Article ID: 180795 - View products that this article applies to.
This article was previously published under Q180795
This article has been archived. It is offered "as is" and will no longer be updated.
One or more servers within the site fail to notify other site servers of replica updates, and other site servers fail in their attempt to communicate with the affected server or servers.
The problem servers log the following Directory event in the Application Event Log during startup:
1166 Error 1720 has occurred (Internal ID 3230203). Contact Microsoft Technical Support for assistance.
Additionally, other site servers may log the following Directory events when attempting to communicate with the problem servers:
1083 Replication warning: The directory replication agent (DRA) couldn't synchronize naming context /o=<some-naming-context> with naming context on directory <some-servername-or-DSA-Signature> 1059 Internal error: The directory replication agent (DRA) call returned error 13.
During directory service startup, the RPC endpoint mapper failed in its attempt to register an endpoint on behalf of the directory service for a specific protocol. This can occur for the ncacn_ip_tcp protocol when the following conditions are true:
If the network environment requires that the Exchange Services have fixed ports assigned to them, carefully plan port assignments to ensure that the ports are not already in use. See MORE INFORMATION, below. Also see the following Microsoft Knowledge Base article:
(http://support.microsoft.com/kb/176466/EN-US/ )XGEN: TCP Ports and Microsoft Exchange: In-depth Discussion
The Exchange Services' default behavior is to allow the RPC endpoint mapper to dynamically assign a TCP port from the pool of ephemeral (short-lived) ports during service initialization. On Windows NT, by default, this pool begins with port 1024. The result is that the port assigned to a service can be different every time the service starts.
Clients attempting to contact a service first query the RPC endpoint mapper (at port 135) to determine at what port the Exchange Service can be contacted (see 176466
Some network environments (typically involving firewalls) require that the ports used by services remain within a fixed range, or in the extreme, also that they remain consistent every time the service executes.
Port Assignment Options and StrategyNOTE: The UDP protocol also has pools of port numbers. These are separate from the TCP pools.
There are two pools of TCP port numbers to draw from for forcing an Exchange Service to use a particular TCP port: the Well-known Ports pool (0 - 1023); and the Ephemeral pool (1024 and above). Either is satisfactory, and a firewall's configuration may dictate which pool to use. When using either pool, adhere to the following:
Well-known (TCP) Ports (0 - 1023)By convention and for interoperability of standard network services, this pool of addresses was set aside per RFC 1060, and many of these ports are reserved for use by specific services, such as echo, ftp, smtp, ldap, pop3, finger, nntp, snmp, and so on. There are 1024 ports in this pool (not all of them assigned yet), and typically only a small fraction of these services is actually implemented within a particular computing environment. It is acceptable to use ports that are not currently (nor ever will be) hosting the service that the port was originally reserved for, or using unassigned ports.
Ephemeral (TCP) PortsDO NOT pick the first ports within this range (1024 [ASCII 150] 65535). Winsock and RPC assign these ports starting at the lowest in the range. By the time the Exchange Services starts up, some of the first ports may have been already assigned (hence the SYMPTOMS above). Unfortunately there is no comprehensive diagnostic that can report which ports have been assigned by either Winsock or RPC. Running "NETSTAT -a" lists all connections and listening ports, but this may not reveal all assigned ports. Experimentation may be required. If in doubt, consider assigning from the Well-known ports pool, or selecting port numbers above 1023 (for instance, the 1200 range).
Other Core Exchange ServicesNeither the Exchange System Attendant nor the Exchange Information Store service logs an event if an RPC endpoint creation fails during service initialization. The operating behavior and performance of these services without the ncacn_ip_tcp protocol enabled is unpredictable and could vary widely from environment to environment. The MTA logs:
9321 - Interface An RPC communications error occurred. An attempt to listen over RPC has failed. NT Error: 1720. [ncacn_ip_tcp BASE IL MAIN BASE 1 504] (14)
Contact us for more help
Connect with Answer Desk for expert help.