Fix it now
Transport asked DNS where to send a message and did not get a usable answer. It is temporary, so mail sits in Retry rather than bouncing. Read the whole error string first: 451 4.4.0 on its own is Exchange’s general temporary next-hop failure, and it is the trailing text that makes it a DNS fault.
Get-Queue | Where-Object {$_.Status -eq 'Retry'} | Format-List Identity,NextHopDomain,MessageCount,LastError
Get-TransportService <servername> | Format-List ExternalDNSAdapterEnabled,ExternalDNSServers,InternalDNSAdapterEnabled,InternalDNSServers,ExternalDNSProtocolOption,MessageExpirationTimeout
Get-SendConnector | Format-List Name,AddressSpaces,SmartHosts,DNSRoutingEnabled,UseExternalDNSServersEnabled
- Query the destination using the resolvers transport is actually configured with, not the machine defaults:
Resolve-DnsName contoso.com -Type MX -Server <one of the servers listed above> - Compare against a public resolver to see whether the problem is your DNS or their records:
Resolve-DnsName contoso.com -Type MX -Server 1.1.1.1 - If you set external DNS servers, also switch the send connector to use them –
Set-SendConnector 'To Internet' -UseExternalDNSServersEnabled $true. The default is $false, so the list is ignored until you do. - Restart the transport service afterwards for the new DNS settings to take effect:
Restart-Service MSExchangeTransport. Do it in a quiet window, because in-flight SMTP conversations are dropped and retried. - Push the queue rather than waiting for the retry timer:
Retry-Queue -Identity '<server>\<queue>' -Resubmit $true
Set-TransportService does not restart anything. Nothing in the cmdlet reference says it does, and no Exchange Set-* cmdlet restarts a Windows service as a side effect – so the restart above is a step, not a consequence.
If the queue drains and resolution works from the server, you are done. If only one destination domain fails, the next section covers the cause Microsoft documents first, which is not your resolver.
Why it happens
Exchange transport keeps its own DNS configuration, separate from the resolvers on the network adapter. It can be told to use the adapter settings, or given an explicit list of servers for external lookups and another for internal ones. That is deliberate: a server in a private namespace often needs one resolver for its own domain and a different one for the public internet. It is also why an administrator can sit at the console, run a successful lookup by hand, and still have a queue full of DNS failures.
Microsoft documents two causes for 451 4.4.0 DNS query failed in Exchange Server, and the first is not the one people reach for. Where a specific remote domain fails, the documented cause is that the remote DNS servers ignore AAAA queries or return unexpected responses to them, and the documented workaround is a send connector for that domain routed through a smart host set to the IP address of the domain’s MX record, with smart host authentication set to None. Where all remote domains fail, the documented cause is the local DNS server not working properly, and the fix is to add public DNS servers to the External DNS Lookups setting on the server and then tell the send connector to use them.
That second half is the step that decides whether the fix works. Set-SendConnector‘s UseExternalDNSServersEnabled parameter controls whether the connector uses the external DNS list from Set-TransportService, and its documented default is $false. Setting ExternalDNSServers and stopping there leaves the connector resolving exactly as it did before.
Read the code carefully as well. 451 4.4.0 is Exchange’s general temporary next-hop failure, and the same code appears with text about a target address that could not be connected to and a failed failover to an alternate host. That is a connection problem, not a DNS one, and the trailing text is the only thing separating them.
One destination domain fails against every resolver
You have this one if The lookup fails from your resolvers and from a public one, and only that domain is affected.
- Look up the IP address of the MX record for that domain from outside your network.
- Create a send connector with that domain as its address space, routed through a smart host set to that IP address, with smart host authentication set to None. This is Microsoft’s documented workaround for remote DNS servers that mishandle AAAA queries.
- Query the domain’s own name servers to see what they publish. An MX pointing at a host name with no address record is theirs to fix, not yours.
- Warn the senders. Messages retry until the expiration timeout and then bounce, so silence is not delivery.
Transport is pointed at a resolver that cannot answer external queries
You have this one if A lookup against the server listed in the transport configuration fails, while the same query against a public resolver succeeds, and every external domain is affected.
- Confirm the internal DNS servers can resolve internet names at all. If they rely on forwarders, test the forwarders directly.
- Add public DNS servers to the external DNS lookup settings:
Set-TransportService <servername> -ExternalDNSAdapterEnabled $false -ExternalDNSServers <ip>,<ip> - Enable them on the send connector that carries the traffic:
Set-SendConnector 'To Internet' -UseExternalDNSServersEnabled $true - Restart MSExchangeTransport, then recheck the queue. Entries leave Retry once resolution succeeds, or immediately if you resubmit them.
Never replace the internal resolver list with a public one. Internal resolution still has to work for domain membership and for Exchange’s own directory access.
Outbound DNS is blocked or intercepted
You have this one if Queries from the Exchange server fail while the same queries from a workstation succeed, and all external domains are affected.
- Confirm the Exchange server can reach the resolver on the DNS port at all, tested from that server rather than from your desk.
- Check whether a firewall rule permits DNS only from designated servers and the Exchange server is not one of them.
- Look for a security product intercepting DNS on the Exchange server itself.
- Retest after each change so you know which one fixed it.
Large DNS answers are being truncated
You have this one if Simple lookups succeed while domains with large record sets fail intermittently, with no obvious pattern.
- Test the same query forcing TCP and compare with the default behaviour.
- If TCP succeeds where the default fails, something in the path is dropping large DNS responses or blocking DNS over TCP.
Set-TransportService <servername> -ExternalDNSProtocolOption UseTcpOnlyis a legitimate workaround while the network path is repaired; the documented values are Any, UseTcpOnly and UseUdpOnly, and the default is Any.- Fix the network path rather than leaving the workaround permanently, because it adds latency to every lookup.
Full reference
Narrowing it down before you change anything
| Test result | Conclusion |
|---|---|
| Lookup fails against every resolver you try, one domain only | Their records are broken, or their DNS mishandles AAAA queries |
| Lookup works publicly, fails against your configured resolver | Your DNS server or its forwarders |
| Lookup works, a port 25 test fails | DNS is fine; this is a network path or firewall problem |
| Only some domains fail | Selective resolution failure, often a broken forwarder or DNSSEC validation |
| Everything external fails, internal is fine | The resolver used for external lookups cannot perform internet resolution |
| The error text is not ‘DNS query failed’ | Not a DNS fault. 451 4.4.0 is also emitted for connection failures |
The transport DNS settings that matter
| Parameter | What it controls | Default |
|---|---|---|
ExternalDNSAdapterEnabled |
Whether external lookups use the network adapter’s DNS settings | $true |
ExternalDNSServers |
The explicit list of external DNS servers, by IP address | Empty |
InternalDNSAdapterEnabled |
Whether internal lookups use the adapter’s DNS settings | $true |
InternalDNSServers |
The explicit list of internal DNS servers | Empty |
ExternalDNSProtocolOption |
Protocol for external DNS queries: Any, UseTcpOnly or UseUdpOnly | Any |
MessageExpirationTimeout |
How long a message may stay in a queue before it is returned as a permanent failure | 2 days |
UseExternalDNSServersEnabled (on the send connector) |
Whether the connector uses the external DNS list at all | $false |
The last row is the one that wastes days. Setting ExternalDNSServers on the transport service changes nothing for a send connector until UseExternalDNSServersEnabled is set to $true on that connector. Both, then restart the transport service.
Restarting transport, deliberately
Set-TransportService changes the configuration and nothing else – it does not restart the Microsoft Exchange Transport service, and the cmdlet reference makes no such claim. The new DNS settings do not take effect until the service is restarted, so Restart-Service MSExchangeTransport is part of the procedure. Do it in a quiet window: in-flight SMTP conversations are dropped and the sending servers retry them, which is safe but noisy, and the queues on that server are unavailable for the duration.
The codes you will see alongside it
| Code | Status |
|---|---|
| 451 4.4.0 with ‘DNS query failed’ | Documented by Microsoft for Exchange Server, with two causes and two fixes |
| 451 4.4.0 with other trailing text | The same general temporary next-hop failure, usually a connection problem |
| 550 5.4.310 | Not present in Microsoft’s Exchange Online non-delivery report reference |
| 550 5.4.312 | Not present in Microsoft’s non-delivery report reference |
| 550 5.4.316 | Microsoft publishes 4.4.316 as ‘Connection refused [Message=Socket error code 10061]’ – a network connection failure at the external server, not an expiry |
| 4.4.7 | ‘Message expired’ – this, not 5.4.316, is the documented code for a message that ran out of queue lifetime |
How long you have
Messages retry until MessageExpirationTimeout is reached and are then returned as a permanent failure. The documented default is 2 days. Read your own value rather than assuming it has been left alone, and if a genuine outage is going to outlast it, extend it deliberately and put it back afterwards rather than discovering a morning’s worth of non-delivery reports.
Checks that are worth doing once
- Resolve the destination from the Exchange server, using the resolvers transport is configured with, not the ones in the network adapter.
- Test port 25 to the resolved host from the Exchange server; a name that resolves and a host you cannot reach look identical from the queue.
- Check whether the failing domains have anything in common – one DNS provider, one country, one record type.
- Confirm the send connector carrying that traffic is the one you are editing; several internet-facing connectors is normal.
- After any change, resubmit rather than waiting, so you find out now whether it worked.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
451 4.4.0 |
Followed by ‘DNS query failed’, the DNS lookup for the next hop failed. The same 451 4.4.0 with different trailing text is a connection failure, not a DNS one | Microsoft Learn |
550 5.4.310 |
Not present in Microsoft’s non-delivery report reference. Treat it as a permanent next-hop DNS failure for the destination domain and test resolution rather than trusting the number | not published by the vendor |
550 5.4.312 |
Not present in Microsoft’s non-delivery report reference. Diagnose it by testing resolution for the destination from the server that queued the message | not published by the vendor |
550 5.4.316 |
Microsoft publishes 4.4.316 as connection refused with socket error code 10061 – a network connection failure at the external server. It is not a message expiry code; 4.4.7 is | Microsoft Learn |
Confirm the fix worked
Resolve-DnsName <destination> -Type MX -Server <the server transport uses>returns records.Get-SendConnectorshowsUseExternalDNSServersEnabledset the way you intended on the connector carrying the traffic.- The transport service has been restarted since the DNS settings changed.
Get-Queueshows the affected queue Active or Ready rather than Retry, and the message count falling.- A test message to the affected domain is delivered and appears with a SEND event in the tracking log.
Questions people ask about this
Will buying anything fix this?
No. This is name resolution and it costs nothing to correct. Either your resolvers need fixing, a firewall rule needs adjusting, or the destination’s own DNS is broken and only they can repair it.
Does Set-TransportService restart the transport service?
No. It changes configuration only. Restart MSExchangeTransport yourself for new DNS settings to take effect, and do it in a quiet window because in-flight SMTP conversations are dropped and retried.
I set ExternalDNSServers and nothing changed. Why?
Because a send connector only uses that list when UseExternalDNSServersEnabled is $true, and its default is $false. Set it on the connector that carries the traffic, then restart transport.
Only one domain fails. Is that my DNS?
Probably not. Microsoft’s first documented cause for this error is remote DNS servers that ignore AAAA queries or answer them unexpectedly, and the documented workaround is a send connector for that domain routed through a smart host set to the IP address of their MX record.
How long before queued messages bounce?
They retry until MessageExpirationTimeout is reached and are then returned as a permanent failure. The documented default is 2 days; read your own value with Get-TransportService.
