Fix it now
These come from the Windows security layer rather than from Exchange, and each names a different point in the same check. 0x800B0109 says the chain ended at a root the machine does not trust. 0x800B0101 says a certificate is outside its validity dates. 0x80090327 is the unclassified one. None of them is fixed by turning encryption off.
Get-ExchangeCertificate | Format-List Thumbprint,Subject,Issuer,NotBefore,NotAfter,Status,HasPrivateKey,Services
certutil -verify -urlfetch C:\temp\cert.cer
certutil -repairstore my <thumbprint>
w32tm /query /status
- Anything whose
Statusis not valid, or whereHasPrivateKeyreads false, is your candidate before you look at anything else. - Read the
certutil -verify -urlfetchoutput for which link fails.-urlfetchretrieves and verifies the issuing certificates and revocation lists named in the certificate, which is what turns a vague trust failure into a specific missing intermediate. - If an intermediate is missing, import it into the local machine’s intermediate certification authorities store, not the current user’s, then re-run the verification.
- Only run
certutil -repairstorewhen the key association is what is broken. It repairs the association between a certificate and its private key rather than replacing the certificate. - Re-apply the service bindings with
Enable-ExchangeCertificate <thumbprint> -Services SMTP, then retest the session.
Do not fix this by disabling TLS, lowering the required protocol version or switching off certificate validation. Those changes are visible to every partner you exchange mail with, are hard to reverse once applications depend on them, and leave the original fault in place.
If the chain builds cleanly and the session completes, you are done. If it does not, the next section takes the four checks in turn.
Why it happens
When a TLS session starts, the receiving side is handed a certificate and has to decide whether to trust it. It builds a chain from that certificate up through any intermediate authorities to a root, checks the root is in its trusted roots store, checks every certificate in the chain is inside its validity dates, and checks none has been revoked. Each of those checks has its own failure code, which is why the codes are worth reading rather than treating as one generic certificate error.
An untrusted root result rarely means the root is genuinely unknown. Far more often an intermediate certificate is missing from the server, so the chain cannot be completed and the layer reports the highest thing it could reach as untrusted. Installing the intermediate is what fixes it, and that is why validating the chain matters more than staring at the leaf certificate.
0x80090327 is the one that does not tell you which check failed. Windows publishes it as an unknown error while processing the certificate, so treat it as a signal to run the chain validation rather than as a diagnosis in itself. 0x80090304 is different again and often misread: it is the security layer reporting that the local security authority cannot be contacted, which points at the machine’s own security subsystem rather than at anything in the certificate. If you see that one, stop examining certificates.
An intermediate certificate is missing from the server
You have this one if 0x800B0109, and certutil -verify -urlfetch reports the chain cannot be built above the leaf.
- Download the issuing authority’s intermediate certificate from the authority that issued yours.
- Import it into the local machine’s intermediate certification authorities store, not the current user’s.
- Re-run
certutil -verify -urlfetchand confirm the chain now completes to a trusted root. - Retest the session. No Exchange change is normally needed once the chain is valid.
Install it on every server that presents or validates the certificate, including ones that only handle transport.
The certificate is expired or not yet valid
You have this one if 0x800B0101, and NotAfter is in the past or NotBefore is in the future.
- Check the server clock as well as the certificate. A server whose time has drifted far enough rejects perfectly good certificates.
- Synchronise time with
w32tm /resyncand confirm withw32tm /query /status. - Install the replacement if the dates really have passed, keeping the same subject and alternative names.
- Enable the replacement for the services the old one served, then retest.
/force is not a parameter of w32tm /resync. Given an unrecognised switch, w32tm prints its usage block and exits without resynchronising anything.
The private key is missing or its association is broken
You have this one if HasPrivateKey reports false, or the certificate looks perfect but session setup fails on this server only.
- Confirm the state with
Get-ExchangeCertificate | Format-List Thumbprint,HasPrivateKey,Status. - If the key is present but not associated, repair the association with
certutil -repairstore my <thumbprint>. - If the key is genuinely absent, re-import the certificate from the original PFX so the key comes with it.
- Run
Enable-ExchangeCertificateagain for the services that use it, which re-applies the permissions those services need on the key.
Exporting a certificate without its private key and importing it on another server is the most common way to produce this exact state.
Revocation checking cannot complete
You have this one if Failures are intermittent or slow, and appear only on servers without outbound internet access.
- Check whether the server can reach the revocation endpoints published in the certificate.
certutil -verify -urlfetchreports these explicitly. - If outbound access is filtered, permit the revocation endpoints rather than switching revocation checking off.
- Confirm the proxy configuration used by the system account, which is not necessarily the one your signed-in session uses.
- Retest and confirm the delay disappears as well as the error.
The local security authority cannot be contacted
You have this one if 0x80090304, and the certificate itself validates cleanly when you check it.
- Stop investigating the certificate. This code is about the machine’s security subsystem, not the chain.
- Check whether the server is under memory or handle pressure, and whether it has recently been rebooted.
- Check the system log for security package and credential errors around the same timestamps.
- Confirm the server’s domain membership and secure channel are healthy before assuming it is a TLS problem at all.
Full reference
What each code actually names
| Code | Constant | Published meaning |
|---|---|---|
0x80090327 |
SEC_E_CERT_UNKNOWN | An unknown error occurred while processing the certificate |
0x800B0109 |
CERT_E_UNTRUSTEDROOT | A certification chain processed correctly but terminated in a root certificate that is not trusted by the trust provider |
0x80090304 |
SEC_E_INTERNAL_ERROR | The Local Security Authority cannot be contacted |
0x800B0101 |
CERT_E_EXPIRED | A required certificate is not within its validity period |
0x800B010F |
CERT_E_CN_NO_MATCH | The certificate’s CN name does not match the passed value |
0x800B010F is worth recognising even though it is not one of this article’s codes: a name mismatch is what a wildcard certificate produces when the name being connected to is not covered by the pattern, and people reach for chain fixes when they see it.
Running the chain validation properly
certutil -verify -urlfetch <file>builds the full chain and verifies it, retrieving the issuing certificates and revocation lists the certificate points at.- Without
-urlfetchit validates against what is already on the machine, which hides exactly the missing-intermediate case you are usually looking for. -sslpolicy <ServerName>enforces the SSL chain policy against a specific name, which is how to reproduce a name mismatch deliberately.- Run it on the server that is failing. Trust is per machine, and a chain that builds on your workstation proves nothing about the mail server.
certutil -repairstore <store> <thumbprint>repairs a key association or updates certificate properties and the key security descriptor. It is not a general-purpose repair.
Why one server fails and another does not
Every part of this check is local state. The intermediate store, the trusted roots, the system clock, the private key and its permissions, and the network path to the revocation endpoints all belong to the individual machine. Two servers holding the same certificate can behave completely differently, and a hardening baseline applied to one group and not another is the usual explanation when the split follows organisational unit boundaries rather than anything technical.
What not to do, and why
| Tempting shortcut | What it actually costs |
|---|---|
| Disable certificate revocation checking | Converts a diagnosable failure into a silent weakness, and the network restriction stays to surprise you elsewhere |
| Drop the required TLS version | Visible to every partner, hard to reverse once something depends on it, and the certificate is still broken |
| Import the leaf into trusted roots | Works on one machine, fails everywhere else, and nobody remembers doing it |
| Turn off validation in the application | Moves the fault out of your logs without moving it out of your estate |
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x80090327 |
SEC_E_CERT_UNKNOWN: an unknown error occurred while processing the certificate. It does not name which check failed, so validate the chain | Microsoft Learn |
0x800B0109 |
CERT_E_UNTRUSTEDROOT: the chain processed correctly but terminated in a root the trust provider does not trust, usually because an intermediate is missing | Microsoft Learn |
0x80090304 |
SEC_E_INTERNAL_ERROR: the Local Security Authority cannot be contacted. This points at the machine’s security subsystem rather than at the certificate | Microsoft Learn |
0x800B0101 |
CERT_E_EXPIRED: a required certificate is not within its validity period, which includes a certificate that is not yet valid | Microsoft Learn |
Confirm the fix worked
certutil -verify -urlfetchbuilds the chain to a trusted root with no errors, run on the server that was failing.Get-ExchangeCertificatereports a valid status and a present private key.- A test session to the affected endpoint completes.
w32tm /query /statusshows the server synchronised against a source you recognise.- The system log shows no new Schannel errors after the change, rather than merely being quiet.
Questions people ask about this
Does fixing this cost anything?
Not in licensing terms. Importing an intermediate, correcting the clock or repairing a key association are all free. A replacement public certificate costs whatever your certificate authority charges, and for purely internal connectors an internally issued certificate is a legitimate alternative.
Why does it work from one server and fail from another?
Because trust is per machine. The intermediate store, the trusted roots, the clock and the private key permissions are all local state, so two servers with the same certificate behave differently.
What does 0x80090304 mean if the certificate is fine?
It means what Windows publishes: the Local Security Authority cannot be contacted. That is the machine’s own security subsystem, not the chain. Chasing certificates on that code wastes an afternoon.
Can I ignore revocation checking to make this go away?
You can, and you should not. It converts a diagnosable failure into a silent weakness, and the underlying network restriction stays in place to surprise you elsewhere.
Is a wildcard certificate a problem here?
Not for chain validation. Wildcards fail for a different reason, when the name being connected to is not covered by the pattern, and that produces a name mismatch code rather than any of these.
