Fix it now
Transport looked for a certificate carrying a particular domain name and did not find one it could use. The name comes from a connector’s FQDN; the certificates it searched are those in the local computer’s personal store that are enabled for SMTP. Until the two agree, that connector cannot offer STARTTLS, and partners that require encryption will start refusing mail.
Get-ExchangeCertificate | Format-List Thumbprint,Subject,CertificateDomains,Services,NotAfter,Status
Get-SendConnector | Format-List Name,Fqdn,TlsCertificateName
Get-ReceiveConnector | Format-List Identity,Fqdn,TlsCertificateName,Bindings
Enable-ExchangeCertificate <thumbprint> -Services SMTP
- Read the event body first and write down the domain name it says it could not find. That name is the whole search key.
- Answer No when
Enable-ExchangeCertificateasks whether to overwrite the default SMTP certificate. Answering yes breaks EdgeSync. - If a certificate does carry the name, enabling it for SMTP is the fix. If none does, decide which is wrong, the connector name or the certificate, before changing either.
- In a hybrid deployment, check
TlsCertificateNameon the receive connector. If it still names a certificate you removed, clear it withGet-ReceiveConnector "SERVER\Default Frontend SERVER" | Set-ReceiveConnector -TlsCertificateName $nulland rerun the Hybrid Configuration Wizard. - Restart the Microsoft Exchange Transport service in a maintenance window, then watch for the event at connector startup.
Do not change the FQDN on the default Receive connector named Default <Server Name> on a Mailbox server. Microsoft warns that with more than one Mailbox server in the organisation, internal mail flow between them fails when you do.
If the event stops appearing and a test message reports an encrypted session, you are done. If not, the next section covers how Exchange picks a certificate at all.
Why it happens
Every send and receive connector has an FQDN. That name is what the server announces in the SMTP banner and the EHLO response, and it is also what transport uses during TLS authentication. When the connector starts, transport searches the local computer’s personal certificate store for a certificate that is enabled for the SMTP service and whose subject or subject alternative names include that FQDN. If nothing matches, you get event 12014 and the connector comes up without STARTTLS.
The important part is that this is a name match, not a thumbprint match, unless TlsCertificateName has been set on the connector explicitly. That is why a certificate renewal breaks things: the replacement is installed and perfectly valid, but the names on it changed, or the old certificate was the one enabled for SMTP and the new one was not. The service is not complaining that a certificate is bad. It is complaining that nothing carries the name it was told to use.
Pinning with TlsCertificateName swaps one failure mode for another. It takes the issuer and subject together in the form "<I>X.500Issuer<S>X.500Subject", so it survives a renewal only if both fields stay identical. Remove the certificate it names and the connector is pointing at nothing, which is the documented cause of a hybrid server losing STARTTLS after a certificate change when the Hybrid Configuration Wizard was not rerun.
The renewed certificate was never enabled for SMTP
You have this one if A certificate with the right names is present and valid, but Services on it does not include SMTP.
- Confirm the thumbprint of the certificate you intend to use.
- Run
Enable-ExchangeCertificate <thumbprint> -Services SMTP, answering No to overwriting the default SMTP certificate. - Confirm with
Get-ExchangeCertificate | Format-List Thumbprint,Services. - Restart transport in a maintenance window and watch for the event at connector startup.
Add IIS to the services list only if that certificate also serves client access, and read the replacement prompt rather than accepting it blindly on a server hosting other sites.
The connector FQDN is a name no certificate carries
You have this one if Every certificate is valid and enabled, but the name in the event appears in no CertificateDomains list.
- Decide which is wrong. Usually the connector is pointing at a namespace that was retired.
- Change the connector with
Set-SendConnectororSet-ReceiveConnectorand the-Fqdnparameter, but never on the default Receive connector namedDefault <Server Name>on a Mailbox server. - If the name is genuinely required, request a replacement certificate carrying it as a subject alternative name.
- Restart transport and confirm the event clears.
On a Mailbox server, changing the FQDN of the Default <Server Name> receive connector breaks internal mail flow between Mailbox servers. Create or correct a different connector instead.
A stale TlsCertificateName is pinning the connector to a certificate that is gone
You have this one if A hybrid server stops offering STARTTLS after a certificate change, and telnet to port 25 shows no STARTTLS verb.
- Enable SMTP on the new certificate first:
Enable-ExchangeCertificate <thumbprint> -Services SMTP. - Clear the stale pin:
Get-ReceiveConnector "SERVER\Default Frontend SERVER" | Set-ReceiveConnector -TlsCertificateName $null. - Rerun the Hybrid Configuration Wizard so the connector is updated with the new certificate information.
This is Microsoft’s documented cause for a hybrid server that cannot receive mail from Microsoft 365 after a certificate is replaced and the wizard was not rerun.
The certificate expired
You have this one if NotAfter is in the past, and the expiry-related transport events appear alongside 12014.
- Install the replacement, keeping the same subject and alternative names wherever possible.
- Enable it for SMTP, and for IIS if it serves the client namespaces.
- Leave the old certificate in place until the new one is confirmed working, then remove it.
- Put the next expiry in a calendar. This failure is entirely predictable and it is the only one on this page that is.
Several certificates match and Exchange picks the wrong one
You have this one if The name exists on more than one SMTP-enabled certificate, and behaviour changed after a renewal without any connector change.
- List every SMTP-enabled certificate and identify the duplicates.
- Pin the connector deliberately:
$TLSCert = Get-ExchangeCertificate -Thumbprint <thumbprint>then$TLSCertName = "<I>$($TLSCert.Issuer)<S>$($TLSCert.Subject)", and set that value on the connector. - Remove genuinely obsolete certificates once nothing references them.
Pinning means the connector needs attention at every renewal. It is the right answer for duplicates and the wrong default for everything else.
Full reference
Matching the event to the connector
| Where the event names the FQDN | Which connector to inspect |
|---|---|
| A public mail domain name | The internet-facing send connector, or the receive connector accepting internet mail |
| An internal server FQDN | The receive connector used for server-to-server traffic on that server |
| A name you no longer use | A connector left behind from an old namespace; correct or remove it |
| A name that is on your certificate | The certificate exists but is not enabled for SMTP, or has expired |
| Nothing recognisable, on a hybrid server | A stale TlsCertificateName pin rather than an FQDN problem |
What the connector FQDN is used for
- The default SMTP banner of the connector.
- The EHLO and HELO response.
- The most recent Received header on incoming mail as it enters the Transport service.
- TLS authentication, which is the use that produces this event.
- Its default value is the FQDN of the Exchange server holding the connector, and that default is deliberately left alone on the
Default <Server Name>connector.
Reading the certificate estate
| Command | What it gives you |
|---|---|
Get-ExchangeCertificate | Format-List Thumbprint,Subject,CertificateDomains,Services,NotAfter,Status |
Everything you need in one view |
Get-ExchangeCertificate | where {$_.Status -eq "Valid" -and $_.IsSelfSigned -eq $false} |
The certificates from a public authority that are currently valid |
Get-ExchangeCertificate | where {$_.IsSelfSigned -eq $true} |
The self-signed certificates Exchange installs for internal use |
Enable-ExchangeCertificate <thumbprint> -Services SMTP |
Grants transport access to the certificate’s key |
Get-ReceiveConnector | Format-List Identity,Fqdn,TlsCertificateName,Bindings |
Which name each connector announces and whether it is pinned |
Restarting transport without causing a second incident
Certificate selection happens when the connector starts, so a restart of the Microsoft Exchange Transport service is what applies the change. That restart stops mail flow on the server while it comes back and can requeue messages in flight. Put the server into maintenance mode first, or do it in a window, and do not do it on impulse because an event log entry annoyed you. The event means the connector cannot offer TLS; it does not mean mail has stopped.
The codes that travel with this one
- Schannel event 36871 reports a fatal error creating a TLS credential. Microsoft’s documented cause includes the SMTP service processing an EHLO when no certificate is assigned, which is the same fault seen from underneath.
451 5.7.3 STARTTLS is required to send mail.is what a receiving Exchange server returns when the session did not establish TLS. It is a consequence of 12014, not a separate problem.- A 530 reply with a 5.7.0 enhanced status is a receiving host refusing the message outright. No Microsoft page defines that pairing for Exchange transport specifically, so read the full text the partner returned rather than the number.
- Events 12016 and 12017 come from the same transport area but have no current published description. Read the event body on your own server; it names the certificate and the FQDN it is unhappy about.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
Event ID 12014 |
Microsoft Exchange could not find a certificate that contains the domain name a connector uses, in the personal store on the local computer | Microsoft Learn |
Event ID 12016 |
Logged by Exchange transport in the same certificate area as 12014. No current vendor description is published; read the event body for the certificate and FQDN it names | not published by the vendor |
Event ID 12017 |
Also logged by transport around certificate selection, with no current vendor description published | not published by the vendor |
451 5.7.3 |
Documented by Microsoft as “451 5.7.3 STARTTLS is required to send mail.” – the receiving side rejecting a session that did not establish TLS | Microsoft Learn |
Event ID 36871 |
Schannel: a fatal error occurred while creating a TLS credential. Microsoft’s documented cause includes SMTP processing an EHLO when no certificate is assigned | Microsoft Learn |
530 5.7.0 |
An SMTP refusal from the receiving host in the security and policy class. No Microsoft page defines this pairing for Exchange transport; read the full text returned | not published by the vendor |
Confirm the fix worked
- Transport restarts without event 12014 reappearing for that connector.
Get-ExchangeCertificate | Format-List Subject,CertificateDomains,Services,NotAftershows the connector FQDN on an SMTP-enabled, unexpired certificate.- A telnet session to port 25 on that connector advertises STARTTLS.
- A test message to an external recipient is reported as encrypted by the receiving side.
- The system log shows no new Schannel errors after the change.
Questions people ask about this
Does this cost anything to fix?
No Microsoft licence changes anything here; it is configuration and certificates. The only possible expense is a replacement certificate from a certificate authority, and an internally issued one is fine for connectors that only serve internal traffic.
Can I use a self-signed certificate?
For internal server-to-server connectors, yes, and Exchange installs one for that purpose. For any connector a partner or a client reaches, no: they will not trust it, and you trade this event for a trust failure.
Is 12014 urgent if mail is still flowing?
It is not an outage, but that connector cannot offer TLS. Mail that should be encrypted is not, and any partner requiring encryption will begin refusing you with a 451 5.7.3. Treat it as important rather than critical.
Why must I not change the FQDN on the default receive connector?
Because Microsoft documents that internal mail flow between Mailbox servers fails when you do, in any organisation with more than one of them. If you need a different name announced, use a different connector.
Do I have to restart transport?
Certificate selection is made when the connector starts, so a restart is the reliable way to apply it. Plan it, because it interrupts mail flow on that server while the service comes back.
