves each CRL and ends without a revocation error; a bad run shows a failed or expired retrieval against one named address.

3. Check the proxy in the system context. The DirectAccess page tells you to read these values under HKEY_USERS\.DEFAULT and HKEY_USERS\S-1-5-18: ProxyEnable, ProxyServer, ProxyOverride and AutoConfigURL.

reg query "HKEY_USERS\.DEFAULT\Software\Microsoft\Windows\CurrentVersion\Internet Settings" /v ProxyEnable

4. On the CA, record how it publishes.

Get-CACrlDistributionPoint

Microsoft notes that PublishToServer and AddToCertificateCdp are separate settings. Where the CA writes a CRL and the address clients are told to use can differ.

5. Check the endpoint's file. Microsoft says to verify at each HTTP endpoint that the expected CRL file is current, complete and readable by the intended clients. Look at the file's modified time and compare it with the CA's last publication.

6. Look for CA publication events. Microsoft's list of AD CS events includes 65 and 66 (could not publish a Base or Delta CRL to a location), 67 (made attempts and will stop until the next CRL is generated), 130 (could not create a CRL) and 68 to 73 (published successfully). The page applies to Windows Server 2003 SP2 through 2012 R2, so confirm the numbers in your own log.

Fix

Unreachable address. Repair the path, not the certificate: DNS, firewall, proxy exclusion or authentication on the web server. For machines that are not on the internal network, an internet-reachable URL has to be in the certificate; Microsoft's DirectAccess page marks its internet CRL location "WAN URL required Internet Access". A CDP change only affects certificates issued after the change, so an older certificate keeps its old address. Keep the old endpoint and current CRL there until no unexpired certificate refers to it.

Risk: a DNS, firewall, proxy or web-server authentication change affects everything else that uses that path. Impact: a wrong rule can cut off other clients or open the endpoint wider than intended. Back-out: record the setting before the change and restore it if the -urlfetch result does not improve.

Unpublished address. For the certificate that is failing now, make the address it carries serve a current CRL. Microsoft's page says: "Keep the old CDP online and restore current CRL content there. A CDP configuration change doesn't repair old certificates." Adding a new retrieval location to the CA helps only certificates issued after the change, which carry the new address. Use the substitution tokens Microsoft shows:

Add-CACrlDistributionPoint `
  -Uri "http://pki.contoso.com/pki/<CAName><CRLNameSuffix><DeltaCRLAllowed>.crl" `
  -AddToCertificateCdp `
  -AddToFreshestCrl

Disruptive step: this changes CA configuration. Microsoft says that if the command reports the CA requires a restart, you restart the CA service in an approved change window. The change affects only certificates issued afterwards. Record Get-CACrlDistributionPoint first. To back out, remove the entry with Remove-CACrlDistributionPoint using the same flags, and run it with -WhatIf first. Without the flags it removes every entry matching the address.

Expired CRL. Publish a new CRL, then fix the delivery chain so the next one arrives on time:

certutil -config "CAHOST\Contoso Issuing CA" -CRL

Disruptive step: this writes new CRL data on the CA according to its configuration. Microsoft documents no way to withdraw a CRL once it is published, so confirm you are on the right CA first. It does nothing for the web server: Microsoft says certutil -CRL "doesn't prove that a web server, replication process, or external transfer makes the new files available". Copy or replicate the file, then repeat step 2.

Risk: copying the new CRL file over the one on the web server replaces what every client now fetches. Impact: a wrong or incomplete file makes the failure worse for all clients. Back-out: keep a copy of the file you replaced and put it back if the new one does not verify.

Files not allowed to be served. Correct the file permissions on the CRL and delta CRL files, the web server's .crl MIME mapping, and double escaping for the CDP location. The DirectAccess page names the IIS request filtering setting allowDoubleEscaping for this. Risk: a web server change affects every site on it. Note the previous value to restore it.

Stale cache. Microsoft's DirectAccess page gives two commands: certutil -setreg chain\ChainCacheResyncFiletime @now for time-validated objects, and certutil -urlcache * delete for the URL cache.

certutil -urlcache * delete

Disruptive step: the certutil reference says delete "deletes relevant URLs from the current user's local cache", so this clears every cached URL of the user who runs it, not only CRLs. If the CDP is still unreachable, that user now has no cached copy at all. There is no way to restore deleted cache entries other than downloading them again. A failure in the system context (a service, or IP-HTTPS) needs the commands run in that context; the DirectAccess page shows a psexec.exe -s example for a registry change in that context.

certutil -setreg chain\ChainCacheResyncFiletime @now

Risk: -setreg writes a registry value, which Microsoft lists as the way to clear the time-validated object cache. Impact: that machine drops its cached time-validated objects and fetches them again. Microsoft documents no undo for it.

Confirm the fix

Run certutil -verify -urlfetch .\server.cer again from the same failing machine and repeat the original action, such as the RDP connection or the netsh interface http show interface check. Test from each network zone that uses the certificate. Then repeat with a certificate issued before any CDP change, to prove older certificates still reach their original address.

Example (illustrative): CA ca01 in the CONTOSO domain publishes a base CRL with a seven-day validity. On Tuesday 3 November at 02:00 it creates a CRL with a next-update time of Tuesday 10 November at 02:00 (3 plus 7 days). The copy job to web server web01 fails on 3 November, so web01 still serves the CRL from 27 October. That file expired on 3 November at 02:00. Every client that checks revocation after that gets 0x80092013, even though ca01 holds a good CRL. Running certutil -CRL on ca01 changes nothing at web01. Repairing the copy job and copying the new file does.

What to watch afterwards

  • Alert before the next-update time, not after it. Microsoft says the response procedure should identify whether the fault is in CA generation, the publication path, the copy or replication process, network reachability or the retrieval service.
  • Check the CRL file's date at every endpoint, not on the CA only.
  • Watch for CA events 65, 66, 67 and 130 on versions where they apply.
  • Add any new network zone, such as a branch site or guest network, to the -urlfetch test list.

Risks and limits

These commands have not been run or lab-tested.

certutil is an administration and diagnostic tool. Microsoft states that it is not meant for production application code. The CA event list comes from a Microsoft page that covers Windows Server 2003 SP2 to 2012 R2, so treat its numbers as a lead on newer versions. The Microsoft pages do not document sample -urlfetch output, so descriptions of good and bad runs here are in words. Deleting a cache or republishing a CRL cannot be reversed. This guidance covers CRL retrieval, not OCSP responders, and it does not cover replacing a certificate after its private key is exposed.

Sources